A prediction model for prehospital triage of patients with suspected cardiac ischemia

The American College of Cardiology recommends that patients with high risk acute myocardial infarction (AMI) be triaged to hospitals with percutaneous transluminal coronary angioplasty capability. However, there are no prehospital triage criteria to select candidates for bypassing community hospitals and being taken directly to "cardiac centers." This article assesses which independent variables predict death within 7 days in patients with suspected AMI transported by EMS. This is a retrospective study of 291 AMI patients transported by ambulance to 3 hospitals during 1996-1997. Included were patients who were (n = 244) > or =18 years of age, had a ED chief complaint of chest pain or dyspnea for whom we had mortality data. Mortality at 7 days, our primary outcome measure, was obtained by using a metropolitan Detroit tricounty death index records. Differences between the survivors and nonsurvivors were assessed using the Student's t-test and chi-square tests. Multiple triage criteria were assessed for optimal identification of high risk patients by constructing a logistic multivariate model. Among the study population, 15% died within 7 days (95% confidence interval (CI) 10.3-19.2), and this group represented 63.2% of all deaths over a 2 year surveillance period. Survivors, compared to nonsurvivors, were 14.1 years younger (P < or = .001) and more often men (P < or = 0.001). The dispatch time to ED arrival was less among survivors than nonsurvivors (42.8 vs. 50.6 min, P < or = .01). EMS vital signs differed by survivor status. Among survivors, HR was lower (-11.9 bpm; P < or = 0.01), RR was lower (-6.7 rpm; P < or = .001), SBP was higher (+14.5 mmHg; P < or = 0.05) and DBP was higher (+13.2 mm Hg; P < or = .01). A multivariate model identified the following as independent predictors of early mortality: female gender (OR = 2.3; P < or = .05), age > or =65 (OR = 5.9; P < or = .01), RR > or = 20 (OR = 4.6; P < or = .001), SBP < 120 (OR = 2.4; P < or = .05). The overall model was 86% sensitive and 53% specific with an area under the receiving operating characteristic curve of 0.8 (P < or = .001). A triage rule based on a multivariate model can identify the group at high risk of early cardiac death. This decision rule needs to be prospectively validated.