Machine learning models predict blood clots in hospitalized adults with high accuracy

Hospitalized patients face a real risk of deep vein thrombosis, a dangerous blood clot that can form in the legs. Doctors need reliable ways to spot this danger early. A recent analysis looked at machine learning tools designed to predict these clots before they cause harm. These computer models use patient data to calculate the chance of a clot forming. The results were promising. When researchers combined data from 17 studies, the models showed an accuracy score of 0.85. This number suggests the tools work well at identifying who is at risk. The analysis included 10 of the studies in a detailed comparison. Safety was not reported in the original research, so we do not know about side effects or risks from using these tools yet. However, the study has important limits. Six of the studies had a high risk of bias. This means their results might be less trustworthy. There was also poor reporting on how the models were tested in new groups of patients. Because of these gaps, doctors cannot yet rely on these tools for everyday care. The findings highlight a need for better reporting and more rigorous testing before these models become standard practice.