Effect of hospital assignment on mortality for AMI patients

International guidelines recommend percutaneous coronary intervention (PCI) to treat acute myocardial infarction (AMI) if PCI can be performed within two hours. PCI is a centralized treatment, and therefore a common trade-off is whether to send AMI patients directly to a hospital that performs PCI, or postpone a potential PCI-treatment by first receiving acute treatment at a local hospital that can not perform PCI. In this paper, we estimate the effect of sending patients directly to a PCI-hospital on AMI mortality. Using nation-wide individual-level data from 2010 to 2015, we studied mortality rates for AMI patients sent directly to a hospital that performs PCI (N=20 336) compared to AMI patients sent to a hospital not performing PCI (N=33 437). Since the underlying health of patients may affect both hospital assignment and mortality, estimates from traditional multivariate risk adjustment models are likely biased. We therefore apply an instrumental variable (IV) model using the historical municipal share sent directly to a PCI-hospital as an instrument for being sent directly to a PCI-hospital. Patients sent directly to a PCI-hospital are younger and have fewer comorbidities than patients who are first sent to a non-PCI-hospital. IV results suggest that those initially sent to PCI-hospitals have 4.8 percentage points decrease (95% CI (- 18.1)-8.5) in mortality after one month compared to those initially sent to non-PCI-hospitals. Our IV results suggest that there is a non-significant decrease in mortality for AMI patients sent directly to a PCI hospital. The estimates are too imprecise to conclude that health personnel should change their practice and send more patients directly to a PCI-hospital. Moreover, the results may be taken to suggest that health personnel navigate AMI patients to the best treatment option.