Deep Learning and Explainable Artificial Intelligence to Predict Patients' Choice of Hospital Levels in Urban and Rural Areas.

Maldistribution of healthcare resources among urban and rural areas is a significant challenge worldwide. People living in rural areas may have limited access to medical resources, and often neglect their health problems or receive insufficient care services. This research uses a deep learning approach to predict patient choices regarding hospital levels (primary, secondary or tertiary hospitals) and interpret the model decision using explainable artificial intelligence. We proposed an autoencoder-deep neural network framework and trained region-based models for the urban and rural areas. The models achieve an area under the receiver operating characteristics curve (AUC) of 0.94 and 0.95, and an accuracy of 0.93 and 0.92 for the urban and rural areas, respectively. This result indicates that region-based models are effective in improving the performance. The result is potentially leading to appropriate policy planning. Further interpretation can be done to investigate the explicit differentiation of the rural and urban scenarios.

Predicting the Travel Distance of Patients to Access Healthcare Using Deep Neural Networks.

OBJECTIVE: Improving geographical access remains a key issue in determining the sufficiency of regional medical resources during health policy design. However, patient choices can be the result of the complex interactivity of various factors. The aim of this study is to propose a deep neural network approach to model the complex decision of patient choice in travel distance to access care, which is an important indicator for policymaking in allocating resources. METHOD: We used the 4-year nationwide insurance data of Taiwan and accumulated the possible features discussed in earlier literature. This study proposes the use of a convolutional neural network (CNN)-based framework to make predictions. The model performance was tested against other machine learning methods. The proposed framework was further interpreted using Integrated Gradients (IG) to analyze the feature weights. RESULTS: We successfully demonstrated the effectiveness of using a CNN-based framework to predict the travel distance of patients, achieving an accuracy of 0.968, AUC of 0.969, sensitivity of 0.960, and specificity of 0.989. The CNN-based framework outperformed all other methods. In this research, the IG weights are potentially explainable; however, the relationship does not correspond to known indicators in public health. CONCLUSIONS: Our results demonstrate the feasibility of the deep learning-based travel distance prediction model. It has the potential to guide policymaking in resource allocation. Clinical and Translational Impact Statement- Deep learning technology is feasible in investigating the distance that patients would travel while accessing care. It is a tool that integrates complex interactive variables with highly imbalanced data distributions.

Adaptation and validation of a measure of health literacy in Taiwan: The Newest Vital Sign.

BACKGROUND: Health literacy (HL) refers to the ability to obtain, read, understand, and use basic health care information required to make appropriate health decisions and follow instructions for treatment. The Newest Vital Sign (NVS) is an instrument developed for assessing aspects of HL relevant to reading and numeracy skills. This study aimed to develop a traditional Chinese version of the NVS (NVS-TC) and assess its feasibility, reliability, and validity in Taiwanese patients with type 2 diabetes. METHODS: The original NVS was translated into traditional Chinese in accordance with established guidelines. A cognitive testing procedure was subsequently performed to evaluate the ease of understanding and acceptability of the test in 30 patients with diabetes. Thereafter, a quantitative survey (N = 232) was administered for validating the NVS-TC against the accepted standard tests of HL and participant education level. RESULTS: The internal consistency (Cronbach's α) was 0.76. In accordance with a priori hypotheses, we found strong associations between the NVS-TC and objective HL and weaker associations between the NVS-TC and subjective HL. The known group validity of the NVS-TC was demonstrated through multivariate regression analyses, which showed that educational differences in the NVS-TC scores remained significant after adjustment for age, gender, and working in healthcare. CONCLUSIONS: The results suggest that the NVS-TC is a reliable and valid tool that facilitates international comparable HL research in Taiwan. The NVS-TC can be used to investigate the role of HL in health care and can be easily incorporated into daily clinical practice for diabetes management.

Stages of change concept of the transtheoretical model for healthy eating links health literacy and diabetes knowledge to glycemic control in people with type 2 diabetes.

AIMS: Health literacy has been recognized as a key construct associated with clinical outcomes; however, few studies have explored the mechanism underlying the association. The transtheoretical model (TTM) has long been considered a useful conceptualization in the process of intentional behavior change. Stages of change lies at the heart of the TTM as studies of change have found that people move through a series of stages when modifying behavior. This study focuses on the role of knowledge and stages of change (SOC) as serial mediators linking health literacy to glycemic control. METHODS: In this cross-sectional survey, a total of 232 patients with type 2 diabetes participated in this study. Participants completed questionnaires for assessing health literacy, readiness to consume healthy foods, and a dietary knowledge test specific to diabetes. RESULTS: Low health literacy was significantly associated with worse glycemic control. Statistical evaluation supported the serial mediation model, in which knowledge and SOC formed a serial mediation chain that accounted for the indirect effect of health literacy on glycemic control. In other words, dietary knowledge significantly motivated participants to move into the later stages of behavior change, which in turn improved the outcome of glycemic control. CONCLUSIONS: The results indicate that the ordering of mediators in the pathway between health literacy and health outcome may be complex, help explain the conflicting results of the past, and form a basis for the development of interventions promoting self-management of diabetes through glycemic control.