For privacy and practical reasons, it is sometimes necessary to minimize sharing of individual-level information in multisite studies. However, individual-level information is often needed to perform more rigorous statistical analysis.Objectives:
To compare empirically 3 analytic methods for multisite studies that only require sharing of summary-level information to perform statistical analysis that have traditionally required access to detailed individual-level data from each site.Research Design, Subjects, and Measures:
We analyzed data from a 7-site study of bariatric surgery outcomes within the Scalable Partnering Network. We compared the long-term risk of rehospitalization between adjustable gastric banding and Roux-en-y gastric bypass procedures using a stratified analysis of propensity score (PS)-defined strata, a case-centered analysis of risk set data, and a meta-analysis of site-specific effect estimates. Their results were compared with the result from a pooled individual-level data analysis.Results:
The study included 1327 events (18.1%) among 7342 patients. The adjusted hazard ratio was 0.71 (95% CI, 0.59, 0.84) comparing adjustable gastric banding with Roux-en-y gastric bypass in the individual-level data analysis. The corresponding effect estimate was 0.70 (0.59, 0.83) in the PS-stratified analysis, 0.71 (0.59, 0.84) in the case-centered analysis, and 0.71 (0.60, 0.84) in both the fixed-effect and random-effects meta-analysis.Conclusions:
In this empirical study, PS-stratified analysis, case-centered analysis, and meta-analysis produced results that are identical or highly comparable with the result from a pooled individual-level data analysis. These methods have the potential to be viable analytic alternatives when sharing of individual-level information is not feasible or not preferred in multisite studies.