Urinary incontinence (UI) is a common condition affecting nearly one fifth of women in the United States.1 The symptoms can be extremely bothersome2 and may be accompanied by numerous adverse psychosocial sequelae, including social isolation3 and depression.4 Overall, women with UI experience substantially reduced quality of life.5,6 In comparison, less is known about the economic impacts of UI. Estimates of the volume and aggregate cost of urogynecologic procedures do not account for indirect costs, such as those related to lost economic productivity.6,7 Women with UI adopt numerous strategies to manage symptoms at work, but their symptoms nonetheless compromise their self-confidence, concentration, and work performance.2,8–11 Faced with these difficulties, women may decide to engage in limited types of work or work fewer hours, or even exit the workforce entirely.
Understanding the indirect costs of UI as reflected in labor force outcomes has long been highlighted as an important area of investigation.7,8 Although the cross-sectional studies cited suggest that UI may have important negative effects on economic outcomes, they are far from conclusive. To address this gap in the literature, we conducted a secondary analysis of data from the Health and Retirement Study, an ongoing population-based cohort of middle-aged and older adults. Given their high risk not only for transitions out of the workforce but also for UI1 and depression,12 this age group represents an important population for in-depth study of the relationships between these variables. Specifically, we sought to estimate the association between UI and depression, work disability, and workforce exit over the course of long-term follow-up.
MATERIALS AND METHODS
The Health and Retirement Study uses a multistage area probability sample; at its inception in 1992, it had a target population of all adults aged 51–61 years residing in households in the contiguous United States.13 Data are collected through biennial telephone and in-person interviews. The response rate to the initial survey was 81.6%, and reinterview response rates have exceeded 93.1% at every subsequent interview cycle. The omnibus questionnaires administered to study participants reflect a broad range of analytic and policy interests related to labor force, health, and family transitions. Because of its expansive scope and sampling strategy, the Health and Retirement Study is frequently used by analysts to investigate questions of clinical relevance to the gynecologic care of middle aged and older women3,4,10,14,15 and to middle-aged and older adults in general.16–18
Because questions about the primary exposure of interest (UI) were not added to the surveys until 1996, we restricted our analysis to data beginning in 1996, by which time study participants were 54–65 years old. UI was assessed with the question, “This might not be easy to talk about, but during the last 12 months, have you lost any amount of urine beyond your control?” Participants who responded affirmatively were then asked to estimate the frequency of episodes with the question, “On about how many days in the last month have you lost any urine?” Responses were grouped into three ordered categories (none or continent, 0–15 days, and more than 15 days), which allowed us to explore potential dose-dependence in the associations between UI and the outcomes of interest.
The primary outcomes of interest were onset of probable depression, work disability, and workforce exit. Depression symptom severity was measured using a modified version of the Center for Epidemiologic Studies–Depression scale,19 which was shortened to eight items while retaining similar reliability,20,21 factor structure,21 and construct validity.21–23 Although the diagnosis of major depressive disorder is a clinical diagnosis that accounts for other pertinent information such as symptom duration, functional impairment, and exclusion of other psychiatric diagnoses, participants' survey responses enabled us to identify probable depression with reasonable accuracy. Specifically, a score of 3 or more on the modified Center for Epidemiologic Studies–Depression scale was used to classify study participants as having symptoms indicating probable depression. This threshold has been found to have 71% sensitivity and 79% specificity when compared with the criterion standard of major depression as determined by the short form of the World Health Organization Composite International Diagnostic Interview.21,24
The outcome of work disability was assessed with the question, “Do you have any impairment or health problem that limits the kind or amount of paid work you can do?” Responses are scored on a dichotomous scale (“yes” or “no”). Its construct validity is supported by previous studies correlating work disability with pain23 and labor force participation.25 Furthermore, little bias was observed when the responses of Health and Retirement Study participants were compared with the criterion standard of U.S. Social Security Administration disability insurance benefit decisions.26 The outcome of workforce exit was based on the “labor force status” variable in the RAND Health and Retirement Study Data file,27 which summarizes the labor force status for each respondent at each interview into mutually exclusive categories of “working full-time,” “working part-time,” “unemployed,” “partly retired,” “retired,” “disabled,” or “not in the labor force.”28 Workforce exit was defined as being retired, partly retired, disabled, or not in the labor force.
