Comparison of alternative approaches to trim subjects in the tails of the propensity score distribution

• Published in: Pharmacoepidemiology and drug safety Vol. 28; no. 10; pp. 1290 - 1298
• Main Authors: Glynn, Robert J., Lunt, Mark, Rothman, Kenneth J., Poole, Charles, Schneeweiss, Sebastian, Stürmer, Til
• Format: Journal Article
• Language: English
• Published: England Wiley Subscription Services, Inc 01.10.2019
• Subjects: Bias; comparative effectiveness; Comparative Effectiveness Research - methods; Computer Simulation; confounding bias; Data Interpretation, Statistical; equipoise; methods; Models, Statistical; pharmacoepidemiology; Population studies; Propensity Score; Sample Size; statistical; Statistical analysis; Tails; statistical; equipoise; methods; confounding bias; comparative effectiveness; pharmacoepidemiology; propensity score
• ISSN: 1053-8569; 1099-1557; 1099-1557
• DOI: 10.1002/pds.4846

Purpose In nonexperimental comparative effectiveness research, restricting analysis to subjects with better overlap of covariate distributions, hence greater treatment equipoise, helps balance the groups compared and can improve validity. Three alternative approaches, derived from different perspectives, implement restriction by trimming observations in the tails of the propensity score (PS). Across approaches, we compared the relationships between the overlap in treatment‐specific PS distributions and the size of the balanced study population after trimming. Methods The three trimming approaches considered were absolute trimming to the range 0.1<PS<0.9, asymmetric trimming to include subjects in both treatment groups with PS above the 5th percentile of the distribution in the target group and below the 95th percentile in the comparison group, and restriction to preference score values between 0.3 and 0.7. Comparisons of approaches used simulated PSs from beta distributions and two example studies. Results The magnitude of the C‐statistic strongly predicted (R2≥.95) the percent of the balanced study population remaining. The balanced study population was largest under trimming at absolute PS levels, unless the target treatment was uncommon. Fewer than half of original study subjects remained after preference score trimming if C≥.80 and after asymmetric trimming if C≥.85. In examples, trimming improved the precision of estimated risk differences and identified apparent treatment effect heterogeneity in the PS tails where covariate balance was limited. Relative amounts of trimming in examples reflected the simulation results. Conclusions Study populations with high PS C‐statistics include only small percentages of subjects in whom valid treatment effects are confidently expected.