A Natural Language Processing-Driven Approach to Developing a Self-Health Management Scale for Healthcare Workers: A Psychometric Study (Preprint)

Abstract

BACKGROUND Healthcare workers face significant mental health challenges due to occupational stressors, yet there is a lack of assessment tools specifically designed for their self-health management. Traditional scale development methods often rely on researcher-led item generation, which may not fully capture the real-world concerns of this unique population. The emergence of Natural Language Processing (NLP) offers a novel, data-driven approach to overcome this limitation. OBJECTIVE To develop and validate a self-health management scale for healthcare workers based on natural language processing (NLP) techniques, aimed at scientifically assessing their health management capabilities. METHODS Multi-source textual data were collected using web crawling technology. Health dimensions were extracted through methods such as LDA topic modeling and sentiment analysis to construct the initial questionnaire. A survey was conducted among 699 healthcare workers, followed by item analysis, exploratory factor analysis, and reliability and validity testing. RESULTS The final scale comprised 20 items divided into two dimensions: "Occupational Stress and Health Response" and "Occupational Resources and Adaptability," explaining a cumulative variance of 60.287%. The Cronbach’s alpha coefficient was 0.940, with all reliability and validity indicators demonstrating good psychometric properties. CONCLUSIONS The developed scale exhibits strong psychometric characteristics and is suitable for assessing self-health management among healthcare workers, providing a scientific basis for health interventions.