Prediction of COVID-19 patients' participation in financing informal care using machine learning methods: willingness to pay and willingness to accept approaches.

Background: Informal care plays an essential role in managing the COVID-19 pandemic. Expanding health insurance packages that reimburse caregivers' services through cost-sharing policies could increase financial resources. Predicting payers' willingness to contribute financially accurately is essential for implementing such a policy. This study aimed to identify the key variables related to WTP/WTA of COVID-19 patients for informal care in Sanandaj city, Iran.

Methods: This cross-sectional study involved 425 COVID-19 patients in Sanandaj city, Iran, and 23 potential risk factors. We compared the performance of three classifiers based on total accuracy, specificity, sensitivity, negative likelihood ratio, and positive likelihood ratio.

Results: Findings showed that the average total accuracy of all models was over 70%. Random trees had the most incredible total accuracy for both patient WTA and patient WTP(0.95 and 0.92). Also, the most significant specificity (0.93 and 0.94), sensitivity (0.91 and 0.87), and the lowest negative likelihood ratio (0.193 and 0.19) belonged to this model. According to the random tree model, the most critical factor in patient WTA were patient difficulty in personal activities, dependency on the caregiver, number of caregivers, patient employment, and education, caregiver employment and patient hospitalization history. Also, for WTP were history of COVID-19 death of patient's relatives, and patient employment status.

Conclusion: Implementing of a more flexible work schedule, encouraging employer to support employee to provide informal care, implementing educational programs to increase patients' efficacy, and providing accurate information could lead to increased patients' willingness to contribute and finally promote health outcomes in the population.