Urinary incontinence as a marker of higher mortality in patients receiving home care services




  • To evaluate urinary incontinence (UI) as a predictor of nursing home admission, hospitalization or death in patients receiving home care services.

Subjects and Methods

  • A total of 699 community-dwelling participants receiving home care services in Geneva were evaluated in Autumn 2004 using the Minimal Data Set-Home Care, a validated instrument that includes grading of UI.
  • Data on death, hospitalization and nursing home admission were collected up until June 2007.
  • The impact of UI on time-dependent outcomes was analysed using survival analysis and multivariate regression Cox models to adjust for age, gender, body mass index, cardiac failure, cognitive impairment, delirium, depression, disability, alcohol and tobacco use, self-rated health, faecal incontinence and number of medications.


  • We found that UI was present in 193 participants (27.8%). After adjustment for confounding factors, UI was associated with a longer length of hospital stay: +36.7 days, (95% confidence interval [CI]: 1.2–72.3) and a higher mortality rate (hazard ratio [HR] 1.6; 95% CI: 1.1–2.6).
  • The HR for death was 1.5 (95% CI: 0.9–2.5) for participants complaining of one episode of urinary leakage per week at most, 2.0 (95% CI: 1.2–3.5) for those presenting with two or more episodes per week and 4.2 (95% CI: 2.3–7.7; P for trend: <0.001) for daily UI compared with participants without UI.
  • Institutionalization (HR 1.1; 95% CI: 0.6–2.2) and hospitalization rates (HR 1.0; 95% CI: 0.7–1.3) were not different between patients with or without UI.


  • In a cohort of patients receiving home care services, UI was a strong predictor of length of hospital stay and mortality, increasing with UI severity.

urinary incontinence


hazard ratio


Institution Genevoise de Maintien à Domicile


Minimal Data Set-Home Care


activities of daily life


instrumental activities of daily living


cognitive performance scale


Urinary incontinence (UI) is frequent in community-dwelling adults and is an important burden on the healthcare system [1]. In a large population-based study performed in five countries, the overall prevalence of UI was 5.4% for men and 13.1% for women, reaching 10.4% in men and 19.3% in women aged ≥60 years [2]. Based on the predicted ageing population, it has been estimated that 423 million individuals will be affected by UI worldwide by 2018 [3].

Older people with UI consume more healthcare resources than those without. UI has been associated with depression [4, 5], decreased work productivity [6] and a decline in quality of life [7], but other associations deserve further evaluation. A systematic review has shown inconclusive data on UI as a potential predictor of nursing home admission [8]. Similarly, studies on mortality have shown conflicting results; most of them found an association in univariate analyses, but this association vanished after adjustment for confounding factors [9].

With a shortening of lengths of hospital stay and the lack of nursing home beds, the frail population receiving home care is expanding. Furthermore, there is a widespread political push to keep people out of institutional care and at home for as long as possible with home care support. This leads to a ‘blurring of the margins’ in terms of the case-mix of older people, either in their own homes or in institutional care; therefore, there is a need to better assess the health characteristics of this specific population and to determine early factors that identify the people at risk for hospital admission, nursing home admission or death. People with a disability and concomitant diseases are especially prone to UI. The objective of the present study was to evaluate if UI was a predictor of nursing home admission, hospitalization, or death, among community-dwelling adults receiving home care services.

Subjects and Methods

Design and Population

Between July and December 2004, all individuals receiving care through the state-funded home care service (Institution Genevoise de Maintien à Domicile [IMAD]) and living in two districts of the state of Geneva were evaluated. This evaluation was carried out at the request of the Geneva State Health Services to obtain better information about home care patients and their needs, for future healthcare policy. These 699 subjects signed an informed consent form and were included in a retrospective cohort, without any exclusion criterion. Urinary symptoms and medical needs were assessed at inclusion. The first day of follow-up was arbitrarily set as 13 November 2004 for all participants. Data on death, hospital admission, and nursing home admission were collected through to 30 June 2007. Two institutional review boards approved the study.


During a routine visit to participants' homes, trained nurses collected baseline characteristics through the Minimal Data Set-Home Care (MDS-HC). The MDS-HC is a widely used validated instrument for the assessment of the health status and care needs of elderly people living in the community [10]. It has been shown to have good interobserver reliability in different languages [11]. The MDS-HC explores multiple functional domains, including grading of UI, disability and general health as well as involvement in social activities and support.

