Urinary Incontinence and Its Association with Death, Nursing Home Admission, and Functional Decline

Jayna M. Holroyd-Leduc, MD, w Kala M. Mehta, DSc,w and Kenneth E. Covinsky, MD, MPH w

OBJECTIVES: To determine whether urinary incontinence (UI) is an independent predictor of death, nursing home admission, decline in activities of daily living (ADLs), or decline in instrumental activities of daily living (IADLs). DESIGN: A population-based prospective cohort study from 1993 to 1995. SETTING: Community-dwelling within the United States. PARTICIPANTS: Six thousand ve hundred six of the 7,447 subjects aged 70 and older in the Asset and Health Dynamics Among the Oldest Old study who had complete information on continence status and did not require a proxy interview at baseline. MEASUREMENTS: The predictor was UI, and the outcomes were death, nursing home admission, ADL decline, and IADL decline. Potential confounders considered were comorbid conditions, baseline function, sensory impairment, cognition, depressive symptoms, body mass index, smoking and alcohol, demographics, and socioeconomic status. RESULTS: The prevalence of UI was 14.8% (18.5% in women; 8.5% in men). At 2-year follow-up, subjects incontinent at baseline were more likely to have died (10.9% vs 8.7%; unadjusted odds ratio (OR) 5 1.29, 95% condence interval (CI) 5 1.021.64), be admitted to a nursing home (4.4% vs 2.6%, OR 5 1.77; 95% CI 5 1.18 2.63), and to have declined in ADL function (13.6% vs 8.1%; OR 5 1.78, 95% CI 5 1.362.33) and IADL function (21.2% vs 13.8%; OR 1.69, 95% CI 1.392.05). However, after adjusting for confounders, UI was not an
From the San Francisco VA Medical Center, San Francisco, California; and w Division of Geriatrics, University of California, San Francisco, California. This work was supported by a grant from the National Institute on Aging (R01AG19827). Dr. Holroyd-Leduc is funded as a VA National Quality Scholar fellow. Dr. Mehta is supported in part by a training grant from the National Institute on Aging (T32-AG00212-08). Dr. Covinsky is supported by an independent scientist award from the Agency for Healthcare Research and Quality (K02 HS00006-01) and is a Paul Beeson Faculty Scholar in Aging Research. An abstract of this study was presented at the 26th Annual Meeting of the Society of General Internal Medicine in Vancouver, Canada, May 2003, and at the American Geriatrics Society 2003 Annual Scientic Meeting in Baltimore, May 2003. Address correspondence to Dr. J. M. Holroyd-Leduc, San Francisco VA Medical Center (181G), 4150 Clement St, Bldg 1, San Francisco, CA 94121. E-mail: Jayna.Holroyd-Leduc@med.va.gov

independent predictor of death (adjusted OR (AOR) 5 0.90, 95% CI 5 0.671.21), nursing home admission (AOR 5 1.33, 95% CI 5 0.862.04), or ADL decline (AOR 5 1.24, 95% CI 5 0.921.68). Incontinence remained a predictor of IADL decline (AOR 5 1.31; 95% CI 5 1.051.63), although adjustment markedly reduced the strength of this association. CONCLUSION: Higher levels of baseline illness severity and functional impairment appear to mediate the relationship between UI and adverse outcomes. The results suggest that, although UI appears to be a marker of frailty in community-dwelling elderly, it is not a strong independent risk factor for death, nursing home admission, or functional decline. J Am Geriatr Soc 52:712718, 2004. Key words: urinary incontinence; mortality; nursing homes; activities of daily living

rinary incontinence (UI) is one of the cardinal geriatric syndromes. Its prevalence in the elderly population is high and increases with advancing age. UI is an important medical condition that has been found to be associated with poor self-rated health,1 impairment in quality of life,2 social isolation,3 and depressive symptoms.4 The presence of UI may be of prognostic signicance, in that it may be a marker of frailty and a predictor of adverse outcomes in older people, but there is surprisingly little known about the relationship between UI and key outcomes, such as nursing home admission and decline in activities of daily living (ADLs) and instrumental activities of daily living (IADLs). Several studies have found that persons with UI have a higher risk of death, but greater illness severity in persons with UI explains this association.57 Nevertheless, studies of the relationship between UI and other adverse outcomes are minimal, and these studies were limited because they were unable to adequately adjust for potential confounders or because the patient sample had limited generalizability. For example, although it is widely assumed that, because of the high prevalence of UI in nursing home residents,8,9 UI is a cause of nursing home admission, prospective studies of this relationship are limited. One of the few studies of this relationship found that UI was a predictor of nursing home admission in two cohorts of patients within the Kaiser

