A public health researcher in Delhi is investigating the association between air pollution (PM2.5 levels) and respiratory symptoms in school children. The study finds a strong positive correlation. However, the researcher notices that children from lower socioeconomic status (SES) schools are overrepresented in the high-exposure group, and lower-SES children also have higher rates of malnutrition and indoor cooking smoke exposure. What is the most appropriate next step to address this methodological concern?

Explanation

## Addressing Confounding in Epidemiological Studies **Key Point:** Stratification is a fundamental analytical technique to control for confounding variables when the confounder is identified during or after data collection. ### Why Stratification Is the Correct Approach Stratification (also called subgroup analysis) involves: 1. Dividing the study population into strata (groups) based on the suspected confounder (SES in this case) 2. Calculating the association between exposure (PM2.5) and outcome (respiratory symptoms) separately within each stratum 3. Comparing stratum-specific estimates to the crude estimate **High-Yield:** If the stratum-specific estimates are similar to each other but different from the crude estimate, confounding is confirmed. If they remain similar to the crude estimate, confounding is unlikely. ### Stratification vs. Other Control Methods | Method | When Used | Advantage | Limitation | |--------|-----------|-----------|------------| | **Stratification** | Post-hoc analysis; identified confounder | Simple, interpretable; allows examination of effect modification | Requires adequate sample size in each stratum; difficult with multiple confounders | | **Matching** | Study design phase | Reduces confounding at recruitment | Loss of information; cannot study confounder effect | | **Multivariate Analysis** | Design or analysis phase | Handles multiple confounders simultaneously | Less transparent; requires assumptions | | **Randomization** | RCT design | Balances known and unknown confounders | Not always feasible | **Clinical Pearl:** In observational studies like this cross-sectional investigation, stratification is the gold standard for post-hoc confounder control when confounding is suspected. ### Mnemonic for Confounding Criteria **CARE** — Confounder must: - **C**orrelated with the exposure (SES linked to pollution exposure) - **A**ssociated with the outcome (SES linked to respiratory symptoms) - **R**elated to both but not in the causal pathway (SES causes both, not caused by PM2.5) - **E**xists independently (not a mediator) SES meets all CARE criteria here, making stratification the appropriate analytical response.