Increasing Omission of Race and Ethnicity in a National Surgical Database.

Abstract

Accurate data collection on race and ethnicity is critical to detecting and characterizing disparities in surgical care. This study explores the prevalence and impact of missing race and ethnicity data in the NSQIP Pediatric (NSQIP-P) database. NSQIP-P from 2016 to 2020 was assessed for patients who were missing data for race (RM), ethnicity (EM), or both (BM). Changes in the proportion of RM, EM, and BM patients were assessed with Cochran-Armitage tests. Surgical outcomes were modeled using linear mixed-effects models, treating CPT codes as random effects. Models were bootstrapped with 1,000 iterations for each outcome. Of 596,571 patients, 18.0% were RM, 10.9% were EM, and 8.3% were BM. BM patients increased in proportion from 5.7% in 2016 to 12.1% in 2020 (p < 0.0001). Outcomes that differed statistically significantly were worse in RM than in non-RM patients. Multivariate analysis found that BM patients had longer hospital length of stay than non-BM patients (b = 0.5 days, p < 0.001). BM patients also had higher odds of readmission (odds ratio 1.22 [1.02 to 1.46]) and any complication (odds ratio 1.27 [1.03 to 1.59]). NSQIP-P exhibits a concerning increase in missing race and ethnicity data that has not been previously described in other literature or database documentation. Controlling for surgery type and patient demographics, patients exhibited worse outcomes with respect to hospital length of stay and both readmission and any complication rate. Further work is necessary to determine the cause of these missing data and provide guidance for researchers examining surgical disparities using NSQIP-P.