Urinary incontinence assessment

Urinary incontinence was evaluated by asking participants about urinary leakage for the last 7 days. The severity was further graded depending on frequency of leakage (grade I: once a week at most; grade II: two or more episodes per week, but not daily; grade III: one or more episode of leakage per day). Use of a bladder catheter was defined as a grade three UI. The use of pads was not considered to be UI unless urine leakages were graded.


Activities of daily life (ADL) and instrumental activities of daily living (IADL) were graded for the last 3 days. The 10 ADL were compiled to form the seven items MDS-HC ADL long-form summary scale (ranging from 0 to 28, with 0 being the best score) [12]. From the six IADL items, housework, meal preparation and telephone use were summed to produce the IADL involvement scale (ranging from 0 to 9, with 0 being the best score) [13]. A decline in ADL status (compared with status 90 days ago) estimated by the caregiver, sensory impairment (hearing and vision), and informal support or professional services required in the last 7 days were recorded. Participants estimated the time spent doing physical activities in the last 3 days (e.g. walking), which was dichotomized into >2 h or <2 h.

Cognitive and mood assessment

We considered antidepressant use or any of four triggers to be indicators of depression: persistent anger with self or others; expression of what appear to be unrealistic fears; sad, pained, worried facial expressions; and repetitive health complaints. Those four items have been independently shown to be correlated with a validated depression scale [14].

The cognitive performance scale (CPS; ranging from 0 to 6, with 0 being the best score), is a hierarchical index to rate an individual's cognitive status and has been validated against minimum mental state examination [15, 16]. The score is obtained through four items of the MDS-HC: short-term memory; cognitive skills for daily decision-making; expressive communication; and dependence for the ADL task of eating.

We used the nursing home confusion assessment method to aggregate the indicators of delirium of the MDS-HC and to estimate the presence of definite delirium (full diagnosis criteria) or subsyndromal delirium (without the full criteria) [17]. This method has been found to be consistent with other validated instruments [17].

Disease and risk factor assessment

Medical diagnoses were not available in the IMAD patient charts. Possible symptomatic heart disease was suspected when leg oedema, dyspnoea or orthopnoea was present. We considered problematic alcohol use when a participant consumed alcohol on awakening, or had physical or legal issues related to the consumption of alcohol. We recorded falls in the last 3 months. Smoking, number of medical drugs (but not therapeutic classes), and client self-estimate of poor health (when asked) were assessed.

Death, Hospital and Nursing Home Admission

Information on death was obtained from the Swiss National Death Registry, the IMAD and the Geneva University Hospitals (the only state public hospital) electronic databases. Data on nursing home admissions were collected through the IMAD database and by calling each nursing home in the Geneva state, individually. We reviewed the charts of the Geneva University Hospitals for hospital admissions and length of stay. Living participants who were not receiving home care by the 30 June 2007, and who were not admitted to the hospital or were not living in a nursing home, were considered lost to follow-up.

Statistical Analysis

We calculated that 699 participants (27% with UI) followed during a 2.5-year period with a 10% expected yearly rate of event in the control group, could detect a hazard ratio (HR) of 1.8 in the UI group with a statistical power of 80% and an α error of 5% (Freedman method).

Comparisons of characteristics between participants with and without UI were performed using the chi-squared test or Fisher's exact test when appropriate for categorical variables. A two-sided t-test was used for continuous variables with normal distribution, and Wilcoxon's rank-sum test for continuous variables not normally distributed. The trend towards UI severity was determined through logistic regression for categorical variables (the binary variable being the dependent variable and grade of UI the independent variable), and through simple regression for continuous variables.

The unadjusted impact of UI on time-dependent outcomes (first hospital admission, nursing home admission and death) was analysed using Kaplan–Meier survival analysis and an unweighted two-sided log-rank test to compare groups. Except for mortality, all analyses were censored in case of death before the closure date or occurrence of the outcome considered. Multivariate Cox models were used to adjust for potential confounding factors: age, gender, body mass index, heart failure symptoms, CPS, ADL summary scale, IADL involvement scale, problematic alcohol use, smoking, sensitive disability (hearing or visual impairment), indicators of depressive mood, delirium (full diagnosis criteria), self-estimate of poor health, faecal incontinence and number of medications. The same analyses were performed considering UI according to its severity. The choice of confounding factors was based on previous publications. Proportional hazards assumption was verified using a Schoenfeld test. The total length of hospital stay was obtained by adding together the length or each individual admission. The association between UI and length of stay was tested using a regression model adjusting for confounding factors, in which length of stay was log-transformed to correct for skewed data. All analyses were performed using stata statistical software, version 12.0 (StataCorp LP, College Station, TX, USA).


