Validation of a Novel Methodology to Estimate Physical Activity and Sedentary Behavior in Free-Living People The Sojourn Method University of Massachusetts Amherst, USA Kate Lyden, Sarah Kozey-Keadle, John Staudenmayer, Patty Freedson
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Validation of a Novel Methodology to Estimate Physical
Activity and Sedentary Behavior in Free-Living People
The Sojourn Method
University of Massachusetts Amherst USA
Kate Lyden Sarah Kozey-Keadle John Staudenmayer Patty Freedson
2 Physical Activity and Health Laboratory
Advanced Statistical Methods for Objective Monitor Data
Adaptive modeling systems
Neural Networks
Hidden Markov models
Support vector machines
Decision tree analysis
o Capable of learning the shape of complex data
o Does not assume a simple parametric relationship
Mannini et al 2010
Countsmin-1
ME
Ts
3 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Lab-Nnet METs
Input Output
Staudenmayer et al (2009)
1 Lag 1-Autocorrelation
o Temporal dynamics of 1-minutersquos worth of second-by-second counts
bull Tests the relationship between adjacent observations (counts)
2 1-minute frequency distribution of counts
o 10th 25th 50th 75th and 90th percentiles of minutersquos second-by-second accelerometer counts
4 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Staudenmayer et al (2009) Freedson et al (2011)
rMSE = 190 METs
Lab-Nnet METs
Input Output
1 Temporal Dynamics
2 Distribution of Counts
Model rMSE (METs)
Lab-Nnet 122
Crouter 161
Swartz 177
Freedson 209
Trained on gt 400 participants
o Wide range of sporting lifestyle and locomotion activities
5 Physical Activity and Health Laboratory
Hours(meanplusmnSD)
NNet
DirectObservation
Sedentary(lt15METs) 00plusmn00 60plusmn18Light(15-299METs) 81plusmn04 26plusmn14
Moderate(3-599METs) 15plusmn04 09plusmn05Vigorous(ge6METs) 02plusmn02 03plusmn04
MET-Hours 246plusmn16 174plusmn45
significantlydifferentthanDO
Identifies no sedentary time
Overestimates light and moderate intensity activity
Overestimates MET-Hours per day
Need to refine Lab-Nnet for use in free-living settings
1 Free-living activities ldquolookrdquo different than activities performed during laboratory calibrations
2 Free-living activities are performed for varying amounts of time
Free-Living Validation Early Results
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
2 Physical Activity and Health Laboratory
Advanced Statistical Methods for Objective Monitor Data
Adaptive modeling systems
Neural Networks
Hidden Markov models
Support vector machines
Decision tree analysis
o Capable of learning the shape of complex data
o Does not assume a simple parametric relationship
Mannini et al 2010
Countsmin-1
ME
Ts
3 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Lab-Nnet METs
Input Output
Staudenmayer et al (2009)
1 Lag 1-Autocorrelation
o Temporal dynamics of 1-minutersquos worth of second-by-second counts
bull Tests the relationship between adjacent observations (counts)
2 1-minute frequency distribution of counts
o 10th 25th 50th 75th and 90th percentiles of minutersquos second-by-second accelerometer counts
4 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Staudenmayer et al (2009) Freedson et al (2011)
rMSE = 190 METs
Lab-Nnet METs
Input Output
1 Temporal Dynamics
2 Distribution of Counts
Model rMSE (METs)
Lab-Nnet 122
Crouter 161
Swartz 177
Freedson 209
Trained on gt 400 participants
o Wide range of sporting lifestyle and locomotion activities
5 Physical Activity and Health Laboratory
Hours(meanplusmnSD)
NNet
DirectObservation
Sedentary(lt15METs) 00plusmn00 60plusmn18Light(15-299METs) 81plusmn04 26plusmn14
Moderate(3-599METs) 15plusmn04 09plusmn05Vigorous(ge6METs) 02plusmn02 03plusmn04
MET-Hours 246plusmn16 174plusmn45
significantlydifferentthanDO
Identifies no sedentary time
Overestimates light and moderate intensity activity
Overestimates MET-Hours per day
Need to refine Lab-Nnet for use in free-living settings
1 Free-living activities ldquolookrdquo different than activities performed during laboratory calibrations
2 Free-living activities are performed for varying amounts of time
Free-Living Validation Early Results
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
3 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Lab-Nnet METs
Input Output
Staudenmayer et al (2009)
1 Lag 1-Autocorrelation
o Temporal dynamics of 1-minutersquos worth of second-by-second counts
bull Tests the relationship between adjacent observations (counts)
2 1-minute frequency distribution of counts
o 10th 25th 50th 75th and 90th percentiles of minutersquos second-by-second accelerometer counts
