Predictive modeling of in-hospital mortality following elective surgery.

Overview

INTRODUCTION: The specific healthcare macroenvironment factors contributing to in-hospital mortality following elective surgery remain nuanced. We hypothesize an accurate global elective surgical mortality model can be created. METHODS: FL AHCA and Hospital Compare (2016-2019) were queried for in-hospital mortality following elective surgeries. Stepwise logistic regression with 47 patient and hospital factors was followed by gradient boosting machine (GBM) modeling describing the relative influence on risk for in-hospital mortality. Deceased and surviving patients were matched (1:2) to perform univariate analysis and logistic regression of significant factors. RESULTS: A total of 511,897 admissions, 2,266 patient deaths and 162 Florida hospitals were included. GBM factors (AUC 0.94) included post-operative patient and hospital factors. In the final regression model, patient age older than 70 years of age and hospital 5-star rating were significant (OR 2.87, 0.47, respectively). Hospitals rated 5-stars were protective of mortality. CONCLUSION: In-patient mortality following elective surgery is influenced by patient and hospital level factors. Efforts should be made to mitigate these risks or enhance those that are protective.