Analysis and interpretation of data IDSP training module for state and district surveillance officers Module 9
Jan 03, 2016
Analysis and interpretation of data
IDSP training module for state and district surveillance officers
Module 9
Learning objectives
• Identify the role, importance and techniques of data analysis
• Sources and management of data for valid conclusions
• Choose appropriate descriptive and analytical methods
• List outcome measures for feedback• Generate reports with tables and graphs
All levels must analyze surveillance data
• Health workers Increase of cases
• Medical officers in primary health centres Outbreak detection Seasonal trends
• District surveillance officers All of the above Advanced analyses
Selected outcomes of data analysis
• Identification of outbreaks / potential outbreaks• Identification of appropriate and timely control
measures• Prediction of changes in disease trends over time • Identification of problems in health systems• Improvement of the surveillance system
through: Identification of regional differences Identification of differences between the private and
the public sectors
• Identification of high-risk population groups
Sources of data
• Sub-Centre• Primary health centre• Community health centre• District• Private practitioners• Private nursing homes• Identified laboratories • Medical colleges• Police departments• State
Types of data
• Syndromic case data• Presumptive case data• Confirmed case data• Sentinel case data• Regular surveillance data• Urban data• Rural data
Periodicity of data collection
• Weekly• High priority (Acute flaccid paralysis)
As soon as a case is detected
• Data on outbreaks are collected and analyzed separately
Analysis of data at the district surveillance unit
• Computer software provides ready outputs• District surveillance officer prepares a report• Technical committee reviews and needs to
bear in mind: The strength and weakness of data collection
methods? Reliability and validity of data
The separate disease profiles The user-friendliness of graphs The need to calculate rates before comparisons
What computers cannot do
Skills• Contact reporting
units for missing information
• Interpret laboratory tests
• Make judgment about: Epidemiologic linkage Duplicate records Data entry errors
• Declare a state of outbreak
Attitudes• Looking• Thinking • Discussing• Taking action
Expressed concerns versus reality
Concerns commonly expressed
• Statistics are difficult• Multivariate analysis
is complex• Presentation of data
is challenging
Mistake commonly observed
• Data are not looked at
Basic surveillance data analysis
1. Count, divide and compare Direct comparisons between number of
cases are not possible in the absence of the calculation of the incidence rate
2. Descriptive epidemiologyA. TimeB. Place C. Person
1. Count, Divide and Compare (CDC)
• Count Count cases that meet the case definition
• Divide Divide cases by the population denominator
• Compare Compare rates across:
• Age groups• Districts• Etc.
2. Time, place and person descriptive analysis
A. Time Graph over time
B. Place Map
C. Person Breakdown by age, sex or personal
characteristics
A. Analysis over time
• Absolute number of cases Does not allow comparisons Analysis by week, month or year
• Incidence Allows comparisons Analysis by week, month or year
Acute hepatitis (E) by week, Hyderabad, AP, India, March-June
2005
0
20
40
60
80
100
120
1 8 15 22 29 4 12 19 26 3 10 17 24 31 7 14 21 28
Num
ber
of
case
s
March April May June First day of week of onset
Interpretation: The source of infection is persisting and continues to cause cases
Absolute number of cases per week
Reported varicella and typhoid cases, Darjeeling district, West Bengal, India,
2000-4Figure 3: Reported varicella and typhoid cases, Darjeeling
district, WB, India, 2000-2004
1
10
100
1000
10000
100000
2000 2001 2002 2003 2004
Years
Number of cases (Log)
Typhoid
Varicella
Interpretation: The parallel increase between varicella (that should be constant) and typhoid suggests that increasing
rates of typhoid are secondary to improved reporting
