The professor of the master's degree in economics, Roberta Wichmann, published together with researchers from USP and members of the IACOV-BR network, the article entitled Improving the performance of machine learning algorithms for health outcomes predictions in multicentric cohorts in the Nature Scientific Reports Magazine . Scientific Reports is the fifth most cited journal in the world, with more than 696,000 citations in 2021, and receives widespread attention in policy documents and the media. ( https://www.nature.com/srep/ ).

The idea for the research arose when the authors wondered whether the predictive performance of machine learning models could be improved just by adding more data to training in cases of predictions in the healthcare area. In other words, would it be possible to generalize a model to different hospitals in a state or different regions of Brazil? Therefore, the authors tested different training strategies for machine learning models, starting with local training with data from a single hospital, to different forms of aggregation with data from other hospitals, to evaluate the predictive performance in identifying the risk of death due to COVID. -19 and if so, would it be possible to generalize the prediction in different regions of the country. 

How was the study carried out? 8 different strategies were then created and data from 18 hospitals from all regions of Brazil were used to train 3 machine learning models (xgboost, catboost and lightgbm) in order to identify the best strategy that maximized predictive performance. 

“ In our study, the best strategy was training with data from a single hospital, obtaining the best performance in 11 (61%) of the 18 hospitals, although in a few cases, there was better predictive performance when aggregating more data ” , he says Professor Roberta . Therefore, using data from a single hospital can result in better performance than adding data from different hospitals with different protocols and socioeconomic differences.

Congratulations Professor Roberta and all the authors and partners involved!