Predicting In-Hospital Survival of Myocardial Infarction

Publisher Website
Google Scholar
Abstract
This study reports on the ability of several indices to predict in-hospital survival from acute myocardial infarction. The following indices were included: Acute Physiological and Chronic Health Evaluation (APACHE II), Medisgroups (MDGRP), Computerized Severity Index (CSI), Patient Management Categories (PMC), Coded Disease Staging (CDS), Ischemic Heart Disease Index (IHDI), and Predictive Index for Myocardial Infarction (PIMI). An arbitrary strategy of predicting that all patients will live was also applied and correctly classified 78% of the cases. Severity indices improve these predictions by up to 6% more. Comparison of relative accuracy of the indices showed that all indices were more accurate than PIMI and, for medically treated patients, CSI was more accurate than MDGRP, CDS, APACHE II, and IHDI. There were no other statistically significant difference in the predictive ability of remaining indices. Indices based on discharge abstracts were as accurate as some of the indices based on physiologic variables, in particular PMC was as accurate as CSI, MDGRP, APACHE, and IHDI, and CDS was as accurate as MDGRP, APACHE, and IHDI. This study was limited in scope and application and should not be generalized to other settings until additional data confirm the findings. We discuss the implications of these findings for measuring quality of care and suggest improvements for design of future severity indices.