Hospital Factors Do Not Explain Away Socioeconomic-based Disparities in Surgical Outcomes
*Kyla M Bennett, *John E Scarborough, Theodore N Pappas, *Thomas B Kepler
Duke University Medical Center, Durham, NC
Objective: Recent studies drawn from Medicare data suggest that hospital-related factors largely explain socioeconomic-based disparities in surgical outcomes. We used an all-payor nationally representative database to test this hypothesis.
Methods: The Nationwide Inpatient Sample from 2001-2006 was used to determine association between postoperative in-hospital mortality and patient- and hospital-level variables for 13 complex cardiovascular and oncologic procedures. Given available predictor variables, a subset of all possible logistic regression models for prediction of mortality was examined exhaustively and the Akaike Information Criterion (AIC) computed for each model. Model selection was performed by minimizing the AIC over all models. Model averaging was performed by weighting each model by AIC-based probability and summing over models.
Results: 1,059,003 patients were included for analysis. After adjustment for patient-related factors (age, gender, race, comorbidity), hospital procedure volume and hospital socioeconomic status (SES), low patient SES remained a significant positive predictor of postoperative mortality for 11 of 13 procedures analyzed. By both model selection and model averaging, patient SES was as important as or more important than hospital SES or hospital procedure volume in predicting postoperative mortality.
Conclusions: In contrast to recent reports, we find that patient SES is as important as or more important than hospital-related factors such as procedure volume and hospital SES in predicting mortality after complex surgery. Policies that attempt to reduce socioeconomic-based disparities by addressing only these hospital-related factors are likely to be inadequate.