ABSTRACT
The novel coronavirus disease (COVID-19) emerged in late 2019 has shown that research done with open data could bethe cornerstone for overcoming the need for collaborative, optimized and urgent analysis. Although several articles have beenpublished, identification of variables that can have correlation with positive PCR results is still a challenge. In this paper weshow a concrete example of open data analysis from 910 patients attended in the hospital undergoing SARS-CoV-2 RT-PCR inthree private institutions in S ̃ao Paulo, Brazil. We performed an exploratory analysis using principal component analysis, featureselection and predictive algorithms to test for associations between a number of laboratory test abnormalities and the SARS-CoV-2 RT-PCR result. More concretely, we found a set of 18 variables that showed some association with a positive PCR result.Among these variables elevated lactic dehydrogenase (LDH) and d-dimer were the most correlated with a positive RT-PCR. Wedeveloped a classifier that achieved 76% mean accuracy, 77% mean precision and 92% mean sensitivity to identify individualswith COVID-19