JAGS 52:712718, 2004 r 2004 by the American Geriatrics Society

0002-8614/04/$15.00

JAGS

MAY 2004VOL. 52, NO. 5

URINARY INCONTINENCE AND ADVERSE OUTCOMES

713

Foundation Health Plan of Northern California.7 Although the study adjusted for age and comorbidity, it did not take into account other key confounders such as baseline functional impairment and sensory impairment. The relationship between UI and functional impairment is even less well understood. A cross-sectional study found that nursing home residents with UI had higher levels of ADL dependency.10 Unadjusted cross-sectional data from Japan also demonstrated that community-dwelling individuals with UI were more likely to have ADL dependence.11 However, these cross-sectional comparisons make it difcult to establish a causal relationship. A subsequent longitudinal study of older Mexican Americans found that baseline UI was not an independent predictor of frailty but that onset of frailty and incident UI appeared to occur together.12 Their denition of frailty was a summary measure that combined self-reported ADL and IADL dependence and physical performance measures. Despite these ndings, the relationship between UI and decline in ADLs and IADLs remains controversial. The relationship between UI and frailty also remains poorly understood. Frailty has been dened as a dynamic model that incorporates a balance between multiple factors, which include chronic illness, dependency or risk of dependency in ADL, and availability of resources and assets.13,14 Frail individuals are at risk of death, nursing home admission, and acute hospitalization. A frailty scale that included bladder incontinence in its classication scheme showed a dose-response relationship between the level of frailty and risk of institutionalization and death.15 The objective of this study was to comprehensively examine the relationship between UI and key adverse outcomes in older persons, including death, nursing home admission, ADL decline, and IADL decline.

Sample minus proxy 6656 sample minus proxy (n = 6,656) (150 missing) Sample with complete continence information (n = 6,506) (593 died) Sample minus deaths (n = 5,913) (41 missing) Nursing home analytic sample (n = 5,872) (392 missing) Activities of daily living decline analytic sample (n = 5,521) (952 missing) Instrumental activities of daily living decline analytic sample (n = 5,509)

Figure 1. Flow diagram of sample populations used in the analysis of the four outcomes. NH 5 nursing home; ADL 5 activities of daily living; IADL 5 instrumental activities of daily living.

In the current study, 791 subjects for whom a proxy respondent was used were excluded. A further 150 subjects were excluded because of missing continence information (Figure 1). For the outcome variable of death there were no missing data, and the 593 who died were excluded from the analysis of the three other outcome variables. These analytic samples consisted of 5,872 in the nursing home sample, 5,521 in the ADL decline sample, and 5,509 in the IADL decline sample. All confounding variables had 1% or less missing information, except for cognitive scores (4.5%).

Measures METHODS Study Population The study included subjects interviewed in 1993 and 1995 as part of the Asset and Health Dynamics Among the Oldest Old (AHEAD) study. This is a prospective cohort study representative of the community-dwelling U.S. population aged 70 and older (N 5 7,447).16 AHEAD was formulated as a supplement to the Health and Retirement communitybased study. The sampling frame was formed from the U.S. birth cohorts of 1923 and earlier. A full description of the sampling and weighting procedures used in the AHEAD study has been described previously.17 Blacks, Hispanics, and Florida state residents were oversampled. Weights were developed for AHEAD to adjust for different probabilities of selection, including oversampling. Respondents aged 70 to 79 were generally interviewed by telephone, and respondents aged 80 and older were generally interviewed in person, but the same instrument was administered to all self-respondents.18 The response rate did not differ signicantly between these two modes of interview, and the in-person interviews were only 4 minutes longer on average. The overall response rate was 80%. The human subjects committee of the University of California, San Francisco, and the San Francisco Veterans Affair research and development committee approved this project. Urinary Incontinence UI was measured at baseline based on the response to a sequence of two questions. The rst question was, The next question might not be easy to talk about, but it is very important for research on health and aging. During the last 12 months, have you lost any amount of urine beyond your control? If subjects responded yes to this question they were then asked, On about how many days in the last month have you lost any urine? These measures were based on questions used in the Medical, Epidemiologic, and Social Aspects of Aging study.19 For purposes of this study, a person was considered incontinent of urine if they responded yes to the rst question and then responded that they had had at least one episode of urine loss in the previous month. One hundred fty-two subjects who responded yes to the rst question were classied as continent (2%), because they responded that they had no episodes of urine loss in the previous month. UI frequency was further classied as occasional (115 days of loss) or frequent (415 days of loss). Death Mortality was assessed over 2 years using the AHEAD survey follow-up procedures.20 Deaths were subsequently conrmed using the U.S. National Death Index.

