Estimating out-of-hospital mortality due to myocardial infarction

We developed a model to estimate out-of-hospital deaths due to myocardial infarction (MI), which was based on a detailed database of MI admissions to Pennsylvania hospitals during 1998. Our estimation method addresses the problem of geographical selection bias in inpatient databases, which occurs when MI patients with poor geographic access are undersampled. A Geographic Information System (GIS) was used to determine travel times between hospitals and patients, based on patients' zip code of residence. Nearness to a hospital was positively associated with in-hospital mortality (P<0.01) and emergency admissions (P < 0.01) and negatively associated with out-of-hospital mortality (P < 0.01). Model predictions were made for a range of input values and validated using empirical data.