Multimorbidity (people having more than one chronic disease) is a major challenge for health systems and hospital institutions, making it necessary to identify patients at higher risk and create prevision models to adjust the response capacity. This implies knowing how to evaluate reality – which was what the Data Sciences team of Hospital da Luz Learning Health did, recently publishing the results from different existing models over 15 years of clinical data at Hospital da Luz Lisboa concerning patients assisted in the Emergency Unit. “ Multimorbidity measurement strategies for predicting hospital visits ” is the title of the study, published last December in journal “BMC Health Services Research”. Resulting from the collaboration between Hospital da Luz Learning Health, INESC-ID and Instituto Superior Técnico following the Intelligent Care project (CMU-Portugal), the study is authored by Bernardo Neves , José M. Moreira, Simão Gonçalves, Jorge Cerejo, Inês Mota, Nuno André Silva, Francisca Leite, and Mário J. Silva. This team has applied three phenotyping strategies and five multimorbidity indexes to data concerning over 925 thousand patients and 9.7 assisting episodes at Hospital da Luz Lisboa, between 2007 and 2022. The goal was to compare and evaluate the different multimorbidity measurement models. It was concluded that: More complete clinical records allow to better identify patients at risk. Crossing information from several sources of electronic health records allows to detect more accurately those who suffer simultaneously from various diseases. The better the identification, the more efficient the prediction of those who will need to resort to the hospital. Considering the “weight” of diseases is more useful than simply counting them. It is not sufficient to know how many diseases a patient has – it is much more important to consider the severity and impact of each disease. The study shows that the methods that weight diseases predict visits to hospital better, although the best approach varies according to the type of hospitalization. Combining the history of illnesses with the history of health services use provides the best results. The combination of the chronic disease profile of a patient and his record of consultations and previous hospitalizations is the more effective strategy to anticipate future visits to hospital. The authors defend that this approach should be systematically adopted to early identify patients at higher risk and thus improve health resources management. In the photo above, a few of the authors of the study: Francisca Leite, Bernardo Neves, Mário J. Silva, José M. Moreira, Jorge Cerejo, and Inês Mota.
