Principles of Epidemiology for Public Health EPID 600

Principles of Epidemiology for Public Health (EPID 600) Confounding Victor J. Schoenbach, Ph. D www. unc. edu/~vschoenb/ Department of Epidemiology Gillings School of Global Public Health University of North Carolina at Chapel Hill www. unc. edu/epid 600/ 4/5/2011 Confounding 1

“Brain Cramps” "interesting" comments by well-known people (Received from Natasha Jamison, EPID 160 graduate)

That’s what can be further from the truth! “I was provided with additional input that was radically different from the truth. I assisted in furthering that version. ” – Colonel Oliver North, from his Iran-Contra testimony.

Equal opportunity employer “We don't necessarily discriminate. We simply exclude certain types of people. ” – Colonel Gerald Wellman, ROTC Instructor.

Quite a high risk, I’d say “If we don't succeed, we run the risk of failure. ” – President Bill Clinton

Why I’m glad no one is taping me “We are ready for an unforeseen event that may or may not occur. ” – Vice President Al Gore

We are here • [Now leaving] Sources of error Confounding • [Now entering] Data analysis and interpretation Causal inference Confounding 7

Setting the scene “The data speak for themselves. ” versus “Our data say nothing at all. ” (Epidemiology guru Sander Greenland, Congress of Epidemiology 2001, Toronto) Confounding 8

Setting the scene • Logically sound inferences involve (1) data + (2) assumptions • No assumptions no inference • So always need a conceptual model Sander Greenland, Congress of Epidemiology 2001, Toronto Confounding 9

Causal inference in everyday living Does exercise make me feel better? • Try getting exercise – how do I feel? • Try not getting exercise – how do I feel? • Try getting exercise again – do I feel better? Confounding 10

Causal inference in everyday living Does getting too little sleep make me irritable? • Try sleeping too little – ask my partner • Try sleeping enough – ask my partner • Try sleeping too little – ask my partner Confounding 11

Desirable attributes of crossover experiments • Exposure is under investigator’s control • Comparison condition is a true control • Can go back and forth, providing some control for secular changes Confounding 12

Constraints on cross-over experiments • Exposures may be harmful or not under our control • Effects may not be quickly reversible • Experimental subjects or the environment may have changed Confounding 13

Key attribute of crossover experiments Can compare what happens to people who are exposed to what happens to the same people when they are not exposed – almost at the same time Confounding 14

People Confounding 15

People with an exposure Confounding 16

Same people without the exposure Confounding 17

With the exposure O Confounding 18

Without the exposure O Confounding 19

Modern formulation of causal inference This comparison provides the best evidence that the exposure causes the outcome. The modern formulation of causal inference and confounding is based on this “counterfactual model”. Confounding 20

Problem of causal inference Problem: cannot observe both conditions Solution: observe a “substitute population”, a population whose experience will represent that of the exposed population without the exposure Confounding 21

“Counterfactual” model Conceptual model for causal inference: • Compare experience of a population exposed to a factor with experience of the same population at the same time but without the exposure • Since cannot do that, compare to experience of a substitute population. Confounding 22

Confounding The substitute population is not equivalent to the counterfactual condition. I. e. , the substitute population does not show the “outcome in the exposed population without the exposure”. Confounding 23

Problem of comparison Confounding is a problem of comparison – we compare the exposed population to a substitute population, but the substitute population does not show the “outcome in the exposed population without the exposure” Confounding 24

Why worry about confounding? • Does air pollution cause bronchitis ? Confounding 25

Why worry about confounding? • Does air pollution cause bronchitis ? Have choices and power Breathe polluted air ? Confounding Develop bronchitis 26

Why worry about confounding? • Does air pollution cause bronchitis ? • Do seatbelts reduce crash injuries? Risk averse Wear seatbelts ? Confounding ↓Injured in a crash 27

Why worry about confounding? • Does air pollution cause bronchitis ? • Do seatbelts reduce crash injuries? • Do STD’s increase HIV transmission? Risky sex STD ? Confounding HIV 28

Why worry about confounding? • Does air pollution cause bronchitis ? • Do seatbelts reduce crash injuries? • Do STD’s increase HIV transmission? • Does smoking lead to illicit drug use? Confounding 29

Three questions 1. What comparison should we make according to the counterfactual model? 2. What comparison will we make instead (i. e. , what substitute population will we use for the comparison)? 3. How likely is this substitute population to show us what will Fhappen in the exposed population without its exposure Confounding 30

