NHANES is a nationally representative annual household survey that assesses health status through interview questionnaires, physical examinations and laboratory tests. We utilized de-identified publicly available 2015-2018 NHANES weighted data for women 20 years or older to estimate UI prevalence. We limited our analysis to women who had completed the mobile examination, including the computer assisted personal interview portion of the exam (administered by trained interviewers) that included the standardized UI questions. UI type was classified according to reported symptoms in the last 12 months as stress incontinence (leakage with activity), urge incontinence (leakage with urgency), mixed incontinence (leakage with both activity and urgency), or unspecified (leakage without activity or urgency). The Sandvik Incontinence Severity Index (1) quantified condition severity.
Socio-demographic and medical history components examined included age in years; race/ethnicity (self-reported into broadly defined categories and separated in this analysis as Non-Hispanic Black, Non-Hispanic White, Latinx, Other); BMI (kg/m2); highest education level (less than high school, high school, some college, college graduate or more); smoking history (never, former, current), prior hysterectomy (yes or no); birth history (no births, any vaginal birth, cesarean only); and currently pregnant (yes or no). Depression was assessed by the Patient Health Questionnaire (PHQ-9) and scored as none/mild, moderate, or severe. Anxiety was assessed by the General Anxiety Disorder-7 (GAD-7) and scored as none/mild, moderate, or severe. Physical activity level was assessed by the Global Physical Activity Questionnaire and scored as none, moderate, or active. Functional limitations were assessed in domains of activities of daily living (4 items), instrumental activities of daily living (3 items), leisure and social activities (3 items), lower extremity mobility (2 items) and general physical activity (8 items). Respondents who answered they had any difficulty on any item in a domain were coded as having dependence in that domain. Respondents were then classified into independent in all domains, dependent in 1-2 domains, and dependent in 3-5 domains. Diabetes mellitus was classified as yes or no, with the “no” category including those who only had gestational diabetes.
Demographics and medical co-morbidities were categorized and evaluated for statistically significant associations with UI using the chi-square test. Multivariable logistic regression modeling determined adjusted associations with UI.