714

HOLROYD-LEDUC ET AL.

MAY 2004VOL. 52, NO. 5

JAGS

Nursing Home Admission During the 1995 interview, subjects or their proxy respondents were asked whether the subject was living in a nursing home or other healthcare facility. Place of residence was further assessed over the 2 years using the AHEAD survey follow-up procedures.20 Functional Decline At baseline and at follow-up, study subjects were asked whether they required help from another person with six ADLs and ve IADLs. These six ADLs were walking across a room, dressing, bathing, eating, getting in and out of bed, and using the toilet. An ADL dependence score was calculated for each interview, assigning one point for each of the activities in which a subject was independent. Functional decline was determined to have occurred if a subject had a greater number of ADL dependencies at follow-up than at baseline. The ve IADLs were help with grocery shopping, making telephone calls, taking medications, preparing hot meals, and difculty managing money. An IADL dependence score was similarly calculated, and functional decline was determined to have occurred if a subject had a greater number of IADL dependencies at follow-up. Confounders Several baseline variables were considered that, based on prior literature,4,8,2128 could confound the association between UI and the outcomes of death, nursing home admission, and functional decline. These measurements were sex, age, socioeconomic status (total net worth and education), race, self-reported smoking and alcohol consumption, body mass index (BMI) based on self-reported height and weight, sensory impairment (vision and hearing), self-reported comorbid disease (stroke, heart disease, hypertension, psychiatric disease, lung disease, arthritis, previous hip fracture, joint replacement, cancer, and diabetes mellitus), depressive symptoms based on the eight-item Center for Epidemiologic Study Depression scale,29 cognitive function measured using a 35-point scale developed for the AHEAD study,30 and baseline dependency in the ve ADLs and the six IADLs. Statistical Analysis The statistical analysis applied sampling weights to AHEAD data to account for the studys complex design. All statistical analysis was performed using STATA 7.0 software (STATA Corp., College Station, TX). The main objective of the study was to determine whether UI was an independent predictor of death, nursing home admission, ADL decline, and IADL decline. Chi-square statistics were performed to compare the baseline characteristics of the entire study population and the sample populations analyzed for each of the outcome variables. Chi-square and Mann-Whitney rank sum test statistics were performed to compare the baseline characteristics of subjects with incontinence with those without incontinence. Relationship Between UI and Outcomes Bivariate logistic regression analyses were conducted to assess the unadjusted associations between UI and the four outcomes of interest: death, nursing home admission, ADL

decline, and IADL decline. To determine whether UI was an independent predictor of each outcome, four separate logistic regression models were used to examine the relationship between UI and each of the four outcome variables after adjusting for confounders. To adjust for the cognitive score variables, which had more than 1% missing information, an indicator variable was incorporated into the logistic regression analyses. Based on the HosmerLemeshow goodness-of-t test, each of the models t adequately. The goodness-of-t tests were performed without application of the study design weights. For the outcome of death, the analysis was repeated on the unweighted sample using Cox proportional hazards modeling. Based on scaled Schoenfeld residuals, proportional hazards assumptions were met.

Relationship Between UI Frequency and Outcomes UI was divided by the frequency of episodes (occasional or frequent). For each of the four outcomes, bivariate and multivariate logistic regression analyses were then performed using indicator variables for the different UI frequency subgroups. Effect Modication To test for the possibility of interaction (effect modication) between UI and each baseline characteristic felt to be a predictor of the outcomes, chi-square statistics and MantelHaenszel test of homogeneity were performed. The presence of interactions was determined by examining the P-value for interaction and comparing the odds ratios across strata. The logistic regression analyses were repeated by stratifying for baseline characteristics that were determined to be effect modiers.