Urinary continence status was available for 694 participants (99.3%). We found that 193 of the 694 patients (27.8%) had UI. Participants with UI were more often women, had a greater number of disabilities and tended to be older (Table 1).

Table 1. Characteristics of participants with and without UI
 Group without UI, N = 499Group with UI, N = 193P
  1. aAssessed using the nursing home confusion assessment method [17].
  2. bFisher's exact test.
Mean age, years (95% CI)76.2 (75.0–77.5)78.5 (76.6–80.4)0.063
Female gender, n (%)366 (73.3)156 (80.8)0.040
Median (IQR) ADL dependence score0 (0–1)1 (0–2)<0.001
Median (IQR) IADL impairment2 (2–3)3 (2–5)<0.001
Decline in ADL status (over the last 90 days), n (%)80 (16.0)46/192 (24.0)0.016
Hearing impairment, n (%)131 (26.2)74 (38.3)<0.001
Visual impairment, n (%)126 (25.2)72 (37.3)<0.001
<2 h exercise for last 3 days), n (%)84 (16.8)58 (30.0)<0.001
Poor self-rating health, n (%)115/490 (23.5)62/189 (32.8)0.013
Mood and cognition
Indicators of depression, n (%)261 (52.3)111/192 (57.8)0.090
Full criteria deliriuma, n (%)3 (0.6)7 (3.6)<0.001b
Short-term memory impairment, n (%)61/496 (12.3)50 (26.0)<0.001
Any cognitive decline: CPS score >0, n (%)106/496 (21.4)80/192 (41.7)<0.001
Severe cognitive decline: CPS score 4–6, n (%)12/496 (2.4)15/192 (7.8)0.001
Disease and risk factors
Median (IQR) no. of medications,5 (3–7)6 (4–8)0.036
Symptoms of cardiac disease, n (%)281/498 (56.4)124 (64.2)0.061
Constipation, n (%)100/496 (20.2)52 (26.9)0.054
Faecal incontinence, n (%)11 (2.2)33 (17.1)<0.001
Nycturia or dysuria, n (%)40/498 (8.0)28 (14.5)0.010
Having fallen in last 3 months, n (%)11/498 (22.3)50 (25.9)0.313
Problematic alcohol use, n (%)16/498 (3.2)5 (2.6)0.808b
Cigarette smoking, n (%)107/498 (21.5)29 (15.0)0.055
Social functioning and support
Feelings of loneliness, n (%)140/491 (28.5)71/190 (37.4)0.025
Distress as result of decline in social activities, n (%)69/495 (13.9)40/191 (20.9)0.025
Primary caregiver expresses feelings of distress, anger or depression, n (%)22 (4.4)18/192 (9.4)0.012

Grade I UI, grade II UI and daily UI were present in 100/694 (14.4%), 63/694 (9.1%) and 30/694 (4.4%) participants, respectively. A higher UI severity was associated with a greater dependence in ADL and IADL, use of services and medication use (Table 2).

Table 2. Severity of UI and patient characteristics
 Group without UIGroup with UIP
Grade I, n = 100Grade II, n = 63)Grade III, n = 30
  1. *Assessed using the nursing home confusion assessment method [17]; IQR: 25–75%.
Median (IQR) age, years80 (70–85)80.5 (71.5–86)83 (72–88)81 (74–86)0.041
Female sex, n (%)366 (73.4)85 (85)51 (81)20 (66.7)0.359
Median (IQR) ADL long-form summary scale score0 (0–1)0 (0–2)0 (0–2)1 (0–5)<0.001
Median (IQR) IADL involvement scale2 (1–3)2 (1–4)2 (1–4)4 (2–6)<0.001
Median (IQR) no. of health/care services used2 (1–3)2 (1–3)2 (1–3)3 (2–4)<0.001
Sensitive disability214 (42.9%)57 (57%)43 (68.3%)17 (56.7%)<0.001
Median (IQR) no. of medications5 (3–7)6 (4–8)6 (3–7)6 (5–8)0.032
CPS score0 (0-0)0 (0–1)0 (0–1)1 (0–2)<0.001
Delirium*3 (0.6%)3 (3%)2 (3.2%)2 (6.7%)0.001
Depression261 (52%)57/99 (57.6%)38 (60.3%)16 (53.3%)0.295

Dependence in at least one area of ADL was present for 54.4% participants in the UI group and 38.1% in the group without UI (control group; P < 0.001). In both groups, <97% of participants needed assistance for one IADL, and >96% received professional aid (Table S1). Apart from home-making, hospital day care and social worker services, participants with UI more often used professional services.

