This study developed and tested a tool to assess the likelihood of patients having acute cardiac ischemia and thus the appropriateness of admitting them to the coronary care unit (CCU). It is valid both for real-time clinical use and for retrospective review: a time-insensitive predictive instrument (TIPI). The authors' earlier acute ischemia predictive instrument, not designed specifically to support retrospective use, could not offer the advantage of a single tool usable by both clinicians and reviewers of care. Over a two-year period, the authors prospectively collected data on 5,773 emergency room patients seen in six New England hospitals for symptoms suggesting acute cardiac ischemia. In the Developmental Phase, based on 3,453 such patients seen during the first year, the authors developed a logistic regression-based TIPI for acute cardiac ischemia. Using seven clinical features reliably ascertainable both in the emergency room setting and by medical record review, the TIPI expressed a patient's probability of having acute ischemia as between 0-100%. In this phase, a risk category system based on the TIPI scale was also devised, which created four similar-sized groups, by cutting at 10%, 25%, and 55%. In the Test Phase, when prospectively tested on the 2,320 emergency room patients seen during the second year, the TIPI showed excellent diagnostic performance. Its receiver-operating characteristic (ROC) curve area of 0.88 was comparable to the original predictive instrument and the ROC curve path suggested performance comparable to physicians as well. Its slope of the relationship between predicted and observed probabilities of having acute ischemia was 1.11 (R2 = 0.97) with a correlation of 0.99 (P < 0.0001), suggesting excellent calibration of predictions throughout the probability range. For patients who proved to have acute ischemia, the average TIPI probability was 59%, whereas for those without ischemia, the average TIPI value was 21% (P < 0.0001). This differentiation was maintained even for those given different (including inappropriate) triage to the CCU, ward, or home (P < 0.0001 for each disposition). When the performance of the four TIPI-based risk groups was prospectively tested on year-two patients, among the 552 patients in the low probability group, only 1.6% had acute cardiac ischemia, including only 0.7% with acute infarctions. Among the 484 patients in the high probability group, 81.6% had acute ischemia, and 53.3% acute myocardial infarctions, suggesting these to be clinically relevant groups for aiding or assessing emergency room triage. In the Hospital Comparison Phase, mean TIPI probabilities of acute ischemia were computed for each hospital's emergency room and CCU patients. The overall study mean TIPI value for emergency room patients was 33% (range among hospitals from 30-43%), and the overall mean TIPI value for CCU patients was 50% (hospital range 42-69%). This 27 point difference between the emergency room and CCU values reflects CCU admission selectivity, representing a 52% increased probability of acute ischemia among CCU patients (hospital range 34-81%). The mean TIPI value for emergency room patients sent home was 18%, (hospital range 16-21%). When hospitals were compared using their TIPI values for patients seen in the emergency room, admitted to the CCU, or sent home, one had values significantly different from the others (P < 0.001). The TIPI is accurate and reliable, and should be useful for real-time CCU triage decisionmaking and also for quality assurance and cost-containment efforts. This dual function may enhance cooperation between clinicians, institutions, payors, and review agencies.