4 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Staudenmayer et al (2009) Freedson et al (2011)
rMSE = 190 METs
Lab-Nnet METs
Input Output
1 Temporal Dynamics
2 Distribution of Counts
Model rMSE (METs)
Lab-Nnet 122
Crouter 161
Swartz 177
Freedson 209
Trained on gt 400 participants
o Wide range of sporting lifestyle and locomotion activities
5 Physical Activity and Health Laboratory
Hours(meanplusmnSD)
NNet
DirectObservation
Sedentary(lt15METs) 00plusmn00 60plusmn18Light(15-299METs) 81plusmn04 26plusmn14
Moderate(3-599METs) 15plusmn04 09plusmn05Vigorous(ge6METs) 02plusmn02 03plusmn04
MET-Hours 246plusmn16 174plusmn45
significantlydifferentthanDO
Identifies no sedentary time
Overestimates light and moderate intensity activity
Overestimates MET-Hours per day
Need to refine Lab-Nnet for use in free-living settings
1 Free-living activities ldquolookrdquo different than activities performed during laboratory calibrations
2 Free-living activities are performed for varying amounts of time
Free-Living Validation Early Results
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
4 Physical Activity and Health Laboratory
Neural Network Laboratory Calibration and Validation
Staudenmayer et al (2009) Freedson et al (2011)
rMSE = 190 METs
Lab-Nnet METs
Input Output
1 Temporal Dynamics
2 Distribution of Counts
Model rMSE (METs)
Lab-Nnet 122
Crouter 161
Swartz 177
Freedson 209
Trained on gt 400 participants
o Wide range of sporting lifestyle and locomotion activities
5 Physical Activity and Health Laboratory
Hours(meanplusmnSD)
NNet
DirectObservation
Sedentary(lt15METs) 00plusmn00 60plusmn18Light(15-299METs) 81plusmn04 26plusmn14
Moderate(3-599METs) 15plusmn04 09plusmn05Vigorous(ge6METs) 02plusmn02 03plusmn04
MET-Hours 246plusmn16 174plusmn45
significantlydifferentthanDO
Identifies no sedentary time
Overestimates light and moderate intensity activity
Overestimates MET-Hours per day
Need to refine Lab-Nnet for use in free-living settings
1 Free-living activities ldquolookrdquo different than activities performed during laboratory calibrations
2 Free-living activities are performed for varying amounts of time
Free-Living Validation Early Results
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
5 Physical Activity and Health Laboratory
Hours(meanplusmnSD)
NNet
DirectObservation
Sedentary(lt15METs) 00plusmn00 60plusmn18Light(15-299METs) 81plusmn04 26plusmn14
Moderate(3-599METs) 15plusmn04 09plusmn05Vigorous(ge6METs) 02plusmn02 03plusmn04
MET-Hours 246plusmn16 174plusmn45
significantlydifferentthanDO
Identifies no sedentary time
Overestimates light and moderate intensity activity
Overestimates MET-Hours per day
Need to refine Lab-Nnet for use in free-living settings
1 Free-living activities ldquolookrdquo different than activities performed during laboratory calibrations
2 Free-living activities are performed for varying amounts of time
Free-Living Validation Early Results
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
6 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produce physical activity estimates in fixed time intervals
o Assumes each interval consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
7 Physical Activity and Health Laboratory
Limitations of Laboratory Calibrations
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
8 Physical Activity and Health Laboratory
Produces physical activity estimates in minute intervals
o Assumes each minute consists of a single activity D
esc
en
d
Sta
irs
Wal
k
Sit
Fai
rly
Sti
ll
Sta
nd
wit
h M
ino
r M
ove
me
nt
Sit
Fai
rly
Sti
ll
Limitations of Laboratory Calibrations
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
9 Physical Activity and Health Laboratory
Purpose
1 To refine the Lab-Nnet to estimate physical activity and sedentary behavior from free-living accelerometer data
The Sojourn Method
a Identify where bouts of activity and inactivity start and stop
b Improve estimates of sedentary behavior
o Sojourn 1-axis (Soj-1x)
bull Vertical Axis
o Sojourn 3-axis (Soj-3x)
bull Vertical anterior-posterior medial-lateral
2 To validate the Lab-Nnet Soj-1x and Soj-3x in a free-living setting
o Criterion Measure Direct Observation
3 To determine the sensitivity of the Lab-Nnet Soj-1x and Soj-3x to detect change habitual activity o Free-living o Three 7-day conditions
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
10 Physical Activity and Health Laboratory
Experimental Procedures Overview
Aim 2 Sensitivity to Change
7-Day Sedentary Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
Aim 1 Refinement of Lab-Nnet N = 6 Aim 2 Performance of algorithms compared to DO N = 7
