External validation of a prediction model for penile prosthesis implantation for erectile dysfunction management.

Introduction: Penile prosthesis implantation (PPI) is the definitive surgical treatment for erectile dysfunction (ED), yet it is often delayed for a variety of reasons. From commercial and Medicare claims data, we previously developed a tool for determining a patient's likelihood of eventually receiving PPI. We validated this instrument's utility by comparing cohorts receiving surgical (PPI) versus non-surgical ED management at a single institution. Material and methods: The prediction model was based on a logistic regression incorporating claims data on demographics, comorbidities and ED therapy. A risk score is calculated from the model as the product of relative risks for the individual variables. The current validation was a retrospective analysis of ED patients seen at this institution from January to December 2012. Inclusion criteria included ED diagnosis and either first-time PPI or non-surgical treatment (controls). Risk scores for patients receiving PPI were compared to those of nonsurgical controls. Results: We established a cohort of 60 PPI patients (mean age 54.4 +/- 9.5) and compared them with 120 non-PPI patients (mean age 53.4 +/- 11.2 years). The median score of the PPI cohort was 5.7 (IQR 2.8-9.9) versus the non-PPI cohort's 1.8 (IQR 0.9-5.5) (p < 0.0001). The area under the receiver operator characteristic curve for predicting eventual PPI was 0.72 (95% CI, 0.64-0.79) (p < 0.0001). Conclusion: The prediction model risk-stratified men who ultimately underwent PPI compared to non-surgically managed controls. This external validation study suggests that the prediction model may be used on an individual patient basis to support a recommendation of PPI for managing ED.