Learning objectives 1. Understand (basic) confounding 2. Recognize potential confounding and actual confounding 3. Know how to control confounding 4. Follow discussions about confounding Confounding 31

Learning objectives - 2 5. Define and explain: – confounding – potential confounder – actual confounder – control of confounding Confounding 32

Conventional perspective Confounding: “mixing of effects” > Some other risk factor may be responsible for at least some of the association under investigation. Confounding 33

Common confounders • Age -- e. g. , exposed persons are older • Sex -- e. g. , more exposure in men • Risk factors - more exposed persons (or unexposed) smoke(-), exercise(+), eat vegetables(+), use drugs(-), . . . Confounding 34

Example of confounding in a cohort Baseline ___________ follow-up Diseased Confounding Not diseased 35

Cohort study – known risk factor Risk factor absent ___________ Risk factor present Confounding 36

Cohort study – known risk factor Risk factor absent ___________ Risk factor present ------ follow-up - - - Diseased Confounding Not diseased 37

Cohort study for a new exposure (Not angry) (Angry) Exposed follow-up Diseased Unexposed follow-up Not Diseased Confounding Diseased Not Diseased 38

Confounding in a cohort (Not angry) (Angry) Exposed Unexposed population is the “substitute population” to tell us what would happen in the exposed population without its exposure - but suppose that the exposed population have another risk factor: Confounding 39

Confounding in a cohort (Not angry) (Angry) Exposed Unexposed Substitute population will not show us what would happen in the exposed population without its exposure Confounding 40

Confounding in a cohort (Not angry) (Angry) Exposed follow-up Unexposed follow-up Not Diseased Confounding Diseased Not Diseased 41

Cohort members without the potential confounder(Not angry) (Angry) Exposed follow-up Unexposed follow-up Not Diseased Confounding Diseased Not Diseased 42

Cohort members with the potential confounder (Not angry) (Angry) Exposed follow-up Unexposed follow-up Not Diseased Confounding Diseased Not Diseased 43

Cohort members without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up Not Diseased 100 2, 400 Confounding follow-up Diseased 166 Not Diseased 8, 134 44

Cohort members without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up Not Diseased 100 (0. 04=4%) 2, 400 Confounding follow-up Diseased 166 (0. 02=2%) Not Diseased 8, 134 45

Cohort members without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up RR = 0. 04 / 0. 02 = 2. 0 Not Diseased 100 (0. 04) 2, 400 Confounding Diseased 166 (0. 02) Not Diseased 8, 134 46

Cohort members with the potential confounder (Not angry) (Angry) Exposed Unexposed 2, 500 follow-up Not Diseased 200 2, 300 Confounding Diseased 68 1, 700 Not Diseased 1, 632 47

Cohort members with the potential confounder (Not angry) (Angry) Exposed Unexposed 2, 500 follow-up Not Diseased 200 (0. 08=8%) 2, 300 Confounding Diseased 68 (0. 04=4%) 1, 700 Not Diseased 1, 632 48

Cohort members with the potential confounder (Angry) (Not angry) Exposed Unexposed 2, 500 follow-up 1, 700 RR = 0. 08 / 0. 04 = 2. 0 Not Diseased 200 (0. 08) 2, 300 Confounding Diseased 68 (0. 04) Not Diseased 1, 632 49

The entire cohort (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 1, 700 follow-up Not Diseased 100+200 8, 300 2, 400+2, 300 Confounding Diseased 166+68 Not Diseased 8, 134+1, 632 50

The entire cohort (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 1, 700 follow-up Not Diseased 100+200 (0. 06) 8, 300 2, 400+2, 300 Confounding Diseased 166+68 (0. 023) Not Diseased 8, 134+1, 632 51

The entire cohort - RR for exposure (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 8, 300 1, 700 follow-up RR = 0. 06 / 0. 023 = 2. 6 Not Diseased 100+200 (0. 06) 2, 400+2, 300 Confounding Diseased 166+68 (0. 023) Not Diseased 8, 134+1, 632 52

Comparing heterogeneous exposure groups (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 8, 300 1, 700 follow-up RR = 0. 06 / 0. 023 = 2. 6 Not Diseased 100+200 (0. 06) 0. 04 2, 400+2, 300 0. 08 Confounding Diseased 166+68 (0. 023) Not Diseased 8, 134+1, 632 53