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
N = 13
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
11 Physical Activity and Health Laboratory
Methods
Measurement
ActiGraph GT3X Accelerometer
o Right hip
o 1-second epochs
o Normal frequency mode
o Vertical Anterior-Posterior Medial-Lateral Axes
Omron Pedometer
o Left hip
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
12 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
o No structured or leisure physical activity
o Limited occupational activity
o Limited time standingwalking
o lt5000 steps per day
Sedentary Condition
Individuals not meeting the PA guidelines
Population researchers might target in an intervention study
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
13 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines
Reflect distinct behavior patterns used in population and surveillance research
Moderately Active Condition
Individuals sufficiently meeting the PA guidelines
Population that performs just enough PA to improve health
o At least 150 minutes of moderate activity or 75 minutes of vigorous activity
o No more than 200 minutes of moderate activity or 100 minutes of vigorous activity
o Structured exercise performed on 5 of the 7 days
o Maintain lifestyle activity
o 8000-10000 steps per day
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
14 Physical Activity and Health Laboratory
7-Day Conditions
Based on the Physical Activity Guidelines recommendation of 150 minutes per week of moderate intensity activity
Reflect distinct behavior patterns used in population and surveillance research
Very Active Condition
Individuals sufficiently meeting the PA guidelines
Population that meets the PA guidelines by at least twice as much as the minimum recommendation
o At least 300 minutes of moderate activity
o No maximum amount of activity
o Increase lifestyle activity
o Limit time sitting
o At least 12000 steps per day
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
15 Physical Activity and Health Laboratory
Methods Criterion Measure
Direct Observation [Noldus Information Technology Netherlands]
Participants were directly observed in their free-living environment o 3 times for approximately 10 consecutive hours each
Hand-held PDA with focal sampling and duration coding
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
16 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 2 Performance of Lab-Nnet Soj-1x and Soj-3x compared to DO
o Repeated measures linear mixed model
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
1 Bias = (estimate ndash criterion)N
o Precision = 95 CI of bias
bull CI spans zero estimate not significantly different than DO (plt005)
2 rMSE = radic(mean square error)
3 Correlation
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
17 Physical Activity and Health Laboratory
Statistical Evaluation
Aim 3 Sensitivity of Soj-1x and Soj-3x to detect change
o Repeated measured linear mixed model with likelihood ratio testing
bull Likelihood ratio test ndash If addition of condition resulted in significantly better fit the algorithms detected change
a MET-Hrs
b Time in sedentary light moderate vigorous and MVPA
c Qualifying minutes ndash ge moderate intensity ge 10 minutes
d Breaks from sedentary time
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
18 Physical Activity and Health Laboratory
Results
Participant Characteristics (mean plusmn SD)
N = 7
Age (yrs) 250 plusmn 49
Body Mass (kg) 710 plusmn 145
Height (cm) 1713 plusmn 92
BMI (kgm-2) 240 plusmn 24
PAS 64 plusmn 05
BMI=Body Mass Index PAS=Physical Activity Status
Participant Characteristics (mean plusmn SD)
N = 13
Age (yrs) 248 plusmn 52
Body Mass (kg) 682 plusmn 131
Height (cm) 1685 plusmn 106
BMI (kgm
-2) 238 plusmn 19
PAS 64 plusmn 07
BMI=Body Mass Index PAS=Physical Activity Status
Aim 2
Aim 3
Aim 1
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
19 Physical Activity and Health Laboratory
1 Identify 5 patterns in accelerometer output
o Identifies departures or sojourns from zero
2 Determine if bout is activity or inactivity
o Pattern of zero and non-zero counts
3 Estimate METs for activities using Lab-Nnet
4 Assign METs to inactivities
o Compendium of physical activities
o Calibration study
Free-Living Accelerometer Output
1 Only Zeros
2 Alternating zeros and non-zeros
3 Rhythmic non-zeros
4 Alternating pattern of non-zeros
5 Short non-zeros
Co
un
ts p
er
Se
con
d
Minutes
Aim 1 Soj-1x and Soj-3x
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
20 Physical Activity and Health Laboratory
Soj-1x vs Soj-3x