RESULTS Characteristics of Subjects The mean age standard deviation of the subjects at baseline was 77 (range 69103); 63% were female, and 86% were white (Table 1). There were no signicant differences in baseline characteristics between the baseline population (n 5 6,506) and the sample populations used in the analysis (nursing home admission, n 5 5,872; ADL decline, n 5 5,521; IADL decline, n 5 5509; P4.9 for all comparisons). The baseline prevalence of UI was 14.8% (18.5% in women; 8.5% in men). The median number of baseline comorbid medical conditions was two. Ten percent were receiving help with one or more ADLs, and 23% were receiving help in one or more IADLs at baseline. At baseline, subjects with UI had more comorbidities than continent subjects (Po.001). They also had higher rates of visual impairment (35.0% vs 22.4%; Po.001) and hearing impairment (39.7% vs 30.5%; Po.001). Subjects with UI were also more functionally impaired at baseline. Signicantly more incontinent subjects had baseline dependency in at least one ADL (20.6% vs 8.4%; Po.001) and in at least one IADL (33.1% v 20.8%; Po.001). Even if toileting was excluded, those with UI still had higher rates of baseline ADL dependency (20.4% vs 8.4%; Po.001).

JAGS

MAY 2004VOL. 52, NO. 5

URINARY INCONTINENCE AND ADVERSE OUTCOMES

715

Table 1. Characteristics of Study Population at Baseline (N 5 6,506)

Characteristic Incontinent Female White Age o75 7579 ! 80 Education 11 years Net worth o$44,000 $44,000150,000 4$150,000 Smoker Current Former Never ! 3 alcoholic drinks/d Fair-poor vision Hearing Fair-poor Hearing aid Comorbidities Hypertension Diabetes mellitus Cancer Lung disease Heart disease Stroke Psychiatric Arthritis Hip fracture Joint replaced Fair-poor self-rated health Receives help with activities of daily living Receives help with instrumental activities of daily living % 14.8 63 86 40 29 31 40 31 34 35 10 42 48 2 24 19 13 50 12 14 11 31 9 11 25 5 7 33 10 23

was not an independent predictor of death, nursing home admission, or ADL decline. It did remain an independent predictor of IADL decline, although adjustment markedly reduced the strength of this association. When examined individually, age, comorbidities, baseline functional status, depression scores, and sensory impairment were the main confounders of the relationship between UI and the outcomes (Table 3). Using Cox proportional hazards modeling, UI was conrmed not to be an independent predictor of death (hazard ratio 5 0.87, 95% condence interval (CI) 5 0.691.09). The adjusted relationship between incontinence and ADL decline was similar when ADL decline was redened as declining in two or more ADLs (odds ratio (OR) 5 1.22, 95% CI 5 0.871.73). The adjusted relationship between UI and IADL decline also was similar when IADL decline was redened as declining in two or more IADL (OR 5 1.38, 95% CI 5 1.071.78). Even if toileting was excluded from the ADL variable, UI was not an independent predictor of ADL decline (OR 5 1.22, 95% CI 5 0.90 1.65).

Relationship Between UI Frequency and Outcomes When UI was stratied by frequency, those with frequent incontinence had higher unadjusted risks for each of the four outcomes (Table 4), but in both the frequently incontinent and the occasionally incontinent strata, adjustment for confounders explained the relationship between UI and adverse outcomes. Effect Modication There were no effect modiers of the relationship between UI and death, ADL decline, or IADL decline. Nevertheless, although there was no independent relationship between UI and nursing home admission overall, there was an association in several subgroups in which interaction terms suggested differential effects. The baseline characteristics determined to be effect modiers were BMI, vision, ADL status, and smoking status (Po.1 and difference in OR41.0). Stratied analysis revealed that UI was an independent predictor of nursing home admission in those subjects in the highest tertile for BMI (AOR 5 2.53, 95% CI 5 1.354.73) and in those with visual impairment (AOR 5 1.99, 95% CI 5 1.103.60). There was also a trend toward increased nursing home admission in those with UI who were

Relationship Between UI and Outcomes UI at baseline was associated with higher unadjusted rates of death, nursing home admission, ADL decline, and IADL decline (Table 2), but after adjusting for confounders, UI

Table 2. Relationship Between Baseline Urinary Incontinence and Outcomes

Continent Outcome Death Nursing home ADL decline IADL decline

Incontinent % Unadjusted OR (95% CI) 10.9 4.4 13.6 21.2 1.29 (1.021.64) 1.77 (1.182.63) 1.78 (1.362.33) 1.69 (1.392.05) Adjusted OR (95% CI) 0.90 (0.671.21) 1.33 (0.862.04) 1.24 (0.921.68) 1.31 (1.051.63)

8.7 2.6 8.1 13.8

Adjusted for sex, age, race, education, net worth, smoking, alcohol, body mass index, sensory impairment, comorbidity, cognitive scores, depression scores, baseline activities of daily living (ADLs), and baseline instrumental activities of daily living (IADLs). After adding an indicator to the cognitive scores variable to adjust for missing data, covariate information was missing from 154 for death, 34 for nursing home, 80 for ADL decline, and 83 for IADL decline. OR 5 odds ratio; CI 5 condence interval.