Mortality Rate

The median follow-up period was 963 days, resulting in 448 participant-years in the UI group and 1234 participant-years in group without UI. No participant was lost to follow-up. A total of 112 deaths occurred during the study period. We found that UI was associated with a significantly higher mortality rate (Fig. 1). Mortality rates for the first and second year and by the end of follow-up were 6.7% (95% CI: 4.0–11.3), 20.2% (95% CI: 15.2–26.6) and 24.9% (95% CI: 19.4–31.6), respectively, in the UI group compared with 3.2% (95% CI: 2.0–5.2), 10.6% (95% CI: 8.2–13.7) and 12.8% (95% CI: 10.2–16.1), respectively, in the group without UI (log-rank test: P < 0.001). The HR of dying for the UI group was 2.08 (95% CI: 1.43–3.02, P < 0.001) and persisted after adjustment for confounding factors (Table 3).

Figure 1.

Survival according to UI severity.

Table 3. Adjusted and unadjusted HRs for participants with and without UI
 Unadjusted HRsAdjusted HRsa
MortalityNursing home admissionHospital admissionMortalityNursing home admissionHospital admission
  1. aAdjusted for age, gender, body mass index, cardiac failure symptoms, CPS, ADL summary scale, IADL involvement scale, problematic alcohol use, smoking, sensitive disability (hearing or visual impairment), indicators of depressive mood, delirium (full diagnosis criteria), self-estimate of poor health, faecal incontinence and number of medications.
  2. bHR given for any one point increase in UI severity.
Any UI2.1 (1.4–3.0)1.3 (0.7–2.3)1.1 (0.8–1.4)1.6 (1.1–2.6)1.1 (0.6–2.2)1.0 (0.7–1.3)
Grade I UI1.5 (0.9–2.5)1.2 (0.5–2.5)1.0 (0.7–1.5)1.4 (0.8–2.6)1.0 (0.4–2.6)1.0 (0.7–1.5)
Grade II UI2.0 (1.2–3.5)1.1 (0.4–2.9)1.1 (0.7–1.7)1.5 (0.7–3.1)1.0 (0.4–2.9)0.9 (0.5–1.5)
Grade III UI4.2 (2.3–7.7)1.8 (0.5–5.8)1.3 (0.7–2.4)3.5 (1.5–8.3)1.7 (0.4–6.7)0.9 (0.4–2.0)
P value for trends1.5 (1.3–1.8); <0.001b1.4 (1.1–1.8); <0.001b

The mortality rate increased gradually with higher severity of UI (log-rank trend test: P < 0.001; Fig. 1). The Schoenfeld test showed a constant proportional hazard (P = 0.38).

Nursing Home Admission

A total of 59 nursing home admissions occurred. The rates of nursing home admission for the first and second year and by the end of follow-up were 1.6% (95% CI: 0.5-4.9), 5.9% (95% CI: 3.2–10.7) and 9.8% (95% CI: 6.1–15.6), respectively, in the UI group compared with 2.2% (95% CI: 1.2–4.0), 5.7% (95% CI: 3.9–8.2) and 7.5% (95% CI: 5.4–10.2) in the group without UI (log-rank test: P = 0.45). The association was not significant after adjustment for confounding factors (Table 3).

Hospital Admission and Length of Stay

By the end of the study, 83/193 participants in the UI group (43.0%) compared with 212/499 (42.5%) in the group without UI were admitted at least once to a hospital. The hospital admission rate was not different between the groups with or without UI (Table 3).

The median (IQR) length of stay for participants in the UI group was 59 (23–131) days vs 44 (15–87) days for those in the group without UI. In a linear regression model with adjustment for confounding factors, we found that UI was associated with a significantly longer length of hospital stay: +36.7 days (95% CI: 1.2–72.3, P = 0.032). In this model it remained the only variable associated with hospital stay, apart from male gender. Six participants (3.1%) in the UI group spent >365 cumulative days in hospital, compared with four (0.8%, P = 0.033 on Fisher exact test) in the group without UI.


