Lecture 10 Epidemiology GPH-GU 2106 Fall 2024 PDF

Lecture 10 Epidemiology GPH-GU 2106 - Fall 2024 Lecture 10 – Confounding 1 Agenda l To understanding confounding. l To differentiate between confounders a...

Lecture 10 Epidemiology GPH-GU 2106 - Fall 2024 Lecture 10 – Confounding 1 Agenda l To understanding confounding. l To differentiate between confounders and confounding. l To differentiate between mediators and confounders. l To understand how to control for confounding in the design and analysis phase of epi studies. 2 1 Lecture 10 Observe an association Is it real? Real Spurious Is it due to presence Bias/Error of 3rd variable? Due to rd 3 variable Bias: Confounding 3 Confounding l Results from what is known as a “mixing of effects” (Rothman) Measure of association l The estimate of effect of the exposure on the outcome is distorted because it is mixed with the effect of an extraneous factor related to both the exposure and outcome l Result of confounding is to distort the true measure of association toward the null (negative confounding) or away from the null (positive confounding). 4 2 Lecture 10 Definition: confounding l Situation in which a non-causal association between an exposure and an outcome is observed as a result of the influence of a third variable known as a confounding variable or confounder l A variable is a confounder if it meets 3 criteria: 1. It is a known risk factor for the outcome 2. It is associated with the exposure 3. It is not a result of the exposure 5 Confounding - Diagram Potential Confounder Exposure of Disease or ? Interest Outcome Relationship of interest being evaluated = CAUSAL relationship = Non-causal relationship (aka an association) 6 3 Lecture 10 Working Example: Gender as risk factor for malaria Gender ? Malaria Relationship of interest Cases Controls Males 88 68 88 / 62 88*82 OR = = = 1.71 68 / 82 62*68 Females 62 82 Adapted from Szklo & Nieto, 1999 7 l Could work environment confound the relationship between gender and malaria? Working Outdoors Gender ? Malaria Relationship of interest 8 4 Lecture 10 Review: meaning of association l If the confounder is a risk factor for the disease: l then risk of disease is different among people with the confounder characteristic compared to those without. l If the characteristic is associated with exposure: l then the distribution of the confounder characteristic is different among people with the exposure compared to people without exposure (unbalanced between groups). What study characteristic does this remind you of? 9 Evaluating work environment as a potential confounder 1. Is working outdoors a risk factor for malaria? Working Cases Controls Outdoors Outdoors 63 18 Indoors 87 132 Gender Malaria 150 150 l What proportion of cases & controls work outdoors? l 42% of cases work outdoors l 12% of controls work outdoors 63*132 OR = = 5.3 18*87 Szklo & Nieto, 1999 10 5 Lecture 10 Evaluating work environment as a potential confounder 2. Is working outdoors associated with gender? Working Outdoors Indoors Outdoors Males 68 88 156 Females 13 131 144 Gender Malaria l Those who work outdoors are more likely to be male than those who work indoors. l Compared with only 9% of females, 44% of males work outdoors 68*131 OR = = 7.8 13*88 Szklo & Nieto, 1999 11 Evaluating work environment as a potential confounder 3. It is not a result of the exposure l Is the potential confounder in the causal pathway between the exposure ® outcome? Working Malaria Gender Outdoors 12 6 Lecture 10 PAUSE: When the 3rd factor is in the causal pathway: l A variable cannot be a confounder if it is a step in the causal chain or pathway between the exposure and the outcome. l Such variables are considered mediators l We WANT to understand influence of mediators Variable X Exposure Outcome ? Diet Cholesterol Heart Disease Level ? 13 PAUSE : What’s a mediator? l When considering mediators, investigators may seek to disentangle the effects as: Variable X Indirect Effect on Outcome ? Exposure Outcome Direct Effect on Outcome 14 7 Lecture 10 Methods used to address confounding l Analysis 1. Stratification 2. Adjustment l Design 1. Matching 2. Randomization (experimental) studies 3. Restriction 15 1. Stratification l Allows investigator to hold the confounder of interest ‘constant’ within a strata (category) of the exposure l group data into (homogeneous) categories of extraneous factor l Examine relationship within strata of the confounder l analyze for each category l combine for summary estimate 16 8 Lecture 10 Stratified analyses examining relationship between gender and malaria controlling for work environment Mostly Outdoor Occupation Cases Controls 53*3 Males 53 15 OR = = 1.06 15*10 Females 10 3 Mostly Indoor Occupation Cases Controls 35*79 Males 35 53 OR = = 1.00 52*53 Females 52 79 Adapted from Szklo & Nieto, 1999 17 1. Stratification l Allows investigator to hold the confounder of interest ‘constant’ within a strata (category) of the exposure l group data into (homogeneous) categories of extraneous factor l Examine relationship within strata of the confounder l analyze for each category l combine for summary estimate l Limitations of Stratification l As number of confounders increase, the size of each stratum gets very small 18 9 Lecture 10 2. Adjustment l Use different statistical technique to estimate what the association would be IF the confounder was not associated with the exposure l Multivariable regression analysis l Adjust for the confounder through statistical modeling l Also allows adjustment for multiple confounders simultaneously 19 Adjustment: The malaria example l What would the OR for malaria in men vs. women be IF men and women were equally likely to work outdoors l Crude OR = 1.71 l Outdoor work adjusted OR = 1.0 (from multivariable logistic regression analysis) 20 10 Lecture 10 Remember: Age adjustment to compare mortality risk in Black vs. White residents of Georgia White Black Population Population Crude Mortality 8.1 per 1,000 7.0 per 1,000 Age Age Adjusted Distribution Mortality 7.3 per 1,000 9.4 per 1,000 Age Group White Pop. % Black Pop. %