716

HOLROYD-LEDUC ET AL.

MAY 2004VOL. 52, NO. 5

JAGS

Table 3. Relationship Between Baseline Urinary Incontinence and Outcomes After Adjusting for Different Confounders
Death Model Unadjusted Adjusted for sex Adjusted for age Adjusted for socioeconomic status and race Adjusted for smoking/alcohol Adjusted for BMI Adjusted for sensory impairment Adjusted for comorbidity Adjusted for cognitive scores Adjusted for depression scores Adjusted for ADL Adjusted for IADL Complete adjustmentz

Nursing Home Admission

ADL Decline

IADL Decline

Odds Ratio (95% Condence Interval) 1.29 (1.021.64) 1.39 (1.081.78) 1.14 (0.881.47) 1.25 (0.971.60) 1.32 (1.041.67) 1.30 (1.011.66) 1.15 (0.911.46) 1.05 (0.821.35) 1.29 (1.021.64) 1.11 (0.871.41) 1.01 (0.781.31) 1.10 (0.851.44) 0.90 (0.671.21) 1.77 (1.182.63) 1.63 (1.082.44) 1.47 (0.982.20) 1.67 (1.112.51)

1.78 (1.362.33) 1.65 (1.262.18) 1.59 (1.222.08) 1.76 (1.332.32) 1.76 (1.342.32) 1.80 (1.362.38) 1.61 (1.232.11) 1.53 (1.172.00) 1.82 (1.382.40) 1.49 (1.131.96) 1.49 (1.121.98) 1.49 (1.131.96) 1.24 (0.921.68)

1.69 (1.392.05) 1.60 (1.311.95) 1.54 (1.251.89) 1.67 (1.402.00) 1.67 (1.372.03) 1.72 (1.412.10) 1.51 (1.231.85) 1.51 (1.231.85) 1.75 (1.472.10) 1.43 (1.181.74) 1.45 (1.181.79) 1.57 (1.281.93) 1.31 (1.051.63)

1.74 (1.162.61)w 1.59 (1.062.38) 1.79 (1.202.67) 1.53 (1.012.30)

1.46 (0.982.20) 1.33 (0.862.04)

Smoking, body mass index (BMI), and activity of daily living (ADL) status were not included in the nursing home (NH) model as confounders because they were determined to be effect modiers. Alcohol was not included in the NH model because no one admitted to a NH drank three or more drinks per day at baseline. Includes hearing only because vision was an effect modier. z After adding an indicator to the cognitive score variable to adjust for missing data, covariate information was missing from 154 for death, 34 for NH, 80 for ADL decline, and 83 for instrumental activities of daily living (IADL) decline.

dependent in at least one ADL at baseline (AOR 5 1.84, 95% CI 5 0.804.26). In the other strata for each of these baseline characteristics, UI was neither predictive of nor protective against nursing home admission. Previous smokers with UI were signicantly more likely to be admitted to a nursing home than previous smokers without UI (AOR 5 2.36, 95% CI 5 1.174.78). UI was not an independent predictor of nursing home placement in nonsmokers (AOR 5 1.03, 95% CI 5 0.571.87) or current smokers (AOR 5 0.68, 95% CI 5 0.123.83).

DISCUSSION This is one of the rst U.S. population-based cohort studies to examine the relationship between UI and key adverse outcomes in community-dwelling elderly individuals. The results demonstrate that elders with UI are at substantially higher risk of death, nursing home admission, ADL decline, and IADL decline, but this relationship between UI and

these adverse outcomes appears to be mostly due to higher baseline illness severity and functional impairment in elders with UI. This would support the notion that UI is a marker of frailty but not the hypothesis that it is an independent predictor of these adverse outcomes. Adjustment for baseline illness and functional impairment explained the relationship between UI and death, nursing home admission, and ADL decline and explained most of the relationship between UI and IADL decline. The nding that these confounders also explained the dose-response relationship seen between UI frequency and the outcomes strengthened these results. Consistent with previous studies,57 these results demonstrate that UI does not appear to be an independent risk factor for death. In addition, baseline UI was not an independent risk factor for ADL decline. These results would suggest that there is not a direct pathophysiological pathway between UI and progression to functional dependency and death.

