Transcript
Dental Risk PreditorKeeping Appointments, Keeping Happy Smiles
Maki S. KoyamaInsight Data Science
PROBLEM
SOLUTION
• Dentists suffer cancellations/no show-up, reducing practice efficiency and profits
• No perceived need in patients
• Educate patients and fill a gap between the perceived need and the actual-need
• Tools/models to estimate dental risks – the probability of tooth loss
Data Features Outcomes Binary Classification
Analysis
US National Health & Nutrition Survey
Age: 20-80yr
Cross-sectional
7 features:1) Age2) Sex3) Insurance 4) Annual Income5) Smoking6) Drinking7) Avoided food
Permanent tooth present or absent
1) Logistic Reg2) Naïve Bayes3) Random Forest
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160
20
40
60
80
100
Tooth Location
Per
cent
age
%
Tooth Loss Percentage (%)
Wisdom Teeth
Upper
Lower
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 320
20
40
60
80
100
Tooth Location
Per
cent
age
%
Wisdom Teeth
Cross-Validation: LR AUC SummaryWisdom Teeth
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 160.0
0.2
0.4
0.6
0.8
1.0
Tooth Location
Pro
babi
lity
The Upper Jaw
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 320.0
0.2
0.4
0.6
0.8
1.0
Tooth Location
Pro
babi
lity
The Lower Jaw
LR AUC: Range = 0.72 ~ 0.8510-fold cross validation: Accuracy = 70~85%
Across 32 tooth location
Tooth #2
LR A
UC
LR A
UC
LR is the best model across 32 locations
Important Features
At risk for losing a tooth (28 teeth)Wisdom Teeth
More Lost More Present
-4 -3 -2 -1 0 1 2
Food Avoided
Smoke Now
High Income
Low Income
Aging
Coefficients-4 -3 -2 -1 0 1 2
Smoke Now
High Income
Insurance
Sex (W)
Aging
Coefficients
More Lost More Present
< 10k
100k >
What is your age (yrs)?
Are you a male or female?
Select your annual House Income ($)Do you currently smoke? If so, how many per day?Did you avoid any particular food due to teethproblems in the last 1 year?
How often did you have a dental checkupin the last 1 year?
How often do you brush and floss your teeth per day?
How often do you have sugary snacks per day?
”Please complete this survey”
35
0
1
M F
Yes NoYes No
0
12
Dentists can spot patient’s dental risksAmy Smith: 35 years old, female
The estimated probability of “losing tooth”
1 2 3 4 5 6 7 8 9 10 11 1213141516171819202122232425262728293031320.0
0.2
0.4
0.6
0.8
1.0
Tooth Location
Pro
babi
ity o
f Too
th L
oss
What is your age (yrs)?
Are you a male or female?
Select your annual House Income ($)Do you currently smoke? If so, how many per day?Did you avoid any particular food due to teethproblems in the last 1 year?
How often did you have a dental checkupin the last 1 year?
How often do you brush and floss your teeth per day?
How often do you have sugary snacks per day?
”Please complete this survey”
35
0
1
M F
Yes NoYes No
0
12
About Me
Maki S. Koyama
PhD in Physiology at University of Oxford, UK
Neuroimaging research, including infant scanning, for neurodevelopmental disorders
1 2 3 4 5 6 7 8 9 10 11 1213141516171819202122232425262728293031320.0
0.2
0.4
0.6
0.8
1.0
Tooth LocationP
roba
biity
of T
ooth
Los
s
Patient 1 Patient 2Age/Sex: 35 yrs / M 35 yrs / MInsurance: No YesIncome: Low HighSmoke now : Yes No
1 2 3 4 5 6 7 8 9 10 11 1213141516171819202122232425262728293031320.0
0.2
0.4
0.6
0.8
1.0
Tooth Location
Pro
babi
ity o
f Too
th L
ossTwo Extreme Cases
Patient 1 Patient 2
top related