Background: A variety of institutional and technological factors have led to reduced lengths of stay (LOS) among patients hospitalized for acute myocardial infarction (MI) over the past two decades, but large international variations in LOS remain. One factor influencing LOS variation may be the availability of post-acute care facilities. When such facilities are unavailable, patients must be discharged in good enough condition to receive care at home. We explore this phenomenon by examining the influence of discharge destination on LOS for MI in 9 countries. Methods: Data are from a multi-site, international registry of patients who presented to acute care hospitals with MI. We used a 2-stage model to estimate the impact of LOS on the likelihood of discharge to home while controlling for patient demographics, medical history, MI severity at presentation, and country. Predicted values of LOS were generated by regressing LOS against the variables listed above, plus the sum of Medicare resource value units (RVUs) associated with procedures performed during the stay. A linear probability model was used to model discharge status as a function of predicted LOS and other variables (excluding RVUs), including country-LOS interactions. Results: Of 5573 observations, 380 patients that died during hospitalization, 435 with missing or invalid lengths of stay, 12 awaiting transplantation and 641 with missing or invalid discharge disposition were excluded, leaving a total of 4105 observations. Mean age was 65.9 years (SD 13.7); 66.8% were male. Mean LOS for the sample was 9.0 days (SD 8.8). Mean LOS was shortest for the US and longest for Canada (7.6 versus 12.0 days, p<0.0001). On average, 86.2% of patients were discharged home, 6.7% to a long-term care facility, and 7.1% to another acute care facility. Patients were most likely to be discharged home in Argentina and least likely to be discharged home in Germany (96.3% vs 59.3%, p<0.0001). The sign of the LOS coefficient in the 2nd stage model confirms the hypothesis that increased LOS is associated with a greater probability of discharge home, independent of country effects (β=0.033, p<0.0001). The likelihood of discharge home with increased LOS was significantly greater for Australia. Other significant variables at p<0.10 included age, gender, race, weight, history of diabetes and previous MI, ST-segment elevation MI, Killip class, and hospitalization in Australia, Canada, Germany, and the Netherlands. Conclusions: These data confirm large differences in LOS at the country level and indicate that patients with longer LOS are likelier to be discharged home than to another facility. This analysis is subject to several limitations, including small sample sizes in several countries, missing discharge data for ~10% of observations, and lack of post-discharge survival data. While our ability to definitely address potential factors affecting LOS across countries is limited, our results suggest a potential inefficient use of health care resources, since acute care hospitalization is comparatively more expensive than skilled nursing or rehabilitation care. Further, this analysis shows the importance controlling for the endogeneity of LOS