Table 4. Relationship Between Baseline Urinary Incontinence Frequency and Outcomes

Occasional Incontinence Outcome Death Nursing home ADL decline IADL decline

Frequent Incontinence % 13.7 5.5 15.1 24.1 OR (95%CI) 1.68 (1.172.40) 2.22 (1.263.91) 2.02 (1.462.79) 1.99 (1.462.71) Adjusted OR (95%CI) 1.02 (0.651.61) 1.33 (0.722.44) 1.12 (0.771.63) 1.32 (0.951.85)

Continent % 8.7 2.6 8.1 13.8

% 8.5 3.6 12.3 18.9

OR (95%CI) 0.98 (0.711.35) 1.40 (0.912.16) 1.59 (1.102.30) 1.46 (1.091.95)

Adjusted OR (95%CI) 0.78 (0.551.10) 1.32 (0.852.06) 1.38 (0.912.07) 1.30 (0.931.82)

Adjusted for sex, age, race, education, net worth, smoking, alcohol, body mass index, sensory impairment, comorbidity, cognitive scores, depression scores, baseline activities of daily living (ADLs), and baseline instrumental activities of daily living (IADLs). OR 5 odds ratio; CI 5 condence interval.

JAGS

MAY 2004VOL. 52, NO. 5

URINARY INCONTINENCE AND ADVERSE OUTCOMES

717

Nevertheless, the psychological and social issues related to UI may affect the adverse outcomes of nursing home admission and IADL decline. This study did not clearly support the ndings of another study that UI is an independent predictor of nursing home placement.7 This may be related to differences in the denition of UI and the resulting difference in prevalence rates. The other study diagnosed UI based on medical chart extraction, and the prevalence of UI was lower (6.9% in women; 5.3% in men). The lower prevalence rates found in that study may be related to the fact that persons with UI often do not seek medical attention31,32 and therefore do not have a diagnosis of UI in their medical chart. The difference in the ndings may also reect the more comprehensive adjustment for other baseline risk factors in this study, but in several subgroup analyses performed because interaction terms suggested differential effects, UI was an independent predictor of nursing home placement. This may be related to the likelihood that these subgroups had higher levels of caregiver dependence and burden. In subgroups with high levels of caregiver burden, UI may act as a nal stressor that leads to nursing home admission. These subgroups included elders with baseline ADL dependence, visual impairment, high BMI, and previous smoking history. Dependence in ADLs, visual impairment, and obesity all increase caregiver burden, making it less likely that caregivers will be able to cope with the additional burden of UI. It was surprising to nd that previous smokers were more likely to be admitted to a nursing home. Although it is possible that this association with UI is due to chance, the nding was highly and statistically signicant, and the effect size was large. This interaction could possibly be due to higher levels of unmeasured illness severity and functional impairment in previous smokers, which may be one reason they quit smoking. Another possibility is that smoking restrictions in nursing homes acted as an admission barrier to current smokers. Psychosocial factors may also have inuenced the relationship between UI and IADL decline. The presence of UI can contribute to social isolation and restriction of household chores,3 resulting in IADL impairment in areas such as shopping and meal preparation, but further research needs to be conducted to investigate the relationship between UI, psychosocial factors, and IADL decline. When classied into different frequencies, UI was not found to be an independent predictor of IADL decline, but this loss of statistical signicance is probably due to the loss of statistical power that resulted from dividing the sample into smaller subgroups. In this study, the prevalence of UI was based on the Medical, Epidemiologic, and Social Aspects of Aging criteria. The prevalence of 18.5% in women and 8.5% in men is approximately within the range reported by other studies. A review of previous literature found that the prevalence of having ever experienced UI ranged between 17% and 55% in older women and 11% and 34% in older men.33 The prevalence of daily UI, in previous studies, ranged between 3% and 7% in older women and 2% and 11% in older men. The variability in prevalence rates appears to be due to differences in the denition of UI and in the age and sex of the population studied.

