Predictors of urinary incontinence in community-dwelling frail older adults with diabetes mellitus in a cross-sectional study

We studied all On Lok enrollees diagnosed with diabetes mellitus between October 2004 and December 2010 (N = 447 participants with n = 2602 UI measurements). On Lok, the original model for Programs for All-inclusive Care for the Elderly (PACE), requires enrollees to be nursing home-eligible, indicating that the participant requires care with full-time supervision of a licensed nurse. On Lok helps nursing home-eligible enrollees remain in the community by providing and coordinating healthcare services, including primary and specialist physician services, adult day health care, home care, hospital care, post-acute rehabilitation care and custodial nursing home care. On Lok provides enrollees with transportation between home and PACE centers where meals, medication management, help with bathing or showering and recreational activities are provided. Further, On Lok centers have physical and occupational therapists, social workers, nurses and physicians on-site. Each enrollee receives a comprehensive health assessment (medical evaluation with assessment of function and geriatric syndromes) upon enrollment and every 6 months thereafter by physicians, nurses, therapists and social workers.Enrollees were eligible for our study if they were enrolled in On Lok during the study period and had a diagnosis of diabetes mellitus on a glucose lowering medication or a hemoglobin A1c (HbA1c) level greater than 6.5% (Figure 1). Diabetes diagnosis was determined according to the International Classification of Diseases, Ninth Revision (ICD-9) code, 250.xx.

We initially identified 460 (N) On Lok enrollees with 2783 (n) UI measurements. UI measurements were excluded from the analysis if enrollees had end-stage renal disease on dialysis (N = 6, n = 59), were receiving end-of-life care (N = 6, n = 112) or had a urinary catheter (N = 1, n = 10). These exclusion criteria led to our final analytic cohort of 447 enrollees with 2602 UI measurements. All data were extracted from electronic medical records. The Committee on Human Research at the University of California San Francisco and the San Francisco VA Research and Development Committee approved this study. Further, they found that this study satisfied federal criteria for waiving informed consent since 1) it poses minimal risks to participants, 2) it would not be practicable to obtain informed consent and 3) participants rights are not adversely affected by waiving informed consent.

To capture the dynamic nature of diabetes and UI in this population, our unit of analysis was measures of urinary incontinence. Thus, a patient who was in our study for 9 months would contribute two UI measurements: the initial admission assessment and the subsequent 6 month follow-up assessment. This allowed us to explore whether risk factors that change over time such as insulin use or HbA1c levels affect incontinence.

The primary outcome was urinary incontinence, evaluated during semi-annual comprehensive health assessments. A nurse or therapist determined the presence and frequency of UI by conducting interviews with enrollees and corroborating with family caregivers and home care aides. Nurses and therapists were trained to code UI occurring less than once a week as “seldom incontinent” and UI occurring more than once per week as “often incontinent”.

We compared measures of “never incontinent” to the combined measure of “seldom incontinent” and “often incontinent”. To explore whether some factors were associated with more severe incontinence, we also compared “never incontinent” to “often incontinent”.

We examined a wide range of potential factors that may be associated with urinary incontinence. Specifically, we examined demographic factors including age, gender and race/ethnicity. We examined geriatric factors including dependence in ambulation, dependence in transferring and cognitive impairment. Ambulation and transferring was assessed by a nurse who made an in-person determination of whether the individual was independent, required supervision, required assistance, was dependent on others or non-ambulatory. All levels except for independent were categorized as “dependent.” Participants with a Mental Status Questionnaire (MSQ) score greater than 4 were considered to have cognitive impairment. Participants without MSQ scores but had an ICD-9 diagnosis of dementia were also considered to have cognitive impairment.

We examined diabetes-related factors including the use of glucose lowering medications, HbA1c levels, and diabetes-related complications such as renal or ophthalmologic complications, peripheral vascular disease, and neurological disease (through ICD-9 codes). To determine the HbA1c level on the day of urinary incontinence assessment, we interpolated HbA1c values, assuming that the HbA1c changes in a linear fashion between 2 measured values. For example, if the first measured HbA1c value is 7.0% and the next measured value 100 days later is 8.0%, the interpolated HbA1c value is 7.1% on Day 10, 7.2% on Day 20, etc. We also examined whether depressive symptoms and diuretic use was associated with urinary incontinence. Presence of depressive symptoms was defined by a short Geriatric Depression Scale score greater than 9.