Comparing heterogeneous exposure groups (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up 1, 700 follow-up RR = 0. 06 / 0. 023 = 2. 6 Not Diseased 100+200 (0. 06) 2, 400+2, 300 Confounding Diseased 166+68 (0. 023) 0. 02 0. 04 Not Diseased 8, 134+1, 632 54

Overall proportions are weighted averages Exposed group (Angry people) Get enough sleep 4% Sleep-deprived 6% 5% 2, 500 7% 8% 2, 500 Confounding 55

Overall proportions are weighted averages Unexposed group (not angry) Get enough sleep 2% 2. 3% Sleep-deprived 2. 5% 3% 3. 5% 4% 1, 700 8, 300 Confounding 56

Confounded comparison of weighted averages Get enough sleep 4% 2, 500 2% Sleep-deprived 6% 5% Exposed group 2. 3% 2. 5% 7% (Angry people) 3% 3. 5% 8% 2, 500 4% 1, 700 8, 300 Unexposed group (not angry) Confounding 57

Comparing within a stratum is valid Get enough sleep 4% 2, 500 2% Sleep-deprived 6% 5% Exposed group 2. 3% 2. 5% 7% (Angry people) 3% 3. 5% 8% 2, 500 4% 1, 700 8, 300 Unexposed group (not angry) Confounding 58

Comparing within a stratum is valid Get enough sleep 4% 2, 500 2% Sleep-deprived 6% 5% Exposed group 2. 3% 2. 5% 7% (Angry people) 3% 3. 5% 8% 2, 500 4% 1, 700 8, 300 Unexposed group (not angry) Confounding 59

Overall comparisons reflect subgroup sizes Get enough sleep 4% 2, 500 2% Sleep-deprived 6% 5% Exposed group 2. 3% 2. 5% 7% (Angry people) 3% 3. 5% 8% 2, 500 4% 1, 700 8, 300 Unexposed group (not angry) Confounding 60

Overall comparisons should use the same weights Get enough sleep 4% 2, 500 2% 2, 500 Sleep-deprived 5% 6% Exposed group 2. 3% 2. 5% Unexposed group Confounding 7% (Angry people) 3% 3. 5% (not angry) 8% 2, 500 4% 2, 500 61

Unconfounded RRs Get enough sleep 4% 2, 500 Sleep-deprived 5% 6% Exposed group 7% (Angry people) 8% 2, 500 RR=2. 0 2% 2, 500 2. 3% 2. 5% Unexposed group Confounding 3% 3. 5% (not angry) RR=2. 0 4% 2, 500 62

Confounded RR Get enough sleep 4% 2, 500 Sleep-deprived 5% 6% Exposed group 7% (Angry people) 8% 2, 500 RR=2. 6 2% 2. 3% 2. 5% 3% 3. 5% 4% Unexposed group (not angry) Confounding 63

RR for confounding Get enough sleep 4% Sleep-deprived 6% 5% 7% Exposed group 2, 500 (Angry people) 2. 3% 2, 500 RR=2. 6 2% 8% 3% 2. 5% 3. 5% 4% Unexposed group (not angry) RRfor confounding = RRcrude / RRunconfounded 1. 3 = 2. 6 Confounding / 2. 0 64

RR for confounding Get enough sleep 2% Sleep-deprived 3% 2. 5% 3. 5% Exposed group 2, 500 (Angry people) 2. 3% 2, 500 RR=1. 3 2% 4% 3% 2. 5% 3. 5% 4% Unexposed group (not angry) RRfor confounding = RRcrude / RRunconfounded 1. 3 = 1. 3 Confounding / 1. 0 65

Confounding in a case-control study (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 1, 700 follow-up Not Diseased 100+200 8, 300 2, 400+2, 300 Confounding Diseased 166+68 Not Diseased 8, 134+1, 632 66

Exposure odds in case group (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 8, 300 1, 700 follow-up Odds in cases = 300 / 234 = 1. 28 Diseased 166+68 100+200 Confounding 67

Exposure odds in controls (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 8, 300 1, 700 follow-up Odds in study base = 5, 000 / 10, 000 = 0. 50 Diseased 100+200 166+68 Confounding 68

Exposure odds ratio in case-cohort study (Not angry) (Angry) Unexposed Exposed 2, 500 follow-up 8, 300 1, 700 follow-up OR = 1. 28 / 0. 50 = 2. 6 Diseased 100+200 166+68 Confounding 69