This study has several advantages over previous studies. First, a large population-based cohort representative of the community-dwelling elderly in the United States was used. Second, there were multiple measures of the key potential confounders, making it possible to determine whether UI is an independent predictor of adverse outcomes. Third, the analysis considered frequency of incontinence, making it possible to determine whether there is a dose-response relationship. Fourth, the analyses considered the possibility that relationships between UI differed in clinically relevant subgroups. One limitation of this study was that UI was measured using self-report. This could have lead to possible reporting bias, in that subjects with UI may have underreported their symptoms, but survey questionnaire is a common method used in the collection of UI information.33 The fact that two different methods of interview were used, based on the subjects age, may also have been a potential source of bias. Those aged 80 and older were interviewed in person because it was felt that age-related increases in hearing impairment, cognitive impairment, arthritis, and frailty would adversely affect the ability to perform telephone interviews with these older subjects.18 Financial limitations prevented every interview from being conducted in person. Previous studies have found that a persons continence status can change over time, in that UI can both develop and remit,34,35 but this study used continence status at baseline and did not adjust for changes over the study period. It was important to establish that UI existed before the onset of the outcomes, because the objective of the study was to determine whether UI was an independent predictor of adverse outcomes. Although multiple comparisons were made in this study, this is not a signicant statistical concern given the overall negative ndings of the study. This study measured frequency of incontinence but did not measure other indicators of severity, such as the amount of urine lost. Another limitation is that the type of UI was not determined. The type of UI has been found to be an important predictor of the adverse outcome of falls.26 Finally, this study did not measure many of the potential psychosocial intermediaries between UI and the measured adverse outcomes. Associations between UI and impaired quality of life,2 social isolation,3 and depressive symptoms4 have been identied in other studies. The effect of UI on such psychosocial variables and the resulting indirect effect on morbidity and independence were not explored in this study. Regardless of whether UI is a direct cause of mortality and functional decline, it remains important to diagnose and treat this condition effectively because of its negative effect on quality of life. In conclusion, although elderly people with UI are at substantially higher risk of death, nursing home admission, ADL decline, and IADL decline, UI does not appear to be a strong independent risk factor for these adverse outcomes. Higher levels of baseline illness severity and functional impairment appear to explain the relationship between UI and these poor outcomes. The study results suggest that, although there may not be a direct pathophysiological link between UI and these adverse outcomes, UI does appear to be a marker of frailty in communitydwelling elderly.

718

HOLROYD-LEDUC ET AL.

MAY 2004VOL. 52, NO. 5

JAGS

ACKNOWLEDGMENTS We would like to acknowledge and thank Karla Lindquist for her assistance with the AHEAD database. REFERENCES
1. Johnson TM II, Kincade JE, Bernard SL et al. The association of urinary incontinence with poor self-rated health. J Am Geriatr Soc 1998;46:693699. 2. Temml C, Haidinger G, Schmidbauer J et al. Urinary incontinence in both sexes. Prevalence rates and impact on quality of life and sexual life. Neurourol Urodyn 2000;19:259271. 3. Wyman JF, Harkins SW, Fantl JA. Psychosocial impact of urinary incontinence in the community-dwelling population. J Am Geriatr Soc 1990;38:282288. 4. Dugan E, Cohen SJ, Bland DR et al. The association of depressive symptoms and urinary incontinence among older adults. J Am Geriatr Soc 2000;48:413 416. 5. Johnson TM II, Bernard SL, Kincade JE et al. Urinary incontinence and risk of death among community-living elderly people: Results from the National Survey on Self-Care and Aging. J Aging Health 2000;12:2546. 6. Herzog AR, Diokno AC, Brown MB et al. Urinary incontinence as a risk factor for mortality. J Am Geriatr Soc 1994;42:264268. 7. Thom DH, Haan MN, Van Den Eeden SK. Medically recognized urinary incontinence and risks of hospitalization, nursing home admission and mortality. Age Ageing 1997;26:367374. 8. Ouslander JG, Kane RL, Abrass IB. Urinary incontinence in elderly nursing home patients. JAMA 1982;248:11941198. 9. Yarnell JW, St Leger AS. The prevalence, severity and factors associated with urinary incontinence in a random sample of the elderly. Age Ageing 1979;8:8185. 10. Ouslander JG, Uman GC, Urman HN et al. Incontinence among nursing home patients. Clinical and functional correlates. J Am Geriatr Soc 1987;35:324 330. 11. Koyano W, Shibata H, Haga H et al. Prevalence and outcome of low ADL and incontinence among the elderly: Five years follow-up in a Japanese urban community. Arch Gerontol Geriatr 1986;5:197206. 12. Miles TP, Palmer RF, Espino DV et al. New-onset incontinence and markers of frailty: Data from the Hispanic Established Populations for Epidemiologic Studies of the Elderly. J Gerontol A Biol Sci Med Sci 2001;56A:M19M24. 13. Rockwood K, Fox RA, Stolee P et al. Frailty in elderly people: An evolving concept. Can Med Assoc J 1994;150:489495. 14. Rockwood K, Stolee P, McDowell I. Factors associated with institutionalization of older people in Canada: Testing a multifactorial denition of frailty. J Am Geriatr Soc 1996;44:578582. 15. Rockwood K, Stadnyk K, MacKnight C et al. A brief clinical instrument to classify frailty in elderly people. Lancet 1999;353:205206. 16. Myers GC, Juster FT, Suzman RM. Asset and Health Dynamics Among the Oldest Old (AHEAD): Initial results from the longitudinal study. Introduction. J Gerontol B Psychol Sci Soc Sci 1997;52(Spec No):vviii. 17. Heeringa SG. Technical Description of the Asset and Health Dynamics (AHEAD) Survey Sample Design. Ann Arbor: MI, University of Michigan, 1995.