The present study identified UI as a risk factor for death, and longer length of hospital stay among a population of community-dwelling patients receiving home care services, which has not been reported previously in such a population. UI remained a strong predictor, even after adjustment for potential confounding factors measured by the MDS-HC. Although other authors have shown high mortality rates and a gradual response for increasing UI severity [9, 18, 19], this association was not found in women [20, 21], or in either gender [9, 18] after adjustment for confounding factors. An explanation for those varying results might come from the differences in the populations studied. We chose a frail population in which dependence on daily activities and the need for professional aid were excessively high (>95%). Furthermore, ∼20% of participants were <65 years old, an age group not evaluated in previous studies. The high incidence of death in a 2.5-year period gave us sufficient statistical power to demonstrate an association after adjustment compared with other studies [18]. In addition, all types of incontinence were considered (urge, stress and mixed), not only urge incontinence, in contrast to the study by Nuotio et al. [9]. Almost half of participants in the UI group had leakage twice a week to daily UI, a frequency considered severe by others. Finally, most studies evaluating UI by severity showed that high grade UI was still associated with mortality after adjustment [20]. UI occurring rarely may not be of clinical significance [22].

Patients and their GPs are reluctant to talk about UI, but the present results stress the importance of early UI detection and severity grading in any standard evaluation, as it is a strong indicator of frailty and predictor of death. We believe that the association between UI and death and the increase in the risk of mortality with higher grade of UI is of utmost importance, even if there is no clear physiopathologically direct link. UI could be associated with other debilitating conditions affecting survival. For example, the increase in the frequency of toilet visits might be associated with a higher risk of fall [23], hip fracture and consecutive mortality, especially in the elderly. Alternatively, persistent or relapsing UTIs [24, 25] could result in frequent sepsis, or could lead to cardiovascular stress such as seen among patients with HIV or rheumatic disease [26]; however, none of these hypotheses have yet been tested.

In the present study, the presence of UI was not associated with nursing home admission, which is a finding consistent with a recent meta-analysis [27]; however, we observed a very low admission rate during the 2.5-year period of follow-up, resulting in a lack of statistical power for this outcome. The decision to admit a patient to a nursing home depends on many factors [27], including personal support (formal and non-formal) and the availability of a nursing home. In fact, like many other Swiss cities, Geneva is facing a shortage of nursing home beds, generating a burden on home care services for those patients awaiting admission. The higher proportion of professional services for UI participants, and higher distress expressed by caregivers, might indicate a less stable condition, with participants being ‘institutionalized at home’. Furthermore, patients with UI had longer lengths of hospital stay. A total of 3% of these patients spent >1 cumulative year in hospital during the follow-up period. Length of hospital stay might further reflect this instable home maintenance. One study that found an impact of UI on nursing home admission, considered long hospital stay to be the equivalent of institutionalization [28].

As in all studies with a retrospective design, several limitations must be considered. The baseline evaluation was performed in 2004, but participants might have developed UI or recovered before the endpoint of interest. Nevertheless, UI seems to be a constant condition, with a potency likely to worsen with time rather than resolve spontaneously [9, 19, 29]. As such, the lack of re-evaluation of UI status should have reduced the association with mortality and not created a spurious association. Adjustment for confounding factors was dependent on information available from the MDS-HC and subject to recall and information bias. For example, we could only adjust for the number of medications, but not for specific medical drugs that were used. Similarly, diagnoses were not known and we were unable to use other validated instability scores such as the CHESS scale (changes in health, end-stage disease, signs and symptoms scale) because of a lack of information on end-stage disease [30]; however, most conditions, such as neurological, cardiac, pulmonary and rheumatic diseases, have an impact on a patient's autonomy (ADL, IADL), the number of medications, and self-rated health. These medical conditions are also linked to smoking or problematic alcohol use. Cardiac disease was suspected on the basis of data collected from the questionnaire. Nevertheless, even with an extensive effort to implement all the important variables in the model, the adjustment might not have been complete (e.g. for diabetes) and residual confounding factors cannot be ruled out. Many items of the MDS-HC evaluated by caregivers are subjective; however the MDS-HC has been shown to have excellent inter-reliability in many languages and is easy-to-use at bedside. Moreover, all measured scales used in the present study have been previously validated against widely used instruments.

In conclusion, in a cohort of ambulatory patients receiving home care services, UI was a strong predictor of hospital stay and death. The mortality rate increased in parallel with the severity of UI. After adjustment for the higher proportion in baseline ADL or IADL dependence, cognitive impairment, delirium, sensorial impairment, self-rated bad health, faecal incontinence or number of medications, the association remained strong.


G.J., E.G and O.R., who are independent of any commercial funder had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

We gratefully acknowledge Dan Lebowitz for his correction of the English manuscript.

Conflict of Interest

None declared.