For each of the three outcomes, we fit two Cox proportional hazards regression models corresponding to the two different specifications of UI described here. Analysis time was extended through 2010–2011. Estimates were adjusted for the following sociodemographic and health status variables measured at baseline: age, race; marital status, educational attainment, household income, body mass index, parity, smoking status, alcohol use, psychiatric medication use, number of physician-diagnosed chronic conditions out of eight (hypertension, diabetes, cancer, chronic lung disease, cardiovascular disease, cerebrovascular disease, arthritis, or psychiatric disorder), and difficulties with activities of daily living. In these regression models, the subsample of participants initially considered “at risk” varied by outcome. For the analyses examining onset of probable depression, only participants who had no depressive symptoms or who had subthreshold symptoms (modified depression score less than 3) at baseline were considered “at risk” for development of the outcome over the course of the follow-up period. For the analyses examining onset of work disability, we limited the analytic sample to participants with no work disability at baseline. Finally, for the analyses examining workforce exit, we excluded participants who were not in the labor force at baseline (ie, they reported being retired, partly retired, or not in the labor force).
For each multivariable regression model, the link test proposed by Pregibon29,30 was used to assess adequacy in specification of the linear predictor. Namely, after estimation of the regression coefficients, the linear predictor and its square were included in a second version of the regression model. A nonstatistically significant estimated regression coefficient on the square of the prediction was taken to be evidence that the proportional hazards assumption was not violated. No violations were noted. All analyses were conducted with the use of Stata statistical software 13.0 and used the survey weights and clustering variables to adjust estimates and standard errors for the complex survey sampling design. Because of survey weighting, some prevalence estimates may not correspond exactly to the ratio of the subpopulation to the sample size and are explicitly labeled here.
Study participants provided written informed consent. Health and Retirement Study procedures were approved by the relevant committees at the University of Michigan and the U.S. National Institute of Aging. The data have been made publicly available for download at http://hrsonline.isr.umich.edu/ as unrestricted datasets, which are purged of secondary identifying information such that they pose no significant threat to respondent anonymity. The specific analysis presented in this article was reviewed by the Partners Human Research Committee and was determined not to meet the definition of human subject research because it was based on anonymous, public-use data with no identifiable information for study participants.
Summary statistics for the sample are described in Table 1. The analytic sample consisted of 4,511 women. The mean age was 59.4 years. The sample consisted largely of Caucasian women, and only a minority of these did not complete high school. Most women reported having three or more children.
Among all study participants, 727 (survey-weighted prevalence, 16.6%; 95% confidence interval [CI] 15.4–18.0) reported any UI at baseline. Of these, 212 participants (survey-weighted prevalence, 29.2%; 95% CI 25.4–33.3) reported losing urine on more than 15 days during the past month. A total of 1,052 participants were categorized as having probable depression (survey-weighted prevalence, 21.6%; 95% CI 19.8–23.6) and 1,276 participants had a work disability (survey-weighted prevalence, 27.1%; 95% CI 25.4–28.9), whereas 1,921 participants were working for pay, either full-time or part-time (survey-weighted prevalence, 43.0%; 95% CI 41.4–44.5); these participants were excluded from the survival analyses depending on the model specification.
Analytically, women were followed-up from 15 November 1995 to 15 May 2011. Among the 3,300 women with absent or subthreshold symptoms of depression at baseline, 1,628 developed probable depression (49.3%). Presence of UI was associated with an increased risk of probable depression (adjusted hazard ratio, 1.43; 95% CI 1.27–1.62) (Table 2). Increasing frequency of incontinence (in terms of days with urine loss) was associated with greater risk.
Among the 3,075 women with no work disability at baseline, 1,398 developed a new work disability (45.5%). Presence of UI was associated with an increased risk of work disability (adjusted hazard ratio, 1.21; 95% CI 1.01–1.45), with increasing frequency of incontinence similarly associated with greater risk. Among the 1,848 women working full-time or part-time at baseline, 1,621 left the workforce (87.7%). Regardless of how the exposure was defined, UI did not have a statistically significant association with workforce exit.
In this nationally representative cohort of women between the ages of 54 and 65 years in the United States followed-up over the course of 14 years, we found that UI was associated with increased risks for probable depression and work disability but not workforce exit. The estimated associations were statistically significant, large in magnitude, increasing in severity of the exposure, consistent with related strands of research, and robust to the inclusion of confounders previously found to be related to both the exposures and outcomes. Our findings have important implications for research and clinical practice in this area.
