Case Control Study: An example

1
Case Control Study An example

• Depression and the Risk of
• Out-of-hospital cardiac arrest
• July, 2005

2
Background - OOH Cardiac Arrest

• Sudden pulseless condition due to heart disease.
• Critical confluence of chronic substrate acute
  trigger
• Chronic Acute
• Accounts for up to 400,000 deaths annually in the
  US.

3
Theme of Research

• How do common non-cardiac clinical
• conditions and treatments influence the risk
• of cardiac arrest?
• Depression alters autonomic tone, increases
• platelet aggregation, and affects health
• behaviors and compliance with care.

4
Hypothesis

• Depression is associated with an increased
• risk of cardiac arrest.

5
Study Design?

• Must consider the exposure, outcome,
• and potential populations when choosing a
• study design.
• Randomized trial
• Cohort Study
• Case control

6
Study Design?

• Exposure cannot be randomized.
• Outcome is 1/1000 person years
• Randomized trial
• Cohort Study
• Case control

7
Exposure, Outcome, and Population

• Exposure ascertainment
• - De novo subject evaluation/interview
• - Other informational resource
• Exposure definition - Ideal versus practical

8
Exposure, Outcome, and Population

• Exposure ascertainment
• De novo subject interview/evaluation
• Medical record
• Health Plan Enrollees
• Definition Medical record review
• - Clinical assessment, measures of severity

9
Exposure, Outcome, and Population

• Outcome ascertainment
• - Emergency Medical Services Records
• - Death Records of the Health Plan
• Outcome definition
• - Validation of a sample of events
• - Agreement between EMS and death records

10
Exposure, Outcome, and Population

• Progress (definition and ascertainment)
• Population- Health Plan Enrollees
• Exposure- Medical record review
• Outcome- EMS and death Records

11
Controls

• The goal is to simulate a randomized trial to try
• and best examine the exposure of interest.
• If possible incorporate what is already known
• about risk factors for the outcome and the
• studys goals when choosing controls.

12
Controls

• Using a random sample would result in substantial
  
• confounding. Overcome confounding in the design
• and sampling stage.

Cardiac arrest risk factors Older age 65
years Male gender 67 Clinical heart disease
67
Random sample of Enrollees Age 45 years Male
gender 50 Clinical heart disease 15
13
Controls
Cardiac arrest risk factors Older age 65
years Male gender 67 Clinical heart disease
67
Selected sample of Enrollees Match on age Match
on gender Match on heart disease status
14
Planning data collection

• Exposure
• - Temporal relationship to outcome
• - Dose response
• Account for potential confounders
• - Specific type of heart disease
• - Established risk factors for cardiac arrest
• - Medication use

15
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services

16
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding

17
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval

18
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification

19
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification
• Medical record review and abstraction

20
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification
• Medical record review and abstraction
• Data entry and cleaning

21
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification
• Medical record review and abstraction
• Data entry and cleaning
• Data analysis

22
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification
• Medical record review and abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript - Interpretation

23
Study Process

• Interface and coordination with the Health Plan
• and Emergency Medical Services
• Funding
• Human Subjects review and approval
• Case (and then) control identification
• Medical record review and abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript - Interpretation
• Publication

24
Results

• Outcome
• Exposed? Yes No
• Yes
• No

a b
c d
25
Data Layout

• Cardiac Arrest
• Depression? Yes No
• Yes
• No

a b
c d
26
Results

• Cardiac Arrest
• Depression? Yes No
• Yes
• No

300 327
1928 3837
Odds ratio aXb/cXd 300X3837/327X1928 Unadjuste
d odds ratio 1.83 Adjusted odds ratio 1.43
27
Results

• Evidence of dose response
• Restriction and adjustment to account for
  confounding.

2
OR
1.5
1
No depress 1º Care Specialty Care
28
Strengths

• Outcome is rare
• Relatively quick and inexpensive
• Study gt 1 exposure

29
Weaknesses

• Cannot estimate incidence directly
• 2000 cases and 4000 controls
• Choice of controls is challenging
• Recall bias

30
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

31
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
32
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
33
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
34
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
35
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
36
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
54 months
37
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
54 months
57 months
38
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
54 months
57 months
60 months
39
Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
54 months
57 months
60 months
72 months
40
Permutations of Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

6 months
24 months
30 months
36 months
48 months
54 months
57 months
60 months
72 months
41
Permutations of Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

9 months
6 months
24 months
Research accelerator
30 months
36 months
48 months
54 months
15 months
18 months
30 months
42
Permutations of Study Process
Time line

• Research idea
• Interface with HP/EMS
• Funding
• Human Subjects
• Subject identification
• Medical record abstraction
• Data entry and cleaning
• Data analysis
• Draft manuscript
• Publication

9 months
6 months
24 months
Young investigator Fortunate investigator Wise
investigator
30 months
36 months
48 months
54 months
15 months
18 months
30 months
43
Summary

• When choosing study design, consider issues
• of exposure, outcome, and population.
• Understand the strengths and weaknesses of
• case control studies and attempt to address
• during the design and then in analysis
• Regardless of study design, recognize the
• hidden/logistic challenges and their time
• requirements.

44

• Thanks to David Siscovick, Bruce Psaty,
• and Mickey Eisenberg
• Contact information
• Tom Rea
• Division of General Internal Medicine
• Box 359945
• Pager 540-8048
• E-mail rea123_at_u.washington.edu