Hospital A reports a higher crude post-operative mortality than hospital B. However, when patients are stratified by pre-operative risk category, hospital A shows LOWER mortality in both the high-risk and low-risk strata. Hospital A receives most of the region's high-risk referrals. This reversal of an association after stratification illustrates:

• A Effect modification by operative risk
• B Random error due to small sample size
• C Detection bias favouring hospital A within strata
• D Confounding by case mix, classically described as Simpson's paradox ✓

Explanation

When the crude association reverses direction within every stratum of a third variable, the third variable (here, proportion of high-risk patients) is a strong confounder, and the phenomenon is Simpson's paradox. Effect modification would mean the true association genuinely differs across strata, not that it uniformly reverses against the crude estimate. Recognition demands stratified or adjusted analysis before comparing hospital performance.

Reference: Rothman, Greenland and Lash, Modern Epidemiology, 3rd ed.

High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP

Written and medically reviewed by the StethoPrep medical team.