Predictions and Outcomes in Congestive Heart Failure. Abstract, Executive Summaryand Final Report

摘要

Background: How well physicians make ICU triage decisions is almost unknown.Current guidelines suggest triage physicians should estimate the probability that a patient will soon develop severe complications, as well as patients' comparative survival probabilities. Study Design: Predictions and Outcomes in Congestive Heart Failure was a cohort study which followed patients with new and existing congestive heart failure (CHF), initially asking their physicians to predict their outcomes. Measurements: The quality of physicians' judgements of the probability of CHF complications in their patients at 4 days, of survival at 90 days, and of survival at one year, as well as the development and validation of predictive models of these outcomes. Results: The prevalence of complications at four days was low (4.3%). Physicians were only moderately able to predict complications and/or survival (ROC curve areas = 0.57 to 0.72; calibration curves suggested both overestimation and underestimation). Models of 90-day and one-year survival based on traditionally used variables, and those based on 'off-the-shelf' models of ICU survival, performed 4 no better or only marginally better (ROC curve areas 0.58 to 0.70).