Introduction: Delayed ejaculation (DE) is a disorder that can cause significant distress for sexually active men and their partners. While several medical conditions and pharmacologic agents have been associated with DE in select case reports and series, the etiology of most cases is idiopathic as the evidence underlying associated conditions is modest. We sought to use machine learning methods to examine a broad group of male health conditions and pharmaceutical treatments to identify those which are associated with DE.
Methods: Using Optum® de-identified Clinformatics Datamart (SES view), a large commercial claims data warehouse, we evaluated all men with a diagnosis of DE and matched them to a cohort of men with other urologic male sexual disorders (i.e. Peyronie’s disease (PD) and erectile dysfunction (ED)). DE men were separately 1:1-matched with two separate control groups, PD and ED, on age at onset of disease, start and end year of enrollment, and race, from 2003 to 2019. All medical diagnoses and prescription use in the six months prior to first disease diagnosis were examined. Given DE’s low prevalence, we incorporated Random Forest for classification of DE versus controls with a plethora of predictors. We used a high-performance generalized linear model using LASSO as model selection for cross validation. We used multivariate logistic models for both methods. Areas Under the Curve were reported to demonstrate classifier performance and odds ratios were shown to indicate risks of each predictors.
Results: 11,602 men with DE were matched to a cohort of men with PD and 11,719 men were matched to an ED cohort. In men with DE, the most common associated diagnosis was neurotic disease. Hypertension, metabolic syndrome, and endocrine disorders also showed increased prevalence. When stratified by race, more black men with DE had higher metabolic scores compared to other men. Clinical factors such as essential hypertension and male infertility were associated with DE ( p < 0.0001) compared to PD. With regards to ED, the presence of male infertility, anxiety , and the use of alpha blockers or anti-depressants, were associated with DE (p < 0.0001). There was a 90% accuracy in predicting a diagnosis of DE compared to PD and 85% accuracy compared to ED when using the identified diagnoses and medications from the model.
Conclusions: DE is associated with multiple medical conditions which may help identify men at risk of developing DE and offer novel treatment options with better understanding of etiology.
.jpg)
.jpg)