Several other factors were considered, but were not included in the final analysis due to a low number of participants and measurements (less than 10%) with these risk factors. These included: obesity, benign prostatic hypertrophy, prostate cancer (determined using ICD-9 codes) and use of a urinary antispasmodic agent.

Subjects with and without incontinence were characterized using descriptive bivariate statistics. We used ANOVA to compare the means of continuous variables (age and HbA1c level) and Chi-square tests to compare categorical variables. We performed multivariate analyses to identify independent risk factors for urinary incontinence using mixed effects logistic regression to account for clustering of incontinence measurements by participant. We adjusted for age, gender, Asian race, dependence in transferring and ambulating, cognitive impairment, use of thiazide or loop diuretics, depression, diabetic medication use, and diabetic complications (renal, ophthalmologic, peripheral vascular, and neurological). All analyses were performed using Stata MP (version 10.1, StataCorp, College Station, TX) and SAS (version 9.2, SAS System of Windows, SAS Institute Inc., Cary, NC).

Table 1 shows the characteristics of the participants at initial assessment by level of urinary incontinence. The overall prevalence of UI was 44% at baseline. The majority of participants were 75 years or older with a high proportion of female and Asian enrollees. Participants had an average of 5 assessments (range 1–19). At the first assessment 56% were “never incontinent”, 28% were “seldom incontinent”, and 16% were “often incontinent”. Participants were frail older adults with complex medical issues. Forty-two percent were dependent with ambulation and 36% had cognitive impairment. Severe urinary incontinence was more common at baseline in older adults with dependence in ambulation and transferring and cognitive impairment. Diuretic use was common in our study population at 25%. Depressive symptoms were present in 25% of participants. Depressive symptoms were present in 57% of those with measurements of “often incontinent” but in fewer than 30% of those with “never incontinent” or “seldom incontinent” measures. Participants had appropriately controlled diabetes according the American Geriatrics Society guidelines [24], with 70% of participants with HbA1c of less than 8% at the first visit within our study. However, forty-eight percent of participants had renal complications related to diabetes while other diabetes-related complications were less common.

Many non-diabetes factors were strongly associated with urinary incontinence (Table 2). Demographic factors (increasing age and African American or other and White race) were associated with urinary incontinence. Although female gender was associated with incontinence in the bivariate analysis, gender was not an independent predictor after adjustment. Geriatric factors were also strongly associated with urinary incontinence. Partial or full dependence with ambulation or transferring were both associated with urinary incontinence with the odds ratio for transferring remaining over 2 after adjustment. Two factors commonly associated with urinary incontinence, diuretic use and presence of depressive symptoms, were not associated with urinary incontinence in this study.

Several diabetes-related factors in the model were significantly associated with urinary incontinence. The strongest association was the use of insulin followed by use of oral glucose-lowering agents. HbA1c level was not found to be associated with increased odds of UI.

The sensitivity analysis performed by comparing “never incontinent” with “often incontinent” showed similar results. However, peripheral vascular disease was found to be independently associated with being “often incontinent” (OR 2.33; 95% CI: 1.35, 4.02) but not when “seldom” was combined with “often incontinent” (OR 1.31; 95% CI: 0.81, 2.10). A second sensitivity analysis performed using ordinal logistic regression with 3 outcome categories (never, seldom and often incontinent) showed that dependence in ambulation (OR 1.61; 95% CI: 1.14-2.27), transferring (OR 3.32; 95% CI: 2.28-4.84) and presence of cognitive impairment (OR 1.73; 95% CI: 1.29-2.32) continued to be statistically significantly correlated with urinary incontinence. The differences were that the age group of 75–80 years old (OR 1.34; 95% CI: 0.86-2.1), White race (OR 1.54; 95% CI: 0.93-2.58), the use of oral glucose-lowering agents (OR 1.21, 95% CI: 0.8-1.82) and insulin (OR 1.48; 95% CI: 0.81-2.68) were no longer statistically significantly associated with urinary incontinence. Having peripheral vascular disease was found to be statistically significant (OR 1.89, 95% CI: 1.1-3.26).