18. Soldo BJ, Hurd MD, Rodgers WL et al. Asset and Health Dynamics Among the Oldest Old: An overview of the AHEAD Study. J Gerontol B Psychol Sci Soc Sci 1997;52(Spec No):120. 19. Herzog AR, Fultz NH. Prevalence and incidence of urinary incontinence in community-dwelling populations. J Am Geriatr Soc 1990;38:273281. 20. The Health and Retirement Study: A Longitudinal Study of Health, Retirement, and Aging Sponsored by the National Institute on Aging [online]. Available at www.umich.edu/ $ hrswww/docs/impute.html. Accessed January 28, 2004. 21. Herzog AR, Fultz NH, Brock BM et al. Urinary incontinence and psychological distress among older adults. Psychol Aging 1988;3:115121. 22. Thom DH, van den Eeden SK, Brown JS. Evaluation of parturition and other reproductive variables as risk factors for urinary incontinence in later life. Obstet Gynecol 1997;90:983989. 23. Brown JS, Seeley DG, Fong J et al. Urinary incontinence in older women. Who is at risk? Study of Osteoporotic Fractures Research Group. Obstet Gynecol 1996;87:715721. 24. Brown JS, Grady D, Ouslander JG et al. Prevalence of urinary incontinence and associated risk factors in postmenopausal women. Heart and Estrogen/ Progestin Replacement Study (HERS) Research Group. Obstet Gynecol 1999;94:6670. 25. Fultz NH, Herzog AR, Raghunathan TE et al. Prevalence and severity of urinary incontinence in older African American and Caucasian women. J Gerontol A Biol Sci Med Sci 1999;54A:M299M303. 26. Brown JS, Vittinghoff E, Wyman JF et al. Urinary incontinence. Does it increase risk for falls and fractures? Study of Osteoporotic Fractures Research Group. J Am Geriatr Soc 2000;48:721725. 27. Diokno AC, Brock BM, Herzog AR et al. Medical correlates of urinary incontinence in the elderly. Urology 1990;36:129138. 28. Sampselle CM, Harlow SD, Skurnick J et al. Urinary incontinence predictors and life impact in ethnically diverse perimenopausal women. Obstet Gynecol 2002;100:12301238. 29. Stefck DE, Wallace RB, Herzog AR et al. Documentation of Affective Functioning Measures in the Health and Retirement Study. Ann Arbor, MI: University of Michigan, 2000. 30. Ofstedal MB, McAuley GF, Herzog AR et al. Documentation of Cognitive Functioning Measures in the Health and Retirement Study. Ann Arbor, MI: University of Michigan, 2002. 31. Roberts RO, Jacobsen SJ, Rhodes T et al. Urinary incontinence in a community-based cohort. Prevalence and healthcare-seeking. J Am Geriatr Soc 1998;46:467472. 32. Stoddart H, Donovan J, Whitley E et al. Urinary incontinence in older people in the community: A neglected problem? Br J Gen Pract 2001;51:548 552. 33. Thom D. Variation in estimates of urinary incontinence prevalence in the community: Effects of differences in denition, population characteristics, and study type. J Am Geriatr Soc 1998;46:473480. 34. Herzog AR, Diokno AC, Brown MB et al. Two-year incidence, remission, and change patterns of urinary incontinence in noninstitutionalized older adults. J Gerontol 1990;45:M67M74. 35. Nygaard IE, Lemke JH. Urinary incontinence in rural older women. Prevalence, incidence and remission. J Am Geriatr Soc 1996;44:10491054.