RRISK is a prospective cohort study of women who were at baseline members of Kaiser Permanente Medical Care Program of Northern California (KPNC), a pre-paid group practice with over 3 million members which provides comprehensive health care to about 25% of general geographic population in the area served. Previous studies have found members of KPNC to somewhat under represent the very poor and very wealthy and to be slightly more educated compared to the general population in the same geographic area, but to be very similar with respect to other demographic characteristics.1
The study was approved by the institutional review (human subjects) boards of both the University of California, San Francisco and Kaiser Foundation Research Institute. A random sample of 10,161 women between 40 and 69 years of age as of January 1, 1999 who had been members since age 18 was generated from membership files. Starting in September 1999, women were screened and recruited with the goal of obtaining weighted sample with equal numbers in each 5 year age strata and a distribution of 40% white, 20% Hispanic, 20% Asian and 20% African-American participants.2
Women were excluded if they did not speak English or Spanish, reported having had less than half of their births within KPNC, were no longer members of KPNC, had moved out of the area, or were demented or otherwise too impaired to participate. Among the 8835 women whose eligibility could be determined, 6018 (68%) were ineligible, primarily for having < 50% of their births within Kaiser (n=2451) or because we had already filled their age-race strata (n=2632), and 2817 (32%) were eligible. Assuming that 32% of women of unknown eligibility were also ineligible, the total number of eligible women would be 3240, of which we enrolled 2109 (65.1%). A second survey conducted approximately 5 years later, between June, 2003 to January 2008, was completed by 1413 (67%) of the original cohort. Of the 696 women not completing the second survey, 446 refused, 163 were ineligible due to poor health or having moved outside the area, and 87 could not be contacted.
During the first (baseline) survey, variables ascertained by pre-interview and in-person questionnaires included demographic characteristics, past history of incontinence, diabetes, chronic obstructive pulmonary disease (COPD) and hysterectomy, pregnancy and parturition history, menopausal status, smoking, and general health, and a detailed description of current incontinence symptoms. Weight and height were measured by the interviewer and used to calculated body mass index (BMI) in kg/m2. Ambulation speed was measured by recording the time needed to walk10 feet, turn around, and return. The second (follow-up) survey ascertained an interim medical, surgical, and pregnancy history, and repeated measures of menopausal status, smoking, general health, BMI, and ambulation speed.
For both surveys, current urinary incontinence was defined as at least monthly incontinence and was further characterized as ‘less then weekly,’ ‘weekly’ and ‘daily.’ Type of incontinence was defined for those women with at least weekly incontinence, according to their response to two questions, one asking if incontinence occurred “with an activity like coughing, lifting, sneezing or exercise” (stress incontinence) and the other asking if incontinence occurred “with a physical sense of urgency” (urgency incontinence). Women were classified as having stress incontinence if they reported the majority of episodes being stress incontinence, and as having urgency incontinence if they reported the majority of the episodes being urgency incontinence. Those reporting both stress and urgency incontinence with neither predominating were classified as mixed incontinence. Women with only other incontinence (n=34) were excluded from the analyses by incontinence type.
Change in incontinence status was defined as incident incontinence (change from continence to any degree of incontinence); progression of incontinence (change from ‘< weekly incontinence’ to ‘weekly or daily incontinence,’ or from ‘weekly incontinence’ to ‘daily incontinence’); regression of incontinence (change from ‘weekly incontinence’ to ‘< weekly incontinence’ or from ‘daily incontinence’ to ‘weekly or < weekly incontinence’); and resolution of incontinence (change from any degree of incontinence to continence). A logistic model was used to test the association between candidate risk variables and change in incontinence status. Variables were evaluated for inclusion in the multivariate model using backwards elimination, and were retained in the final model if p remained <0.2. Continuous variables were also examined as categorical variables to evaluate for non-linear associations with change in incontinence status. Parity was categorized as 0, 1 or 2+ as there was virtually no additional association between number of deliveries beyond the second and change in incontinence status. Each model included only women for whom the specific change in continence status was possible. For example, the model for progression of incontinence was limited to women with incontinence < daily incontinence at baseline and the dependent variable was therefore women who had progression of their incontinence vs. all other women in this group. Associations between independent variables and change in incontinence were expressed as odds ratio (OR), and 95% confidence intervals (95% CI). All p-values were two-sided. Model fit was assessed using the Hosmer-Lemeshow ‘goodness of fit’ test and the c statistic.3
All analyses were carried out in SAS Version 8.02 (SAS Institute, Cary, NC).