Hospital volume and outcomes after code-defined inpatient ventral hernia repair: A machine-learning analysis of the national inpatient sample, 2016–2022

Abstract

Abstract Purpose The National Inpatient Sample (NIS) does not capture outpatient hernia surgery or clinical features that define complex incisional hernia. We evaluated the continuous hospital-volume association in an inpatient code-defined abdominal wall repair (CAWR) cohort and assessed its robustness with an instrumental-variable analysis. Methods Adults in the 2016–2022 NIS were included when an ICD-10-CM K43.x ventral-hernia diagnosis occurred with an abdominal-wall mesh or component-separation/transversus abdominis release-adjacent procedure code ( n = 85,426). Double/debiased machine-learning partially linear regression estimated adjusted associations. A leave-one-out Census-division-by-year CAWR-density instrument was used as a sensitivity analysis; its estimate has a local-average-treatment-effect interpretation only if instrumental-variable assumptions hold. The primary outcome combined in-hospital mortality, coded surgical-site infection, and an intensive-care-equivalent severe course; components were also examined separately. Results Per 10 additional annual CAWR cases, the adjusted association was − 0.09% points for the composite outcome (95% confidence interval [CI], -0.19 to 0.01; P = 0.079) and − 0.05% points for mortality (95% CI, -0.09 to -0.01; P = 0.013). Length of stay increased by 0.13 days and charges by approximately 3.1%. In the instrumental-variable analysis, composite and mortality estimates attenuated toward the null, whereas length-of-stay and charge estimates persisted. Component-specific instrumental-variable estimates were heterogeneous. Conclusion Within this partial measure of inpatient hospital volume, higher volume was associated with greater inpatient resource use, whereas the mortality association was not robust to instrumental-variable analysis. NIS data cannot establish total institutional hernia volume, anatomical complexity, or a referral threshold.