Deep Learning Approaches For Cost-Benefit Analysis Of Vision And Dental Coverage In Comprehensive Health Plans

Abstract

We propose to use two very recent deep learning approaches, including vision transformer and deep variational continuous factor analysis, for cost-benefit analysis of seeking health insurance coverage for dental and vision in addition to comprehensive medical coverage. We based our analyses on medical Expenditure Panel Survey data, and the classification results reached 0.972, 0.979, 0.983, and 0.992 in terms of the area under the receiver operating characteristic curve, the area under the precision-recall curve, accuracy, and F1 score, respectively. We also found that lower dental and vision risk is associated with older age. We finally argue that our work has the potential to improve decision support algorithms healthcare providers use in order to differentiate attracting patients who would be profitable from those who would not. This paper presents trade-offs of seeking health insurance coverage for dental and vision in addition to comprehensive medical coverage. To carry out the analysis, we present two case studies of cost-benefit analyses using the consecutive Medical Expenditure Panel Survey data from 2016 to 2020: (1) a classification analysis of making dental and vision appointments versus not making dental and vision appointments among women for preventive care; and (2) a hierarchical clustering analysis for dental and vision diagnosis-classified patterns of seeking, obtaining, delaying, and avoiding appointments. To obtain accurate and dependable analysis results as much as we can, we utilized the deep learning algorithms in supervised and unsupervised data analysis and showed a deep learning-based cost-benefit analysis health informatics system. The high prediction findings show the reasonableness and potential usefulness of the deep learning and deep variational continuous factor analysis vision dental cost-benefit analysis amidst its cost negativity. We argued that future work would investigate other advanced deep generative deep learning algorithms.