A major component of the growth in size and cost of the U.S.'s health care system is the increasingly aggressive treatment of acute and chronic disease at more advanced stages of illness. Most of this care is concentrated in hospitals an much is located within Intensive Care Units, whose size and contribution to total health care costs is increasing dramatically. The goal of this proposal is to advance the analytic capabilities of basic health services research as they relate to survival analysis or risk prediction for critically ill hospitalized adults (CIHAs) treated in hospitals and ICUs and to link these findings to the evaluation of patient outcome, costs, quality of care, utilization and practice patterns and, ultimately, to the improvement of clinical decision making. The long term objective is to use clinically accurate data bases to explore limitations and shortcomings of current efforts at survival analysis and risk prediction and suggest improvements. Such progress will be of substantial benefit to researchers, quality assurance for CIHAs. We will accomplish this by making use of two contemporary, clinically accurate data bases containing detailed disease and physiologic information on over 25,000 CIHAs. Within these data bases we will determine how variations in missing values, selection of the patient for treatment, and the interaction of major predictive components influence the overall explanatory power of survival analyses. We will also explore how various multivariable techniques and length of follow up influence determinations of the quality, cost-efficiency and appropriateness of care. From these analyses we will propose analytic approaches, develop guidelines, and suggest standards that could be used to evaluate the quality of evidence rom inter-institutional evaluations of CIHA as well as evaluating the usefulness of individual patient outcome predictions for use in conjunction with traditional clinical decision making.
