A researcher is conducting a case-control study on the association between oral contraceptive use and venous thromboembolism (VTE) in Indian women. After data collection, the analysis shows a strong association (OR = 3.2, 95% CI 2.1–4.8). However, the researcher notices that cases were recruited from tertiary care centers (where VTE patients are more likely to be hospitalized), while controls were recruited from the community. What is the most appropriate next step to address this potential bias?

Explanation

## Identifying and Addressing Selection Bias **Key Point:** The mismatch between case and control recruitment sources (tertiary care vs. community) introduces **selection bias** — specifically, differential selection into the study based on exposure status or disease severity. ### Why This Is Selection Bias - **Cases** recruited from tertiary care are likely to be more severe or complicated VTE cases. - **Controls** from the community represent the general population. - This creates a **non-comparable** case-control pair, violating the fundamental assumption that cases and controls come from the same source population. ### Why Sensitivity Analysis Is the Best Next Step **High-Yield:** When selection bias is suspected *after* data collection, the immediate action is **sensitivity analysis** — a method to assess how robust the findings are to different assumptions about the bias. Sensitivity analysis allows the researcher to: 1. Estimate the magnitude of bias that would be needed to explain the observed association. 2. Determine whether the association remains significant under plausible bias scenarios. 3. Decide whether the bias is likely to have materially altered the conclusion. ### What NOT to Do | Approach | Why It's Wrong | | --- | --- | | **Stratified analysis** | Addresses confounding, not selection bias. The fundamental comparability problem remains. | | **Switch to hospital controls** | Both cases and controls would now be hospital-based, introducing **Berkson's bias** (collider bias). Hospital patients are selected by disease, making exposure–disease associations spurious. | | **Repeat as cohort study** | Appropriate for future research, but does not address the current data. | **Clinical Pearl:** Selection bias in case-control studies is often detected *post-hoc* during interpretation. Sensitivity analysis is the epidemiologist's tool to quantify its impact without discarding the data. [cite:Park 26e Ch 9]