In proposing a one-year continuation of our grass% 'Assessment of Coronary Care Unit (CCU) Use in Different Hospitals"""""""" (NCHSR grant #R01HS05549), we have two primary goals: 1) to appropriately execute the one remaining of our five originally-planned studies, """"""""Who Admits to the CCU?', and 2) to further develop the methodology for assessing performance of logistic regression models. Project 1: In our original proposal's three-year plan, we proposed to complete five studies: 1) Who admits to the CCU?; 2) CCU admitting practices and severity of AIHD at teaching and non-teaching hospitals; 3) Non-clinical (sociodemographic) factors affecting CCU admitting practices; 4) Clinical factors with inordinate impact on CCU admission decision making; 5) The impact of the lack of hospitalization on patients with acute myocardial infarction. The funding ultimately granted was for two years, and we do expect to complete all but one of these studies within this period. However for the study """"""""Who Admits to the CCU?"""""""", which describes and examines the impact of different physician types on CCU triage decisions, more time is needed. After analysis of our triaging physician variable, it is apparent that to avoid generating misleading results in this currently hotly-debated area, more data about the structure and process of physician participation in care, and supervision, are needed. Thus a major undertaking in the proposed third year of this grant will be returning again to each of the six study sites to obtain such data to support an accurate patient-level analysis for this study. Project 2: Although logistic regression has been used for 20 years, and has mushroomed in its use in health services, epidemiologic, and medical research, methods for assessing logistic models' performance are still not well-developed. Having been involved in early efforts in developing and testing logistic regression, clinical decision aids, and having a uniquely-suited database for developing and testing logistic regressions, in the proposed third year, we plan to rigorously explore and develop the methodology for testing their performance. Project 1 should provide important information on the imp act on triage practices of emergency room physician type and level of trainee supervision. Project 2's results should be of use to those who develop and test predictive probabilistic predictive methods, and to those using logistic regression in health services, epidemiologic, and other research.
