The Evaluation of Hospitalist Care for Patients with Aspiration Pneumonia Using Risk-adjusted Performance Indicators...

Objective High-quality evidence proving the superiority of hospitalist services is lacking. We developed risk-adjusted performance indicators from a multilevel prediction model using a nationwide inpatient database to evaluate hospitalist medical care for patients with aspiration pneumonia. Methods We extracted cases diagnosed with aspiration pneumonia between 2014 and 2021 from the Diagnosis Procedure Combination (DPC) database. Hospital-level risk-adjusted performance indicators were the observed-to-expected ratio of the following outcomes using a multilevel prediction model containing both patient- and hospital-level variables: death or transfer in poor condition within 30 days (Poor outcome), in-hospital death within 30 days, and discharges within the 25th and 50th percentiles for length of stay defined by the DPC system. Using the predicted numbers of each outcome without random intercept as denominators of both indicators, the numerators of Indicator 1 were observed numbers of each outcome, while those in Indicator 2 were "smoothed" predicted numbers of outcomes estimated by the fitted model with random intercept. The ratio of the number of outcomes for each hospital to the mean number of outcomes among participating hospitals was used as a reference. We applied these indicators to Takatsuki General Hospital (TGH) as a working example. Results A total of 526,245 patients were analyzed. Compared with Indicator 1, Indicator 2 showed greater stability in the mean ratio and bootstrapping confidence interval (CI). Indicator 2 of Poor outcome and discharges within the 25th percentile in 2017 at TGH were 1.110 (95% CI 0.784-1.375) and 1.458 (95% CI 1.272-1.597), respectively. Conclusion Utilizing a nationwide inpatient database, we developed risk-adjusted performance indicators using a multilevel prediction model to evaluate hospitalist medical care for patients with aspiration pneumonia. Given the reliable results shown in the working example, these indicators have potential benefits for the accurate evaluation of the quality of medical care.