The potential association between depression and either UI or the severity of incontinence has been well-studied. Most analyses in this literature have been cross-sectional in nature,4,31–33 but two longitudinal studies have shown null findings.15,34 Thom et al34 measured UI using chart diagnoses, but it is well-known that fewer than one half of women with urine leakage actually seek care for their symptoms.35 The analysis by Melville et al15 was based on the same underlying dataset as ours, but we believe there are four potential explanations for the conflicting findings. First, in estimating the association between UI measured in 1996 and probable depression measured during 1998–2004, Melville et al15 did not account for attrition from the study during the 8-year period. Second, our use of repeated-measures, time-updated data regarding UI minimized the potential for misclassification of exposure status (that would have occurred had a study participant's exposure status changed after the baseline interview). Third, in our analysis, the participants were followed-up for a longer duration of time. The Cox proportional hazards regression models likely provided greater statistical power than the logistic regression model of Melville et al,15 because study participants contributed data until they experienced the outcome of interest or exited the study. Fourth, Melville et al15 used a different threshold on the depression scale (ie, with 6 or more indicating probable depression). Any of these analytic modifications, either alone or in combination with the others, could have explained the discrepant results.
A second important finding is that UI predicted the development of a work disability. Given that previous work has shown depression to be one of the most important causes of work absenteeism,36,37 the analysis presented here extends previously published findings that only indirectly suggest the potential adverse labor force effects of UI.2,8–11 The mechanisms explaining the association between UI and work disability are likely varied and may be attributable to symptom-related bother,8,9 avoidance and limiting behaviors,2 or restricted physical activities in general.10,11 Importantly, there was no statistically significant association between UI and workforce exit. It may be that the threshold for exiting the workforce because of incontinence-related bother may be higher than the threshold for engaging in different types of work or cutting back on the amount of work.
If these estimated associations are causal, the findings presented in this analysis suggest that early diagnosis and management of UI may have important psychosocial and economic benefits for women. Given that depression is a treatable condition, obstetricians and urogynecologists should be attentive to its signs and symptoms among their patients presenting with UI. With regard to the economic implications of the study, some observers may question how many more years of economic productivity the women in this study had left to contribute before exiting the workforce. However, if effective management of UI prevents the loss of only 1 year and potentially up to 11 years of economic productivity (ie, before Medicare eligibility), this still could have important economic consequences for individual women. At the level of population health, the macroeconomic implications for the United States would depend on the aggregate number of women with UI for whom effective intervention delayed or prevented their reliance on public insurance programs like Medicare and Social Security Disability Insurance. Certainly, the analyses presented here cannot establish causality. It may be that UI is simply a marker for frailty heralding subsequent functional decline,38 but this would not undermine the economic benefits of early diagnosis and management.
Interpretation of our findings is subject to several limitations. First, the analytic sample consisted of a relatively homogeneous group of women with an age range of 54–65 years at baseline. On average, they may have fewer economically productive years compared with younger women, so it is likely that the potential economic effects identified in our study are minimized. UI is likely to exert even more powerful labor force effects at younger ages and among women of lower socioeconomic status, but this possibility would need to be explored with a different dataset. Second, all of the variables were self-reported. Notably, labor force participation was not verified with employer surveys or access to employment records. However, it is also important to note that the outcome measures were generic rather than condition-specific. To the extent that outcome measures based on condition-specific questions (eg, “Does fear of odor or smell restrict your activities?”39) would have been more closely correlated with UI,31,40 our estimates are likely biased toward the null. Third, this analysis did not distinguish between different types of incontinence. To the extent that urge UI has greater adverse effects on quality of life compared with stress UI,41–43 the estimates presented here may be biased toward the null. Finally, although the outcome of probable depression (as determined by responses to a screening instrument) is accepted in the field and is frequently used in large-scale epidemiologic studies,4,15 it does not equate to a clinical diagnosis of major depressive disorder or dysthymia consistent with the Diagnostic and Statistical Manual of Mental Disorders. To explore the sensitivity of our findings to alternative definitions of probable depression, we used a threshold of 4 or more on the Center for Epidemiologic Studies–Depression scale, which is the threshold estimated by Steffick44 to correspond to the conventionally adopted threshold of 16 or more on the full 20-item scale.19,45 This procedure actually yielded point estimates of increased magnitude and statistical significance, so we chose to report the more conservative estimates in this article.
