Practice-based evidence for factors associated with urinary incontinence following prostate cancer care.

Abstract

106 Background: Urinary incontinence (UI) is a common complication following treatment for localized prostate cancer. Past studies evaluating UI risk factors use surveys or chart abstraction, which may be costly and lack generalizability. Electronic health records (EHR) allow us to examine UI at a population level. We applied data mining methods to EHR data to: (1) evaluate rates of UI following prostate cancer treatment; and (2) evaluate potential risk factors for posttreatment UI. Methods: We conducted a retrospective analysis of patients undergoing prostatectomy or radiation therapy for localized prostate cancer between 2009-2016, and who received follow-up care at our medical center. Our cohort was constructed from the institutional EHR and the California Cancer Registry. The primary outcome was the presence of UI, measured in three-month intervals from the start of first-line treatment. The secondary outcome was UI 12-24 months following treatment (“late UI”). UI was assessed using natural language processing of EHR clinician notes. UI was also assessed with the EPIC-26 quality of life survey, which a subset of patients had prospectively completed. Results: Our cohort consisted of 2783 men, of whom 1907 (69%) underwent surgery and the remainder received radiation; of this cohort, 609 (22%) had data on late UI status. UI prevalence was higher among surgery than radiation patients across all posttreatment time points, and 278 of 434 (64%) surgery patients had late UI compared to 78 of 175 (45%) radiation patients (p < 0.001). Univariable analyses showed an association between pretreatment and late UI among surgery patients as measured in the EHR (OR 2.5, 95% CI 1.0-6.5, p = 0.05) and by EPIC-26 (OR 8.1, 95% CI 1.8-36.5, p = 0.01). Only surgery (compared to radiation) was a significant predictor of late UI (OR 5.8, 95% CI 1.1-32.3, p = 0.05) in multivariable regression with EHR data. Conclusions: Using EHR data, we found that treatment modality was a significant predictor of late UI among prostate cancer patients who underwent prostatectomy or radiation therapy. These results suggest the utility of EHRs in patient-centered outcomes research in prostate cancer care, and should be validated at other sites.