Exploring the relative importance of the factors associated with menopausal symptoms using a random forest model: a cross-sectional study

Abstract

Purpose: This study aimed to identify key factors associated with menopausal symptoms among middle-aged women by examining a comprehensive set of physical, psychological, and lifestyle variables. Methods: A cross-sectional study was conducted with 94 women aged 45 to 55 years in Seoul, South Korea. Data were collected through physical assessments, self-reported questionnaires, and monitoring using a wearable device (Fitbit Charge 5, Fitbit Inc.). The Menopause Rating Scale was used to assess symptom severity, with scores dichotomized into no/mild symptoms (0–8) and moderate/severe symptoms (9–44). Random forest analysis was applied to evaluate the relative importance of various factors in relation to menopausal symptoms. Results: Fourteen significant predictors were identified from an initial set of 57 variables using recursive feature elimination with cross-validation. The final random forest model achieved balanced predictive performance, with an accuracy of 74.1%, an area under the curve of 75.7%, sensitivity of 78.9%, and specificity of 62.5%. Age emerged as the most influential predictor, followed by psychological well-being and loneliness as the second and third most important factors. Among physical characteristics, relative grip strength (fourth) and body fat percentage (fifth) were significant predictors. Lifestyle factors, including moderate physical activity (sixth) and health-conscious dietary behaviors (seventh), showed moderate importance, while socioeconomic factors demonstrated lower importance. Conclusion: The findings highlight the multifaceted nature of menopausal symptoms and suggest that effective management strategies should incorporate physical, psychological, and lifestyle interventions. These results provide evidence for developing comprehensive digital healthcare applications that incorporate monitoring and intervention features across multiple domains for effective menopausal symptom management.