Identifying predictors of formal help-seeking for premenstrual symptoms: A machine learning analysis of symptom, functional impairment and barriers data

TL;DR

This study uses machine learning to identify factors that predict whether people seek formal help for premenstrual symptoms, finding that social impairment and perceived severity are key predictors.

Contribution

The novel contribution is applying machine learning to identify predictors of formal help-seeking for premenstrual symptoms, revealing actionable insights for healthcare improvement.

Findings

Impaired social functioning and perceived symptom severity are strong predictors of formal help-seeking.

A previous poor care experience increases the likelihood of seeking help for premenstrual symptoms.

The predictive model achieved fair performance with an AUROC of 0.75.

Abstract

Despite the potential severity and burden of premenstrual symptoms, few appear to seek formal care. Given that access to many therapeutic interventions requires formal help-seeking, it is important to understand predictors of this health behaviour. This study employed machine learning to identify symptoms, functional impairment, and barriers to accessing care that predict formal help-seeking for premenstrual symptoms. Data was collected from a UK-based sample using online survey software and explored using descriptive analysis. Group differences in ordinal and categorical data between those who have and have not sought formal help specifically for premenstrual symptoms were examined using Mann-Whitney U tests and Chi-square tests, respectively. Predictive models of help-seeking were built using the decision tree-based machine learning method, Extreme Gradient Boosting (XGBoost). A total of 592 participants with complete data who endorsed premenstrual symptoms in consecutive cycles were included for analysis. Of those, 57.26% (n = 339) had previously seen a healthcare professional specifically for premenstrual symptoms. The model predicting formal help-seeking demonstrated fair performance, with an area under the receiver operating characteristic curve (AUROC) of 0.75 (SD = 0.06), a sensitivity of 0.65 (SD = 0.10), and a specificity of 0.71 (SD = 0.11). The strongest predictors of formal help-seeking were impaired social functioning, thinking that symptoms were severe, impairment in work/studies, and a previous poor care experience for gynaecological/reproductive conditions. These insights may be leveraged to encourage help-seeking behaviour, potentially reducing unnecessary distress or impairment. Improved knowledge of premenstrual symptoms and disorders is vital to facilitate identification of severe symptoms and impaired functioning related to the menstrual cycle. Additionally, improved guidance on when to seek help is required to increase the rate of formal help-seeking, particularly for individuals with high-risk psychological symptoms such as suicidality. More work is needed to determine the specific mechanism by which previous poor care experiences drive further help-seeking for premenstrual symptoms.

Figures