Prediction models for postpartum urinary and fecal incontinence in primiparous women.

Abstract

OBJECTIVES: This study aimed to develop and internally validate a nomogram that facilitates decision making between patient and physician by predicting a woman's individual probability of developing urinary (UI) or fecal incontinence (FI) after her first delivery. METHODS: This study used Childbirth and Pelvic Symptoms Study data, which estimated the prevalence of postpartum UI and FI in primiparous women after vaginal or cesarean delivery. Two models were developed using antepartum variables, and 2 models were developed using antepartum plus labor and delivery variables. Urinary incontinence was defined by a response of leaking urine "sometimes" or "often" using the Medical, Epidemiological, and Social Aspects of Aging Questionnaire. Fecal incontinence was defined as any involuntary leakage of mucus, liquid, or solid stool using the Fecal Incontinence Severity Index. Logistic regression models allowing nonlinear effects were used and displayed as nomograms. Overall performance was assessed using the Brier score (zero equals perfect model) and concordance index (c-statistic). RESULTS: A total of 921 women enrolled in the Childbirth and Pelvic Symptoms Study, and 759 (82%) were interviewed by telephone 6 months postpartum. Two antepartum models were generated, which discriminated between women who will and will not develop UI (Brier score = 0.19, c-statistic = 0.69) and FI (Brier score = 0.10, c-statistic = 0.67) at 6 months and 2 models were generated (Brier score = 0.18, c-statistic= 0.68 and Brier score = 0.09, c-statistic = 0.68) for predicting UI and FI, respectively, for use after labor and delivery. CONCLUSIONS: These models yielded 4 nomograms that are accurate for generating individualized prognostic estimates of postpartum UI and FI and may facilitate decision making in the prevention of incontinence.