4 Assign METs to inactivities ndash 4 types
a Sitting or lying still
b Sitting with minimal movement
c Standing still
d Standing with minimal movement
Soj-1x
o To distinguish sitting and standing
bull non-zero counts from vertical axis
bull Duration of bout
Soj-3x
o To distinguish sitting and standing
bull Neural network
bull Inputs from vertical anterior-posterior and medial-lateral axes
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
21 Physical Activity and Health Laboratory
Aim 2 Results Bias
(Estimatendash Criterion)N
Lab-nnet
Soj-1x
Soj-3x
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
22 Physical Activity and Health Laboratory
Lab-Nnet
Soj-1x
Soj-3x
068
077
091
537
501
262
Correlation rMSE (Min)
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
055
075
086
550
497
276
Correlation rMSE (Min)
Sedentary Light
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
23 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
079
091
093
54
10
11
Correlation rMSE (MET-Hrs)
Lab-Nnet
Soj-1x
Soj-3x
063
098
095
455
40
78
Correlation rMSE (Min)
MET-Hrs MVPA
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
24 Physical Activity and Health Laboratory
Aim 2 Results Correlation and rMSE
Lab-Nnet
Soj-1x
Soj-3x
-
099
096
-
14
73
Correlation rMSE (Min)
Lab-Nnet
Soj-1x
Soj-3x
-
075
084
-
121
61
Correlation rMSE (Breaks per day)
Qualifying Minutes Breaks
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
25 Physical Activity and Health Laboratory
Aim 3 Results
Aim 2 Sensitivity to Change
N = 13
7-Day Sedentary
Condition
7-Day Moderately Active Condition
7-Day Very Active Condition
10-Hours Direct Observation
10-Hours Direct Observation
10-Hours Direct Observation
Evaluate Soj-1x and Soj-3x only
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
26 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition
Tim
e
700
649 583
704 648
588 234
248
273
245 252
264
Sedentary
Moderately Active
Very Active
520
812
1171
391
788
1218
Min
ute
s
ME
T-H
ou
rs
198 227
270
182
223
276
Sedentary Light
T
ime
MVPA MET-Hours
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
27 Physical Activity and Health Laboratory
Aim 3 Results Mean estimates by condition M
inu
tes
108
379
708
61
425
828
Bre
aks
pe
r D
ay 556 548 559
385 401 410
Sedentary
Moderately Active
Very Active Qualifying Minutes Breaks
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
28 Physical Activity and Health Laboratory
MET-Hrs per Condition Compared to Direct Observation
1013 participants increased MET-Hours from Sedentary to Moderately to Very Active
o Soj-1x ndash Identified 90
o Soj-3x ndash Identified 90
313 did not increase MET-Hours as expected
o Soj-1x ndash identified 667
o Soj-3x ndash identified 100
Assume participants were compliant with conditions
o Expect variability across days for a participant
o Expect variability between participants within a given condition
bull Valid tool will detect change when it has occurred and will remain stable when it has not
Sojourn methods are sensitive to change and remain stable when no change has occurred
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
29 Physical Activity and Health Laboratory
Summary and Discussion
Methods developed in the laboratory do not translate to free-living o Need free-living calibration and validation studies
Sojourn Methods improved free-living physical activity estimates
o MET-hours
o MVPA
o Sedentary ndash Sojourn-3x
bull Sedentary and light most often grouped into ldquolow intensityrdquo category
o Qualifying minutes
o Breaks
Sensitive to change in habitual activity
o Detect change subsequent to an intervention
o Distinguish activity levels in surveillance research
One ActiGraph Accelerometer
Open Source Statistics Environment ndash R
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
30 Physical Activity and Health Laboratory
Sojourn Methods Future Directions
Next Steps
o Activity type
bull Locomotion sport lifestyle sedentary
o Duration and frequency of sedentary and active bouts
o Validate in different populations
bull Older individuals children overweight
o Raw acceleration
bull 32-100 Hz
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
31 Physical Activity and Health Laboratory
Acknowledgements
Physical Activity and Health Laboratory
Patty Freedson PhD
John Staudenmayer PhD
Dinesh John PhD
Sarah Kozey PhD
Jeffer Sasaki MS
Amanda Libertine
Cori Oliver
Mari Mavillia
Natalia Petruski
Funding Source
NHLBI 1RC1 HL099557-01
Participants
32 Physical Activity and Health Laboratory
Thank-You
32 Physical Activity and Health Laboratory
Thank-You
Related Documents
