Application of Machine Learning in Prediction of COVID-19 Diagnosis for Indonesian Healthcare Workers

The COVID-19 pandemic poses a heightened risk to health workers, especially in low- and middle-income countries such as Indonesia. Due to the limitations of implementing mass RT-PCR testing for health workers, high-performing and cost-effective methodologies must be developed to help identify COVID-19 positive health workers and protect the spearhead of the battle against the pandemic. This study aimed to investigate the application of machine learning classifiers to predict the risk of COVID-19 positivity (by RT-PCR) using data obtained from a survey specific to health workers. Machine learning tools can enhance COVID-19 screening capacity in high-risk populations such as health workers in environments where cost is a barrier to the accessibility of adequate testing and screening supplies. We built two sets of COVID-19 Likelihood Meter (CLM) models: one trained on data from a broad population of health workers in Jakarta and Semarang (full model) and tested on the same, and one trained on health workers from Jakarta only (Jakarta model) and tested on both the same and an independent population of Semarang health workers. The area under the receiver-operating-characteristic curve (AUC), average precision (AP), and the Brier score (BS) were used to assess model performance. Shapely additive explanations (SHAP) were used to analyse future importance. The final dataset for the study included 5,393 healthcare workers. For the full model, the random forest was selected as the algorithm choice. It achieved cross-validation of mean AUC of 0.832 ± 0.015, AP of 0.513 ± 0.039, and BS of 0.124 ± 0.005, and was high performing during testing with AUC and AP of 0.849 and 0.51, respectively. The random forest classifier also displayed the best and most robust performance for the Jakarta model, with AUC of 0.856 ± 0.015, AP of 0.434 ± 0.039, and BS of 0.08 ± 0.0003. The performance when testing on the Semarang healthcare workers was AUC of 0.745 and AP of 0.694. Meanwhile, the performance for Jakarta 2022 test set was an AUC of 0.761 and AP of 0.535. Our models yielded high predictive performance and can be used as an alternative COVID-19 methodology for healthcare workers in Indonesia, therefore helping in predicting an increased trend of transmission during the transition into endemic.