Abstract 14379: A Hospital Outcome Prediction Model in Percutaneous Coronary Intervention: Volume-Specific Analysis Based on Adverse Ratios and Risk Adjusted Mortality

Abstract

Introduction: Percutaneous coronary intervention (PCI) hospital volume is an important quality surrogate for clinical outcomes. Yet PCI volume also inflates or deflates ratio analysis that is fundamental to outcomes assessment. The hospital volume-based PCI outcome prediction has not been reported.

Hypothesis: We hypothesize that, based on adverse ratios and risk adjusted mortality the scores of discriminant analysis account for variations in patient population, physician skills and care management of a hospital to provide a fair comparison between outcomes across hospitals and reflect the overall quality of care provided.

Methods: The 860,892 PCI procedures from 1,231 hospitals submitted to CathPCI Registry® successfully and consecutively between 1/1/2007 and 12/31/2008 were studied. We analyzed the adverse outcomes and hospital PCI volume using ratio analysis and z-scores to eliminate the impact of volume on outcomes assessment. We used the Risk-adjusted Mortality (RAM) model to compute patient risk scores associated with each PCI performed, and then introduced the discriminant analysis to develop a discriminant function model which was based on hospital related major adverse ratio, patient-sick-level related major adverse ratio, hospital & patient related major adverse ratio, hospital related minor adverse ratio, hospital patient risk ratio, and RAM.

Results: The prediction model was developed for the hospitals with annual PCI volume ≥ 200 procedures. Four optimized threshold values classified the outcomes into 4 classes: Exceptional, Above-average, Average, and Below-average. This proposed model provides a direct interpretation in estimating hospital’s PCI outcomes. Our statistical results indicate the validity of the model and its predictive accuracy: 94.3% preliminary grouped cases correctly classified and 92% test grouped cases correctly classified.

Conclusions: The proposed model provides a new method to predict a hospital’s PCI outcome with a comparison to its peers by eliminating the inflating or deflating effects of hospital PCI volumes.