In Reply The potential for unmeasured confounding is a limitation of all observational analyses. Propensity scores can at best achieve balance in measured confounders.1 Dr Xue and colleagues correctly assert the importance of including all known factors in the matching model. We were unable to include all covariates that may have an effect on postoperative outcomes (eg, laboratory measurements and chronic treatment prior to surgery), as these were not available in the database. Although frailty and functional status are potentially the most important unmeasured confounders, our sensitivity analysis restricted to patients younger than 60 years who were living at home suggests it is unlikely that this explains the differences in outcomes. This analysis demonstrated an increased risk of mortality (relative risk [RR], 5.80 [95% CI, 2.96-11.35]; P < .001) with hip fracture surgery relative to elective total hip replacement.
Le Manach Y, Collins GS, Devereaux PJ. Hip Fracture Surgery vs Elective Total Hip Replacement—Reply. JAMA. 2016;315(9):942-943. doi:10.1001/jama.2015.17861