Predicting Hospital Employees' Quality of Life using Explainable Machine Learning on Psychosocial Work Environment Data

Nida Aslam^1*Arwa Alumran¹Bashayer Alshahrani²Irfan Ullah Khan¹Rana Alshedayed¹Dina Alfrayan¹Rand Alessa¹Samiha Mirza¹Fatimah Alshakhs¹

• ¹Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia
• ²Johns Hopkins Aramco Healthcare (JHAH), Dhahran, Saudi Arabia

Abstract: Health-Related Quality of Life (HRQL) embodies the impact of an individual's health on their ability to live a fulfilling life. Quality of Life (QoL) is influenced by a range of factors, including physical functioning and well-being, psychological functioning, work environment, lifestyle, and social relations. Various studies have found that job-related factors can be an essential predictor of an individual's HRQL. Furthermore, the Psychosocial Work Environment (PWE) can affect workers' well-being and contribute to the company's sustainability. In the context of healthcare providers, the quality of health services provided is affected by PWE and QoL. Therefore, the relationships among QoL, PWE, and healthcare quality need to be assessed to identify factors that improve overall patient healthcare service quality. This relationship has not been extensively evaluated in the Saudi context. Therefore, in the current study, we aim to employ machine learning (ML) techniques to predict employee QoL using PWE data from a hospital in the Kingdom of Saudi Arabia (KSA). Several ML models have been developed to predict HRQL effectively and their . The study aims to identify the significant attributes; thus, the several experiments were carried out, with and without feature engineering. The Naïve Bayes (NB) classifier achieved the highest precision of 1.0 (95% CI: 0.81-1.0) in predicting employees' QoL using PWE and demographic variables. Whereas the selected Work Environment (WE) features using the Xverse voting selector with a SVM classifier achieved the best results, with accuracy, recall, precision, F1, and receiver operating characteristic (ROC) reaching 0.92 (95% CI: 0.88-0.95), 0.90 (95% CI: 0.86-0.98), 0.95 (95% CI: 0.86-0.99), 0.92 (95% CI: 0.88-0.95), and 0.9, respectively. Post-hoc Explainable Artificial Intelligence (XAI) was used to alleviate the black-box nature of SVM and add transparency to the model. In conclusion, this study provides a robust, explainable tool for predicting employee QoL that can help healthcare organizations to improve quality.