Actors Associated with Deaths of Covid-19 Patients: Case study of Rayong Hospital

Abstract

Background : The outbreak of Coronavirus Disease 2019 (COVID-19) has become a significant public health crisis with widespread global impacts. In Thailand, the number of confirmed cases has continued to rise, leading to major challenges in the healthcare system, including limited hospital bed capacity, shortages of medical resources, and constantly evolving treatment guidelines. Screening and risk assessment of patients have therefore become crucial steps for prioritizing treatment and allocating resources effectively. However, most reference data are derived from international studies, which may not accurately reflect the context of Thailand's population and healthcare system.

Objective : To analyze the factors associated with mortality among COVID-19 patients using appropriate predictive models for estimating the probability of death.

Material and methods : This quantitative study utilized data from COVID-19 patients admitted to and discharged from Rayong Hospital between April 1, 2021, and January 31, 2022. Data were retrieved from the hospital’s Health Information System (HIS). The factors studied were categorized into three groups: (1) Personal characteristics (e.g., gender, age, nationality, body mass index), (2) Initial clinical presentation (e.g., chest X-ray results, blood pressure, respiratory rate, and blood oxygen saturation), and (3) Treatment-related factors (e.g., medication usage, mechanical ventilation, underlying diseases, and complications recorded in medical records). Data were analyzed using descriptive statistics and three classification methods: binary logistic regression, decision trees, and multilayer perceptron (MLP) neural networks.

Result : The MLP neural network model, consisting of three hidden layers using ReLU and Sigmoid activation functions, provided the highest predictive performance with an accuracy of 0.95. The most influential factors associated with mortality among COVID-19 patients were invasive mechanical ventilation, chronic kidney disease, age, the number of days from diagnosis to discharge, and the use of hydrocortisone.

Conclusion : Factors positively associated with mortality included advanced age, underlying conditions such as diabetes and hypertension, low blood oxygen saturation, and delayed access to treatment—particularly among patients requiring mechanical ventilation. The findings of this study can be used to develop screening protocols and care strategies for high-risk patients.