These caveats notwithstanding, our analysis of population-based, longitudinal data shows that UI among women between the ages of 54 and 65 years is associated with increased risks for onset of probable depression and work disability. Our findings are unlikely to be explained entirely by unobserved confounding and, if anything, are stronger and of greater magnitude than the estimates presented here. Therefore, improved diagnosis and management of UI may yield significant economic and psychosocial benefits.
1. Nygaard I, Barber MD, Burgio KL, Kenton K, Meikle S, Schaffer J, et al.. Prevalence of symptomatic pelvic floor disorders in US women. JAMA 2008;300:1311–6.
2. Fultz NH, Burgio K, Diokno AC, Kinchen KS, Obenchain R, Bump RC. Burden of stress urinary incontinence for community-dwelling women. Am J Obstet Gynecol 2003;189:1275–82.
3. Yip SO, Dick MA, McPencow AM, Martin DK, Ciarleglio MM, Erekson EA. The association between urinary and fecal incontinence and social isolation in older women. Am J Obstet Gynecol 2013;208:146 e1–7.
4. Nygaard I, Turvey C, Burns TL, Crischilles E, Wallace R. Urinary incontinence and depression in middle-aged United States women. Obstet Gynecol 2003;101:149–56.
5. Temml C, Haidinger G, Schmidbauer J, Schatzl G, Madersbacher S. Urinary incontinence in both sexes: prevalence rates and impact on quality of life and sexual life. Neurourol Urodyn 2000;19:259–71.
6. Subak LL, Brown JS, Kraus SR, Brubaker L, Lin F, Richter HE, et al.. The “costs” of urinary incontinence for women. Obstet Gynecol 2006;107:908–16.
7. Wyman JF. The “costs” of urinary incontinence. Eur Urol 1997;32(Suppl 2):13–9.
8. Fultz N, Girts T, Kinchen K, Nygaard I, Pohl G, Sternfeld B. Prevalence, management and impact of urinary incontinence in the workplace. Occup Med (Lond) 2005;55:552–7.
9. Margalith I, Gillon G, Gordon D. Urinary incontinence in women under 65: quality of life, stress related to incontinence and patterns of seeking health care. Qual Life Res 2004;13:1381–90.
10. Fultz NH, Fisher GG, Jenkins KR. Does urinary incontinence affect middle-aged and older women's time use and activity patterns? Obstet Gynecol 2004;104:1327–34.
11. Nygaard I, Girts T, Fultz NH, Kinchen K, Pohl G, Sternfeld B. Is urinary incontinence a barrier to exercise in women? Obstet Gynecol 2005;106:307–14.
12. Blazer DG. Depression in late life: review and commentary. J Gerontol A Biol Sci Med Sci 2003;58:249–65.
13. Juster FT, Suzman R. An overview of the Health and Retirement Study. J Hum Resour 1995;30:S7–56.
14. Lindau ST, Drum ML, Gaumer E, Surawska H, Jordan JA. Prevalence of high-risk human papillomavirus among older women. Obstet Gynecol 2008;112:979–89.
15. Melville JL, Fan MY, Rau H, Nygaard IE, Katon WJ. Major depression and urinary incontinence in women: temporal associations in an epidemiologic sample. Am J Obstet Gynecol 2009;201:490 e1–7.
16. Baker DW, Sudano JJ, Albert JM, Borawski EA, Dor A. Lack of health insurance and decline in overall health in late middle age. N Engl J Med 2001;345:1106–12.
17. McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ. Health of previously uninsured adults after acquiring Medicare coverage. JAMA 2007;298:2886–94.
18. Hurd MD, Martorell P, Delavande A, Mullen KJ, Langa KM. Monetary costs of dementia in the United States. N Engl J Med 2013;368:1326–34.
19. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas 1977;1:385–401.
20. Kohout FJ, Berkman LF, Evans DA, Cornoni-Huntley J. Two shorter forms of the CES-D (Center for Epidemiological Studies Depression) depression symptoms index. J Aging Health 1993;5:179–93.
21. Turvey CL, Wallace RB, Herzog R. A revised CES-D measure of depressive symptoms and a DSM-based measure of major depressive episodes in the elderly. Int Psychogeriatr 1999;11:139–48.
