Is one question enough to detect female sexual dysfunctions? A diagnostic accuracy study in 6,194 women.

摘要

INTRODUCTION: Literature shows that recognition of sexual dysfunctions in women is insufficient and existing instruments to aid detection are mostly too extensive to be used in routine practice. AIM: To develop a brief and accurate screening instrument to detect female sexual dysfunctions in routine care. METHODS: The initial item pool for the index test consisted of 15 items. In a 4-year period, a total of 12,957 persons filled out the test on a specifically designed web-site. Six thousand one hundred ninety-four complete data sets could be used for statistical analysis. The validated German version of the Female Sexual Function Index (FSFI-d) served as reference standard to estimate the accuracy of the screening test. In order to test several possible ways of combining items a multi-step procedure employing univariate analyses, multiple logistic regression, and classification and regression tree analysis was applied to a learning sample and cross-validated in a test sample. MAIN OUTCOME MEASURES: Diagnostic performance (sensitivity, specificity, positive and negative predictive value, accuracy, diagnostic odds ratio as well as adjusted odds ratio) of the items and resulting models to discriminate women with sexual dysfunction from those without were calculated. RESULTS: One dichotomous question for overall satisfaction proved to show high accuracy as a stand-alone instrument and played also a central role in multivariate models. It may be recommended as a one-question screening test (76.4% sensitivity and 76.5% specificity in the test sample). A hierarchical two-question test yielded higher sensitivity (93.5%) and lower specificity (60.1%). A slightly more extensive version consists of five questions (83.1% sensitivity and 81.2% specificity). CONCLUSIONS: Despite some methodological limitations of our study all developed tests showed acceptable to good diagnostic performance, all are very short and could therefore be easily implemented into routine care. Further tests of psychometric properties in other settings are needed.