The potential value of observational studies of elective surgical interventions using routinely collected data

Abstract

PURPOSE: To assess the apparent validity of observational studies of elective arthroplasty interventions. METHODS: Data from the nationwide Dutch Arthroplasty Register were used. The first case study compared surgical approaches for total hip arthroplasty (posterolateral approach vs. straight lateral approach), where allocation of the intervention was assumed to be mostly independent of patient characteristics. The second case study compared fixation methods (cemented vs. uncemented), where choice of fixation method was expected to depend on patient characteristics. The potential for confounding was quantified by differences between intervention groups and the impact of confounding adjustment. RESULTS: The study of posterolateral approach versus straight lateral approach included 73,750 and 16,557 patients, respectively, and showed no meaningful differences in patient characteristics between treatment groups (standardized mean differences <0.1) and also no relevant impact of confounding adjustment (Z-scores <1). The study of cemented versus uncemented total hip arthroplasty (THA) included 29,579 and 79,360 patients, respectively. Several meaningful imbalances were observed in patient characteristic between the two treatment groups (standardized mean differences >0.1), as well as a relevant impact of confounding adjustment (Z-scores >2). CONCLUSIONS: This study provides insight in the reasoning behind the credibility of observational studies of surgical interventions using routinely collected data and when confounding is expected to have a major impact and thus additional precautions to limit confounding are needed.