Incidence by year
2. Analysis by place
• Number of cases by village or district Does not control for population size Spot map
• Incidence of cases by village or district Controls for population size Incidence map
Mangalore
Nallur
Vridha-chalam
Kattumannar Kail Kumaratchi
Parangipattai
Kamma-puram
Panruti
Cuddalore
Annagraman
Kurinjipadi
Bhuvanagiri
Keerapalayam
Interpretation: Cases were reported from tsunami affected
non-affected areas, thus the cluster was not a consequence of
the tsunami
Reported cases of measles, Cuddalore district, Tamil Nadu, Dec
2004 – Jan 2005
Spot map of absolute number of cases
20-49
50-99
100+
1-19
0
Attack rate per100,000 population
Pipeline crossing open sewage drain
Open drain
Incidence of acute hepatitis (E) by block, Hyderabad, AP, India, March-
June 2005
Interpretation: Blocks with hepatitis are those supplied by pipelines
crossing open sewage drains
Incidence by area
3. Analysis per person
• Distribution of cases by: Age Sex Other characteristics
(e.g., Ethnic group, vaccination status)
• Incidence by: Age Sex Other characteristics
81%
19%
Immunized Unimmunized
Immunization status of probable measles cases, Nai, Uttaranchal,
India, 2004
Interpretation: The outbreak is probably caused by a failure to vaccinate
Distribution of cases according to a characteristic
Probable cases of cholera by age and sex, Parbatia, Orissa, India,
2003Number of cases Population Incidence
0 to4 6 113 5.3%5 to14 4 190 2.1%15 to24 5 128 3.9%25 to34 5 144 3.5%35 to44 6 129 4.7%45 to54 4 88 4.5%55 to64 8 67 11.9%
Age group(In years)
> 65 3 87 3.4%Male 17 481 3.5%SexFemale 24 465 5.2%
Total Total 41 946 4.3%
Interpretation: Older adults and women are at increased risk of cholera
Incidence according to a characteristic
Seven reports to be generated
1. Timeliness/completeness2. Description by time, place and person3. Trends over time4. Threshold levels5. Compare reporting units6. Compare private / public7. Compare providers with laboratory
Report 1: Completeness and timeliness
• A report is said to be on time if it reaches the designated level within the prescribed time period Reflects alertness
• A report is said to be complete if all the reporting units within its catchment area submitted the reports on time Reflects reliability
Interpretation of timeliness and completeness
Scenario Interpretation
Reporting unit A is timely and complete
•Ideal
Reporting unit B timely but regularly incomplete
•Medical officer of B understands the importance•Sort out problem of non reporting sites
Reporting unit C is late but complete
•Medical officer C don’t understand the importance of timeliness. He needs to be educated
Reporting unit D is late and incomplete
•Major problem. Urgent action required
Report 2: Weekly/ monthly summary report
• Based upon compiled data of all the reporting units
• Presented as tables, graphs and maps• Takes into account the count, divide
and compare principle: Absolute numbers of cases and deaths are
sufficient for a single reporting unit level Incidence rates are required to compare
reporting units
Epidemiological indicators to use in weekly / monthly summary report
• Cases• Deaths• Incidence rate• Case fatality ratio
Report 3: Comparison with previous weeks/ months/ years
• Help detect trend of diseases over time• Weekly analysis compare the current
week with data from the last three weeks Alerts authorities for immediate action
• Monthly and yearly analysis examine: Long term trends Cyclic pattern Seasonal patterns
Acute hepatitis by week of onset in 3 villages, Bhimtal block, Uttaranchal,
India, July 2005
0
10
20
30
40
50
60
70
80
901s
t w
eek
2nd
w
eek
3rd
we
ek
4th
wee
k
1st
wee
k
2nd
w
eek
3rd
we
ek
4th
wee
k
1st
wee
k
2nd
w
eek
3rd
we
ek
4th
wee
k
1st
wee
k
2nd
w
eek
3rd
we
ek
4th
wee
k
1st
wee
k
May June July August September
Week of onset
Num
ber
of c
ases
Interpretation: The second week of July has a clear excess in the number of cases, providing an early warning signal for the
outbreak
Example of weekly analysis
Malaria in Kurseong block, Darjeeling District, West Bengal, India, 2000-
