A multipurpose machine learning approach to predict COVID-19 negative prognosis in São Paulo, Brazil

Abstract

Introduction The new coronavirus disease (COVID-19) is a challenge for clinical decision-making and the effective allocation of healthcare resources. An accurate prognostic assessment is necessary to improve survival of patients, especially in developing countries. This study proposes to predict the risk of developing critical conditions in COVID-19 patients by training multipurpose algorithms. Methods A total of 1,040 patients with a positive RT-PCR diagnosis for COVID-19 from a large hospital from Sao Paulo, Brazil, were followed from March to June 2020, of which 288 (28%) presented a severe prognosis, i.e. Intensive Care Unit (ICU) admission, use of mechanical ventilation or death. Routinely-collected laboratory, clinical and demographic data was used to train five machine learning algorithms (artificial neural networks, extra trees, random forests, catboost, and extreme gradient boosting). A random sample of 70% of patients was used to train the algorithms and 30% were left for performance assessment, simulating new unseen data. In order to assess if the algorithms could capture general severe prognostic patterns, each model was trained by combining two out of three outcomes to predict the other. Results All algorithms presented very high predictive performance (average AUROC of 0.92, sensitivity of 0.92, and specificity of 0.82). The three most important variables for the multipurpose algorithms were ratio of lymphocyte per C-reactive protein, C-reactive protein and Braden Scale. Conclusion The results highlight the possibility that machine learning algorithms are able to predict unspecific negative COVID-19 outcomes from routinely-collected data.

Prediction models for COVID-19 disease outcomes

“…The algorithms were selected based on their demonstrated utility in healthcare prediction models [ 54–57 ] and evaluated using Scikit-Learn [ 58 ]. Random Forest, which has been commonly used for predicting hospitalizations [ 59 ] and ICU admission [ 60–62 ], is a decision tree-based model that can run multiple parallel trees. This technique averages predictions of all trees generated, allowing for a reduction of overfitting [ 63 ].…”

Effects of Increasing Levels of Total Tannins on Intake, Digestibility, and Balance of Nitrogen, Water, and Energy in Hair Lambs

“…Based on these results, the following calculations were performed: Digestible energy = Gross energy ingested − Gross energy of feces; and Metabolizable energy = digestible energy − gross energy lost in urine and gases. Meanwhile, gas energy was estimated as 8% of the total energy [35].…”

Machine Learning based COVID-19 Diagnosis from Blood Tests with Robustness to Domain Shifts

“…We additionally report the predictive ability for mortality risk of the COVID-19 positive samples on the basis of the blood tests only, again with no additional expensive features [33][34][35]38,40,41,53,54 . Compared to previous studies 33,36,37,39 , our mortality models are trained on a large number of COVID-19 positive patients.…”

“…COVID-19 and the patient's prognosis can be predicted from chest CT-scans, X-rays [11][12][13][14] or sound recordings of coughs or breathing [15][16][17] . Furthermore, it has been shown that ML models based on blood tests are capable of detecting COVID-19 infection [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32] and predicting other outcomes, such as survival or admission to an intensive care unit [33][34][35][36][37][38][39][40][41] .…”