# Using artificial intelligence to predict sexual health outcomes in endometriosis: a decision tree model algorithm

## TL;DR

This study uses AI to identify factors that predict sexual dysfunction in women with endometriosis, helping to guide personalized treatment strategies.

## Contribution

The novel contribution is the development of a decision tree model that accurately identifies key predictors of sexual dysfunction in endometriosis patients.

## Key findings

- The decision tree model showed strong performance in predicting sexual dysfunction with an AUC of 0.96 in training data.
- Chronic pelvic pain, dyspareunia, infertility, and digestive symptoms were identified as major predictors of sexual dysfunction.
- Certain subgroups, such as those with severe menstrual cramps and no treatment, had a 100% probability of dysfunction.

## Abstract

Endometriosis profoundly impairs sexual function through complex interactions between pain, hormonal disturbances, psychological distress, and sociodemographic factors.

To develop and validate a decision tree–based model identifying key predictors of sexual dysfunction in women with endometriosis.

We conducted a cross-sectional online survey among 1586 women with endometriosis recruited through social media in France between November 2023 and January 2024. Participants completed sociodemographic and clinical questionnaires and the Female Sexual Function Index (FSFI), with sexual dysfunction defined as FSFI < 26.55. The predictors included pain characteristics, menstrual symptoms, digestive symptoms, infertility, BMI, lifestyle factors, and treatment history. A classification and regression tree model was trained on 70% of the sample and validated on the remaining 30%. Model performance was assessed using area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. Subgroups at high or low risk of sexual dysfunction were identified through terminal nodes (rules) of the decision tree.

The primary outcome was classification into sexual dysfunction versus normal sexual function.

Of the 1586 respondents, 1358 (86%) met the criteria for sexual dysfunction. The decision tree demonstrated strong discrimination in the training dataset (AUC = 0.96; sensitivity = 98%; specificity = 82%) and acceptable performance in the validation dataset (AUC = 0.79; sensitivity = 72%; specificity = 77%). Chronic pelvic pain, dyspareunia, worsening pain over time, heavy menstrual bleeding, infertility, BMI, digestive symptoms, education level, and treatment history were identified as major determinants. Several terminal nodes showed a 100% probability of sexual dysfunction, most notably women with severe menstrual cramps, heavy menstrual bleeding, no treatment, normal weight or obesity, infertility, or pronounced digestive symptoms. Conversely, women with no dyspareunia, no urinary or bowel pain, high education, and overweight exhibited a 0% probability of dysfunction.

AI-driven decision trees can support early identification of high-risk profiles and guide individualized management strategies to improve sexual health in women with endometriosis.

Strengths include a large sample, comprehensive symptom profiling, and transparent model interpretability. Limitations include reliance on self-reported diagnosis, potential selection bias inherent to online recruitment, lack of geographic and clinical verification data, and the cross-sectional nature of the analysis, preventing causal inference.

A decision tree–based model accurately identified key predictors of sexual dysfunction in endometriosis, supporting its potential for personalized risk stratification and clinical decision support.