A time-insensitive predictive instrument for acute hospital mortality due to congestive heart failure: development,...

The purpose of this study was to develop a "time-insensitive" predictive instrument (TIPI) for acute hospital mortality due to congestive heart failure. In Phase 1, based on prospectively collected data on 401 congestive heart failure patients among 5,773 study patients who presented to six New England hospitals over a 2-year period whose chief complaints were chest pain, shortness of breath, or other cardiac symptoms, a multivariable logistic regression was used to develop the TIPI for acute mortality. Discrimination between patients who lived and those who died was reflected by receiver-operating characteristic (ROC) curve area of 0.90. Predicted mortality was found to not vary significantly from actual mortality rates across deciles of predicted probabilities from 0% to 100%. In Phase 2, the six hospitals' actual mortality rates for their congestive heart failure patients were compared to their respective rates predicted by the TIPI. Actual hospital mortality rates ranged from 3.6% to 11.3%, with no hospital having a statistically significantly higher rate. Predicted mortality rates ranged from 4% to 9%, with one hospital having a significantly lower predicted rate (P = .01), and one hospital having a borderline significantly higher predicted rate (P = .07). Individual hospitals' differences between actual and predicted mortality ranged from -3.8% to +4.7% (all NS). When grouped by hospital type, respectively for urban teaching, smaller city teaching, and rural non-teaching hospitals, the actual mortality rates were 5.1%, 10.5%, and 5.4%, (NS). The predicted mortality rates were 8.3%, 6.1%, and 5.4%, respectively, with the rate for urban major teaching centers being significantly higher (P = .03). No hospital type had significant differences between their actual and predicted mortality rates (NS). This congestive heart failure mortality TIPI (CHFM-TIPI) shows potential for risk-adjusted studies of hospitals, mortality for multi-hospital groups, hospital-to-hospital comparisons, and potentially for within-hospital assessment and if further validated, potentially also for real-time clinical use.