This eight-month budget is necessary for completing the prospective comparison of the accuracy and utility of three methods for predicting intensive care unit (ICU) mortality - multiple logistic regression (MLR) models (developed during the first budget year), the Acute Physiology Score (APS), and the Therapeutic Intervention Scoring System (TISS) - using a cohort of approximately 2000 ICU patients. Additionally, a revised version of the Acute Physiology Score, developed in 1984, will be included in formal comparisons of predictive efficacy. A second objective is to complete the six-month follow-up of all cohort members alive at hospital discharge to ascertain vital status, return-to-work status, and level of self-care. This study is being conducted prospectively on patients admitted to the ICUs at Baystate Medical Center, Springfield, MA. All study participants are being evaluated with respect to their probability of survival using the four methods noted above. The three methods will then be statistically compared on the basis of actual patient outcome. At six months post-hospital discharge a questionnaire is mailed to each patient to ascertain vital status, return-to-work status, and level of self-care. A telephone follow-up is conducted for all patients who do not return the questionnaire. Given formal sample size calculations based on the sensitivity and specificity of each method, it will be necessary to have a cohort of 2000 patients to detect a significant difference between the MLR and the APS methods. This eight-month supplement to the current grant is intended to extend data collection to obtain an adequate sample size. A small sub-study will be performed to assess the ability of a variety of non-data collector nurses to abstract accurate information for each of the predictive systems of interest.
