self experiments analysis of patients’ causal diaries
TRANSCRIPT
Self ExperimentsSelf Experiments
Analysis of Patients’ Causal Analysis of Patients’ Causal DiariesDiaries
System Change for System Change for Exercise Maintenance in Exercise Maintenance in Older Cardiac Patients Older Cardiac Patients
National Heart and Blood Institute National Heart and Blood Institute 09/1/2006- 09/1/2009 09/1/2006- 09/1/2009
PI: Shirley Moore RN, PhDPI: Shirley Moore RN, PhDCo-PI: Farrokh Alemi, PhDCo-PI: Farrokh Alemi, PhD
Exercise Causes & Exercise Causes & Constraints Constraints
“… I realized how many great reasons I had to skip my workout today. First: rain… Second: low quality sleep after a night spent with my 17-pound cat getting tangled in the blinds. How could anyone exercise after a night like that?”
Paige Waehner
Exercise Causes & Exercise Causes & Constraints Constraints
“What makes me exercise is that I have to take a shower and the only place I can take a fun shower, with lost of water, is at the gym. My shower at home does not have much water pressure. I have no choice but to go to the gym.”
68 years old woman
Exercise Causes & Exercise Causes & Constraints Constraints
“When I bike, I do not exercise. I commute to work. ”
42 year old man
Exercise Causes & Exercise Causes & Constraints Constraints
People have different reasonsPeople have different reasons– One solution does not fit allOne solution does not fit all
People have wrong perceptionsPeople have wrong perceptions– I fail because of my environmentI fail because of my environment– I succeed because of myselfI succeed because of myself
Exercise Causes & Exercise Causes & Constraints Constraints
People have different reasonsPeople have different reasons– One solution does not fit allOne solution does not fit all
People have wrong perceptionsPeople have wrong perceptions– I fail because of my environmentI fail because of my environment– I succeed because of myselfI succeed because of myself
We cannot succeed, if we do not know why we
have failed
Post-cardiac Exercise Post-cardiac Exercise PatternsPatterns
0
1020
30
4050
60
70
8090
100
Rehab 6-months
% E
xe
rcis
ing
ObjectivesObjectives Health
Sustain exercise post rehab
ObjectivesObjectives Health
Sustain exercise post rehab
Understand causes of & constraints for exercise
ObjectivesObjectives Health
Sustain exercise post rehab
Understand causes of & constraints for exercise
Conduct self experiments: Maintain diary, analyze data, repeat the process. Gain insight.
Self ExperimentsSelf Experiments
List possible causes/constraintsList possible causes/constraints Trace occurrences Trace occurrences Analyze dataAnalyze data
– Small data sets of 10-14 data pointsSmall data sets of 10-14 data points
ExampleExample
DiaryDiary
Check
if cause
Other causes (please specify):
List Constraints
Consider as constraints factors that prevent you from completing structured exercise. For example, rainy weather may prevent you from going for a run or walk.
List Causes
Consider as possible causes factors that enable you to complete structured exercise. Think through the impact of daily routines (e.g. commuting, sleeping patterns, bathing patterns), medication changes, weather and environmental changes, socialization routines and other physical changes in the environment. Do not list your motivation or your time as causes but list factors that improve your motivation or available time.
Check
if present
Day 1
YYMMDD//
Did you exercise? Yes No Not sure
Steps taken today: ? ? ? , ? ? ? ,? ? ?
