Clinical validation of the Bladder Health Survey for urinary incontinence in a population sample of women

Introduction and hypothesis Our aim was to clinically validate the Bladder Health Survey (BHS) for detecting urinary incontinence (UI) in population-based surveys. Methods A random sample of women >= 40 years was recruited from primary care practices. We assessed the BHS content validity with an expert advisory board. Test-retest reliability of UI questions was measured. BHS UI definitions included noncases, active (more than three symptoms in the prior 6 months), inactive (past but no current symptoms), and incident (new onset over the past 2 years) cases. To assess criterion validity, we compared BHS diagnosis to an expert clinical diagnosis using structured history, pelvic exam, voiding diary, and urodynamics (if needed). Construct validity was assessed comparing the BHS UI score and case status to Sandvik's score. Results Among 322 patients, the BHS identified 17 % as noncases, 70% as active, 10% as inactive, and 3% as incident cases. Using the clinical diagnosis as the gold standard, the percent of true-positive UI cases was 98 % (active), 84 % (inactive), and 80 % (incident). A total of 75 % of BHS noncases were true negatives. The receiver operating characteristic c-statistic was 0.86. Sensitivity and specificity of the BHS were 91% and 84 %, respectively. The Sandvik score for active cases (median = 4) was significantly greater than it was for inactive (median = 1), incident (median = 1), and noncases (median = 0) (p < 0.001). The BHS UI score was significantly correlated with the Sandvik severity score (r = 0.68, p < 0.01). Conclusion The BHS is highly reliable, with robust content and construct validity for detecting UI for use in population samples.