22. Han B. Depressive symptoms and self-rated health in community-dwelling older adults: a longitudinal study. J Am Geriatr Soc 2002;50:1549–56.
23. Emptage NP, Sturm R, Robinson RL. Depression and comorbid pain as predictors of disability, employment, insurance status, and health care costs. Psychiatr Serv 2005;56:468–74.
24. Kessler RC, Andrews G, Mroczek DK, Ustun B, Wittchen HU. The world health organization composite international diagnostic interview short-form (CIDI-SF). Int J Meth Psych Res 1998;7:171–85.
25. Webber DA, Bjelland MJ. The impact of work-limiting disability on labor force participation. Health Econ in press [Epub ahead of print on December 5, 2013]. doi: 10.1002/hec.3020.
26. Benítez-Silva H, Buchinsky M, Chan HM, Cheidvasser S, Rust J. How large is the bias in self-reported disability? J Appl Econom 2004;19:649–70.
27. St. Clair P, Bugliari D, Campbell N, Chien S, Hayden O, Hurd M, et al.. RAND HRS data documentation, version L. Santa Monica (CA): Labor & Population Program, RAND Center for the Study of Aging; 2011.
28. Doshi JA, Cen L, Polsky D. Depression and retirement in late middle-aged U.S. workers. Health Serv Res 2008;43:693–713.
29. Pregibon D. Goodness of link tests for generalized linear models. Appl Statist 1980;29:15–24.
30. Pregibon D. Logistic regression diagnostics. Ann Statist 1981;9:705–24.
31. Fultz NH, Herzog AR. Self-reported social and emotional impact of urinary incontinence. J Am Geriatr Soc 2001;49:892–9.
32. Melville JL, Delaney K, Newton K, Katon W. Incontinence severity and major depression in incontinent women. Obstet Gynecol 2005;106:585–92.
33. Sung VW, West DS, Hernandez AL, Wheeler TL 2nd, Myers DL, Subak LL, et al.. Association between urinary incontinence and depressive symptoms in overweight and obese women. Am J Obstet Gynecol 2009;200:557 e1–5.
34. Thom DH, Haan MN, Van Den Eeden SK. Medically recognized urinary incontinence and risks of hospitalization, nursing home admission and mortality. Age Ageing 1997;26:367–74.
35. Harris SS, Link CL, Tennstedt SL, Kusek JW, McKinlay JB. Care seeking and treatment for urinary incontinence in a diverse population. J Urol 2007;177:680–4.
36. Kessler RC, Greenberg PE, Mickelson KD, Meneades LM, Wang PS. The effects of chronic medical conditions on work loss and work cutback. J Occup Environ Med 2001;43:218–25.
37. Goetzel RZ, Long SR, Ozminkowski RJ, Hawkins K, Wang S, Lynch W. Health, absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting U.S. employers. J Occup Environ Med 2004;46:398–412.
38. Holroyd-Leduc JM, Mehta KM, Covinsky KE. Urinary incontinence and its association with death, nursing home admission, and functional decline. J Am Geriatr Soc 2004;52:712–8.
39. Norton C. The effects of urinary incontinence in women. Int Rehabil Med 1982;4:9–14.
40. Fultz NH, Herzog AR. Epidemiology of urinary symptoms in the geriatric population. Urol Clin North Am 1996;23:1–10.
41. Coyne KS, Zhou Z, Thompson C, Versi E. The impact on health-related quality of life of stress, urge and mixed urinary incontinence. BJU Int 2003;92:731–5.
42. Monz B, Chartier-Kastler E, Hampel C, Samsioe G, Hunskaar S, Espuna-Pons M, et al.. Patient characteristics associated with quality of life in European women seeking treatment for urinary incontinence: results from PURE. Eur Urol 2007;51:1073–81.
43. Coyne KS, Kvasz M, Ireland AM, Milsom I, Kopp ZS, Chapple CR. Urinary incontinence and its relationship to mental health and health-related quality of life in men and women in Sweden, the United Kingdom, and the United States. Eur Urol 2012;61:88–95.
44. Steffick DE. Documentation of affective functioning measures in the Health and Retirement Study. Ann Arbor (MI): Survey Research Center, University of Michigan; 2000.
45. Comstock GW, Helsing KJ. Symptoms of depression in two communities. Psychol Med 1976;6:551–63.