This large study examined how well the HEART score predicts major adverse cardiovascular events within 30 days for adults presenting with chest pain. Data came from six different emergency departments involving over 57,000 encounters. The team compared a federated diagnostic meta-analysis, which keeps patient data local, against a centralized patient-level analysis that pools all data together.
The federated method yielded a discrimination score of 0.759, while the centralized analysis scored 0.785. These numbers are very close, suggesting that federated approaches can provide reliable risk-stratification estimates without moving sensitive patient information to a central server. This alignment allows for scalable evaluation of risk tools while accommodating differences between various practice settings.
No safety concerns were reported because the study focused on diagnostic accuracy rather than medication side effects. Readers should take from this that federated analysis is a promising way to evaluate tools like the HEART score across different hospitals. However, the study notes modest variability between sites and differences in how often outcomes occur, which means results may vary in other locations.