Exposure odds in cases without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up Odds in cases = 100 / 166 = 0. 60 Diseased 166 100 Confounding 70

Exposure odds in controls without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up Odds in study base = 2, 500 / 8, 300 = 0. 30 Diseased 166 100 Confounding 71

Odds ratio in people without the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 8, 300 follow-up Odds = 0. 60 / 0. 30 = 2. 0 Diseased 166 100 Confounding 72

Exposure odds in cases with the potential confounder (Not angry) (Angry) Unexposed Exposed 2, 500 1, 700 follow-up Odds = 200 / 68 = 2. 94 Diseased 68 200 Confounding 73

Exposure odds in controls with the potential confounder (Not angry) (Angry) Exposed Unexposed 2, 500 1, 700 follow-up Odds = 2, 500 / 1, 700 = 1. 47 follow-up Diseased 68 200 Confounding 74

Odds ratio in persons with the potential confounder (Not angry) (Angry) Exposed Unexposed 2, 500 1, 700 follow-up OR = 2. 94 / 1. 47 = 2. 0 follow-up Diseased 68 200 Confounding 75

Potential confounder Determinant or risk factor for the outcome (or its detection). Must have potential to provide an alternative explanation for observed association. Confounding 76

Actual confounder The potential confounder becomes an actual confounder when one exposure group has more of it than the other, so it’s not fair to compare them Confounding 77

Causal models A X B (X confounding) A X (X intervening) Confounding B 78

What is a confounder - 2? A confounder is: 1. “associated with the exposure and the disease” – it causes “guilt by association”. 2. capable of being an “alternate explanation”, i. e. , the “real culprit”. Confounding 79

Control of confounding Controlling confounding means doing something to make comparison fair: • Exclude people who have the risk factor (“restriction”) • Stratified analysis (adjustment, standardization) • Mathematical modeling (e. g. , regression) Confounding 80

Control of confounding – hard to control unknown risk factors • These methods can control only known potential confounders. • Only random assignment of exposure can control for unknown potential confounders. Confounding 81

Limitations in ability to control Effective control of confounding requires: • Knowing the causal pathways • Knowing all relevant causal factors • Measuring all relevant causal factors – accurately Confounding 82

Limitations in ability to control Effective control of confounding requires assumptions, such as the mathematical form of relationships between covariables and outcome Large, randomized experiments uniquely powerful for causal inference but. . . Confounding 83

Confounded confounding! • Does overweight increase CHD risk independently of cholesterol, hypertension, and diabetes? Confounding 84

Confounding – key concepts 1. Interpreting data requires assumptions about causal relations (including what factors are potential confounders, i. e. , what factors affect incidence and are not themselves caused by the exposure). Confounding 85

Confounding – key concepts 2. If exposed people and unexposed people differ on factors that affect disease incidence, then those factors may confound (distort) the observed relation between exposure and disease (i. e. , actual confounding). Confounding 86

Confounding – key concepts 3. We can control confounding by study design if we can make the exposed and unexposed groups similar in respect to all disease determinants, though matching or randomized assignment of exposure. Confounding 87

Confounding – key concepts 4. We can control confounding in the analysis if we can stratify the data by disease determinants that are not themselves caused by the exposure (i. e. , not causal intermediates). Confounding 88

Confounding – key concepts 5. The best way to understand a case -control study is to analyze it as a window into a cohort and to be aware that many books and teachings still follow the traditional and somewhat misleading perspective. Confounding 89

Keep hope alive! Confounding can be confounding – do not be discouraged if you do not understand it yet. Confounding 90

Dietary advice The Japanese eat very little fat and suffer fewer heart attacks than the British or Americans. On the other hand, the French eat a lot of fat and also suffer fewer heart attacks than the British or Americans. Confounding 91

Dietary advice The Japanese drink very little red wine and suffer fewer heart attacks than the British or Americans. On the other hand, Italians drink excessive amounts of red wine and also suffer fewer heart attacks than the British or Americans. Confounding 92

Dietary advice - conclusion Conclusion: Eat & drink what you like. It appears that speaking English is what kills you. (submitted by Natasha Jamison, EPID 160 student) Confounding 93

Hmmm. "Your food stamps will be stopped effective March 1992 because we received notice that you passed away. May God bless you. You may reapply if there is a change in your circumstances. ” – Department of Social Services, Greenville, South Carolina