Did you talk about exercise with others? Yes, with friends or family members Yes, with project staff No
Start time of exercise:
MMHH:
End time of exercise:
MMHH:
Type of exercise:
Level of exertion: Very, very light Somewhat hard Very light Hard Light Very hard Fairly light Very, very hard
Did you reach your prescribed exercise zone? Yes No Not sure
Any changes in your medications? Yes No Not sure
Any contact with health services? Emergency Room Office visit Hospital
Was the contact heart related? Yes No
Bike to work
Shower at gym
Rain
Sleep early
14-day Diary14-day Diary
Day Rain
Plan to commute with bike
Plan to shower at
gymSleep early
Exercise pattern
kept
1 1 1 0 0 0
2 0 0 1 1 1
3 0 0 1 1 0
4 0 1 1 0 1
5 1 1 1 0 1
6 1 1 0 0 0
7 0 0 0 0 0
8 0 0 1 0 1
9 0 0 0 0 0
10 1 1 0 1 1
11 0 0 0 0 0
12 0 0 1 0 1
13 1 0 1 0 1
14 0 1 0 0 1
14-day Diary14-day Diary
Day Rain
Plan to commute with bike
Plan to shower at
gymSleep early
Exercise pattern
kept
1 1 1 0 0 0
2 0 0 1 1 1
3 0 0 1 1 0
4 0 1 1 0 1
5 1 1 1 0 1
6 1 1 0 0 0
7 0 0 0 0 0
8 0 0 1 0 1
9 0 0 0 0 0
10 1 1 0 1 1
11 0 0 0 0 0
12 0 0 1 0 1
13 1 0 1 0 1
14 0 1 0 0 1
Too little data for most statistical methods
of analysis
Obvious LessonsObvious Lessons
No variation in outcomes: No variation in outcomes: – No exercise in the entire 2 weeksNo exercise in the entire 2 weeks– Exercise every dayExercise every day
No variation in causes:No variation in causes:– Always present causeAlways present cause– Always absent causeAlways absent cause
Causal AnalysisCausal Analysis
1.1. SequenceSequence Cause precedes exerciseCause precedes exercise
2.2. AssociationAssociation When the cause is present, exercise When the cause is present, exercise
should be likely should be likely Counter-factualCounter-factual
If the cause is absent, and no other If the cause is absent, and no other causes are present, exercise should be causes are present, exercise should be unlikelyunlikely
Methods of AnalysisMethods of Analysis
1.1. Logistic regressionLogistic regression
2.2. Bayesian networksBayesian networks
3.3. Causal analysisCausal analysis
Method 1: Logistic Method 1: Logistic RegressionRegression
Log [pi /(1-pi )] = α 0 + α1xi + εi
Log [pi /(1-pi )] = α 0 + α2yi + εi
Log [pi /(1-pi )] = α 0 + α3zi + εi
Using the coefficients estimated in logistic regression, the various conditional probabilities of success, the main effect of the cause on success, can be calculated as:
P(S | xi=1) Max
= e α 0+ α1 / (1+ e α 0+ α1 ) P(S | xi=1)
Min = e0+1 / (1+ e0+1 )
P(S | yi=1) Max
= e α 0+ α2 / (1+ e α 0+ α2 ) P(S | yi=1) Min = e0+2 / (1+ e0+2 ) P(S | zi=1)
Max = e α 0+ α3 / (1+ e α 0+ α3 )
P(S | zi=1) Min = e0+3 / (1+ e0+3 )
Log [pi /(1-pi )] = β0 + β1xi+β2yi+β3zi + ...+ εi
Method 2: Bayesian Method 2: Bayesian NetworkNetwork
Markov blanketMarkov blanket Use of conditional probabilitiesUse of conditional probabilities
– Serial conditional independenceSerial conditional independence– Common causeCommon cause– Common effectCommon effect
Method 3: Causal AnalysisMethod 3: Causal Analysis
Xined cause unconstracases withNumber of
e Xained caush unconstr cases witsuccessfulNumber of MaxX)|p(S
causeotherno Xined cause unconstracases withNumber of
ther causee X & no oained caush unconstr cases witsuccessfulNumber of
&X)|p(S Min
1-Cou
nter
fact
ual
Conditional
14-day Diary with No 14-day Diary with No ConstraintsConstraints
Day
Plan to commute with bike
Plan to shower at
gym Sleep earlyExercise
pattern kept
1 0 0 0 0
2 0 1 1 1
3 0 1 1 0
4 1 1 0 1
5 0 1 0 1
6 0 0 0 0
7 0 0 0 0
8 0 1 0 1
9 0 0 0 0
10 0 0 1 1
11 0 0 0 0
12 0 1 0 1
13 0 1 0 1
14 1 0 0 1
Method 3: Causal AnalysisMethod 3: Causal Analysis
Day
Plan to commute with bike
Plan to shower at
gym Sleep early
Exercise pattern
kept
4 1 1 0 1
14 1 0 0 1
Exercise Pattern on Days in which Client was Ready to Bike
Probability of Success Given the Cause
1-Counterfactual Conditional
Plan to commute with bike 1 1
Method 3: Causal AnalysisMethod 3: Causal Analysis
Day
Plan to commute with bike
Plan to shower at
gym Sleep earlyExercise
pattern kept
2 0 1 1 1
3 0 1 1 0
4 1 1 0 1
5 0 1 0 1
8 0 1 0 1
12 0 1 0 1
13 0 1 0 1
Exercise on Days in which the Client was Ready to Take Shower at Gym
Probability of Success Given the Cause
1-Counterfactual Conditional
Plan to shower at gym 0.80 0.86
Method 3: Causal AnalysisMethod 3: Causal Analysis
Probability of Success Given the Cause
1-Counterfactual Conditional
Plan to commute with bike 1 1
Plan to shower at gym 0.80 0.86
Sleep early 0.50 0.67
Probability of Success Associated with Different Causes
Method 3: Causal AnalysisMethod 3: Causal Analysis
Study Phase IStudy Phase I
Which of the methods is most Which of the methods is most accurate?accurate?
Which method is easier to Which method is easier to understand?understand?
Which method is easier to use?Which method is easier to use?
Questions & Questions & CommentsComments