Machine Learning Predicts Subjective Cognitive Decline During Menopause: Novel Framework for Early Detection

A machine learning model developed to detect severe subjective cognitive decline (SCD) in nurses at midlife going through the menopause transition identified menopausal symptoms and menopausal stage as the strongest predictors of cognitive complaints, with performance of the model suggesting it reliably detects high-risk individuals.¹

Ping Li, PhD
Courtesy of Shandong University

In a cross-sectional analysis of 1,264 Chinese nurses aged 40 to 60, a support vector machine (SVM) model achieved an area under the curve (AUC) of 0.846, 78.9% accuracy, 75.3% sensitivity, and 80.2% specificity, using 13 selected features. The findings, published in the journal Menopause, suggest a scalable framework for early detection of cognitive vulnerability in women experiencing menopause-related neuropsychological changes, the authors said.¹

The study was conducted by a team from Shandong University and focused on nurses, a population known to experience high occupational stress and heightened vulnerability to SCD during the menopause transition.² SCD, defined as a self-reported decline in memory or cognition not evident on objective testing, is increasingly recognized as a potential early marker of future cognitive deterioration, including risk for Alzheimer disease.³ Given “high-intensity physical and mental demands of their work, long-term irregular schedules, and the physiological and psychological changes brought about by menopause, nurses face particularly pronounced challenges during this transition,” lead author Ping Li, PhD, professor and associate dean for graduate education, school of nursing and rehabilitation, Shandong University, China, and colleagues wrote. They emphasize the cohort represents a key population for early cognitive screening.

The researchers used secondary analysis of data gathered via questionnaire from nurses working at 16 hospitals. Potential participants were excluded if they had significant medical or psychiatric comorbidities or objective cognitive impairment. Participants completed the SCD-Q9 questionnaire, which quantifies cognitive concerns on a 0–9 scale. Nurses with scores of 7.5 or greater, corresponding to the 75th percentile, were categorized as having severe SCD. The sample was randomly split into training and validation sets (75% and 25%, respectively), and the Boruta algorithm was applied to identify key predictors among 24 sociodemographic, work-related, menstrual, lifestyle, and mental health variables.

Machine learning models, variables. Li and colleagues constructed 7 machine learning models, optimizing them using 5-fold cross-validation. Among them, the support vector machine (SVM) model outperformed alternatives, including random forest, logistic regression, and k-nearest neighbors. The most influential variables, determined via SHAP (Shapley Additive Explanations) analysis, were severity of menopausal symptoms, menopausal stage, economic status, sleep satisfaction, and positive emotion. The researchers also identified nonlinear associations, including a plateau effect in the impact of worsening menopausal symptoms on cognitive complaints.

FINDINGS

Li and team found that women with more intense vasomotor and psychological menopausal symptoms were significantly more likely to report severe SCD. SHAP plots illustrated that as symptom burden rose, in particular sleep disturbance, mood swings, and hot flashes, the likelihood of SCD increased sharply before leveling off. The stage of menopausal transition also contributed independently; risk was highest during the menopause and early postmenopausal stages. Nurses with limited financial resources or poor sleep quality were also more likely to report cognitive concerns. Conversely, Li et al reported that greater resilience and higher levels of positive emotion were associated with lower SCD risk.

Among the 1,264 respondents (mean age 46 years), the average SCD score was 5.38, with 26.9% classified as experiencing severe SCD. Notably, nearly 37% of those with severe SCD had chronic medical conditions, and 25% had experienced a major life event in the past year. Sleep patterns also had an impact with duration of 5 hours or less and low sleep satisfaction found to be more common among those with higher cognitive complaints. Emotional well-being also played a substantial role: higher negative affect and neuroticism scores were strongly linked with SCD.

Among the study’s limitations, the authors acknowledged that the sample consisted entirely of nurses in a single province in China, which may limit generalizability. The cross-sectional design precludes causal inference, and the reliance on self-reported data introduces subjectivity. Moreover, objective cognitive testing was not used to confirm impairment, and biological factors such as hormone levels or neuroimaging were not assessed.

Novel guidance for interventions. Nonetheless, the study demonstrates the utility of machine learning and SHAP analysis in stratifying risk for cognitive decline using accessible, noninvasive data. By integrating validated questionnaires and statistical modeling, the approach enables early identification of at-risk individuals. Further, the authors said, their findings “provide a novel guidance for interventions designed to preserve cognitive health in women undergoing the menopause transition. Future research should aim to further validate these results and investigate additional potential influencing factors to develop more effective strategies for promoting cognitive health during the menopause transition,” they concluded.

References

1. Siangyu A, Shen X, Jia F, He X, Zhao D, Li P. Using machine learning models to identify severe subjective cognitive decline and related factors in nurses during the menopause transition: a pilot study. Menopause. 2025;23:00. doi: 10.1097/GME.0000000000002500

2. Hayashi K, Ideno Y, Nagai K, et al. Complaints of reduced cognitive functioning during perimenopause: a cross-sectional analysis of the Japan Nurses' Health Study. Women’s Midlife Health. 2022;8:6. doi: 10.1186/s40695-022-00076-9

3. Reuben R, Karkaby L, McNamee C, Phillips NA, Einstein G. Menopause and cognitive complaints: are ovarian hormones linked with subjective cognitive decline? Climacteric. 2021;24:321‐332. doi:10.1080/13697137.2021.1892627