2004
0
5
10
15
20
25
30
35
40
45
Janu
ary
Feb
ruar
y
Mar
chA
pril
May
June
July
Aug
ust
Sep
tem
ber
Oct
ober
Nov
embe
r
Dec
embe
rJa
nuar
yF
ebru
ary
Mar
chA
pril
May
June
July
Aug
ust
Sep
tem
ber
Oct
ober
Nov
embe
r
Dec
embe
r
Janu
ary
Feb
ruar
yM
arch
Apr
ilM
ayJu
neJu
lyA
ugus
t
Sep
tem
ber
Oct
ober
Nov
embe
r
Dec
embe
rJa
nuar
y
Feb
ruar
yM
arch
Apr
ilM
ay
June
July
Aug
ust
Sep
tem
ber
Oct
ober
Nov
embe
rD
ecem
ber
Janu
ary
Feb
ruar
yM
arch
Apr
il
May
June
July
Aug
ust
Sep
tem
ber
Oct
ober
Nov
embe
rD
ecem
ber
2000 2001 2002 2003 2004
Months
Inci
denc
e of
mal
aria
per
10,
000 Incidence of malaria
Incidence of Pf malaria
Example of monthly and yearly analysis
Interpretation: There is a seasonality in the end of the year and a trend towards increasing incidence year after year
Report 4: Crossing threshold values
• Comparison of rates with thresholds• Thresholds that may be used:
Pre-existing national/international thresholds
Thresholds based on local historic data • Monthly average in the last three years
(excluding epidemic periods)
Increasing trends over a short duration of time (e.g., Weeks)
Report 5: Comparison between reporting units
• Compares Incidence rates Case fatality ratios
• Reference period Current month
• Sites concerned Block level and above
Interpretation of the comparison between reporting units
Scenario Interpretation
Incidence rate and case fatality ratio in various reporting units are similar
•May be indicative of good reporting mechanism
Markedly low incidence rate and case fatality ratio in a reporting unit
•Quality of data needs review•Possibility of under-reporting
Markedly high incidence rate and case fatality ratio in a reporting unit
•Quality of data needs review•Possibility of an outbreak•Possibility of a data entry error
Report 6: Comparison between public and private sectors
• Compare trends in incidence of new cases/deaths Incidences are not available for private
provider since no population denominators are available
• Good correlation may imply: The quality of information is good Events in the community are well represented
• Poor correlation may suggest: One of the data source is less reliable
Report 7: Comparison of reports between the public health system
and the laboratory
Elements to compare
Public health system
Laboratories
Validation of reporting
•Number of cases seen by providers
•Number of laboratory diagnoses
Water borne disease
•Cases of diarrheal diseases
•Water quality
Vector borne disease
•Cases of vector borne diseases
•Entomological data
Frequency of reports and analysisReports Daily Weekly Monthly Yearly
Report 1:Timeliness/completeness
Report 2: Description
Report 3:Trends over time
Report 4:Threshold levels
Report 5:Compare reporting units
Report 6:Compare private / public
Report 7:Compare with laboratory
Review of analysis results by the technical committee
• Meeting on a fixed day of every week• Review of a minimum of:
4 reports weekly 7 reports monthly
• Review by disease wise • Search for missing values• Check the validity• Interpret • Prepare summary reports and share• Take action
Limitations in analysis of surveillance data
• The quality of data may be problematic Poor use of case definition Under-reporting
• There may be a time lag between detection, reporting and analysis
• Under-reporting occurs However, if the level of under-reporting is constant,
trends may still be analyzed and outbreaks may still be detected
• The representativeness may be poor Engage the private sector to diversify reporting sources
Conclusion
• Analysis is a major component of surveillance – links data collection and program implementation
• While it is important to analyze data, its also important that analyzed reports are sent to the appropriate authorities Higher level Lower level
Points to remember
• Surveillance data identifies outbreaks and describe conditions by time, place and person
• Surveillance helps monitor disease control and assess the impact of services
• Data analysis must occur at each level• Analyzed data is presented in tables,
graphs with comparisons with previous data