Secondary Logo

Journal Logo

Original Research

Predictors of Urinary Incontinence in a Prospective Cohort of Postmenopausal Women

Jackson, Sara L. MD, MPH1,2; Scholes, Delia PhD3,4; Boyko, Edward J. MD, MPH2,5; Abraham, Linn MS3; Fihn, Stephan D. MD, MPH1,2

Author Information
doi: 10.1097/01.AOG.0000236446.17153.21
  • Free

Urinary incontinence is highly prevalent among postmenopausal women. Current knowledge about the epidemiology of incontinence in this group derives largely from cross-sectional, prevalence data. Longitudinal studies of incontinence incidence and remission have been few, and analyses typically are not adjusted for important confounding factors. Reported annual incidence rates have ranged from 1 to 35%,1–8 with rates of remission 0 to 38%.1,3–5,8 In unadjusted analyses for confounders or multiple risk factors, longitudinal studies have identified age,5 estrogen with and without progesterone,2,8,9 obesity,3,7 smoking and certain dietary factors,7 poor pelvic floor contractility,3 gynecologic surgery excluding hysterectomy,3 and hysterectomy during the follow-up period6 as risk factors for urinary incontinence in previously continent women. To better understand risk factors for urinary incontinence while adjusting for multiple confounding factors, we used data from a large study of urinary tract infection among a population-based cohort of 1,017 community-dwelling postmenopausal women followed up prospectively for 2 years.


This analysis used longitudinal data from a study of urinary tract infection risk among postmenopausal women, conducted at Group Health Cooperative from 1998 to 2002. Group Health Cooperative is a group model health maintenance organization with approximately 450,000 members. Women aged 55–75 years were eligible to participate if they had no natural menstrual cycle in the preceding 12 months, resided in Pierce, King, or Snohomish counties of Washington state, and had been enrolled in Group Health Cooperative for at least one year.

The study cohort was comprised of women randomly selected using Group Health Cooperative enrollment files, with supplemental sampling of women with diabetes from the Group Health Cooperative diabetes registry, as previously described.10 Exclusion criteria for all potential participants included residential nursing care, restriction to a wheel chair, dementia or a severe psychiatric disorder, indwelling or intermittent urinary catheterization, end-stage renal disease requiring dialysis, active malignancy other than skin cancer, acute cystitis in the preceding 90 days, or chronic antibiotic use. The Human Subjects Committees of Group Health Cooperative and the University of Washington approved all procedures.

After obtaining informed consent, data from detailed survey data and research clinic visits at baseline, 12- and 24-month follow-up visits, in conjunction with computerized laboratory, hospital, and pharmacy records (for identification and verification of diabetes), were used to identify and measure potential risk factors.

The baseline interview assessed general medical and surgical health information, estrogen exposure (including oral estrogen with or without progesterone, and vaginal estrogen cream), history of urinary tract infection, and diabetes status. Body mass index was obtained from a Group Health Cooperative mammography surveillance database, based upon self-reported height and weight, using the measurement closest to study enrollment. During research clinic visits (baseline, 12-, and 24-month), women submitted clean-catch midstream urine specimens that were cultured for aerobic bacteria and vaginal introital swabs were obtained and cultured for aerobic gram-negative rods using previously described methods.11 Last, postvoid residual bladder volume was measured using a portable ultrasound device (BladderScan BVI 2500+, Diagnostic Ultrascan, Bothell, WA).

Diabetes presence was ascertained by self-report of physician diagnosis or inclusion in the Group Health Cooperative Diabetes Registry. Details of selection method and confirmation of diabetes are described elsewhere.10,12 Women with diabetes responded to a supplemental questionnaire at each visit with queries regarding type of treatment (diet, pill, or insulin) and presence of peripheral neuropathy or retinopathy.

Our primary outcome measures were report of any urinary incontinence or severe incontinence at the 12 or 24-month follow-up visits. Any urinary incontinence was ascertained by asking “Have you had accidental leakage of your urine during the past year?” If yes, amount and severity were determined as previously described.10 Due to variety in incontinence definitions in the literature, we also approximated a validated and internationally recommended measure of incontinence severity, the Sandvik index,13,14 which accounts for both amount and frequency of urinary incontinence, to define severe incontinence. We matched our measures of incontinence amount to the two levels defined by the Sandvik index (1—drops or small splashes [none or few drops or damp underwear] and 2—more [moderately to completely wet underwear or outer clothes wet or leakage to the floor]) and four levels of frequency (1—less than once per month [none in past month], 2—few times per month [once per month or once per week], 3—few times per week [2 or 3 times per week to 4–6 times per week], 4—every day or night or both [once per day or 2 or 3 times per day or 4 or more times per day]) and then multiplied to yield a score of 1–2 (slight), 3–4 (moderate), or 6–8 (severe).13

Initially, study group characteristics at baseline were compared with characteristics of those returning for 12- and 24-month follow-up visits using univariable analyses (χ2 tests and Fisher exact tests when appropriate). We then estimated the proportions of women reporting any, severe or no urinary incontinence at 12- and 24-month follow-up visits, conditional upon urinary incontinence status and diabetes status at baseline (Table 1).

Table 1
Table 1:
Percentage of Women Reporting Any, Severe, or No Incontinence in the Past Year at 1- and 2-Year Follow-up, Conditional Upon Continence Status and Diabetes Status at Baseline and Incidence Density of Incontinence Outcomes

Each characteristic was evaluated as a potential risk factor for urinary incontinence reported at follow-up. We used transitional logistic regression models to account for episodes of urinary incontinence reported at baseline and multiple times during follow-up.15 This method involved evaluating the outcome(s) “(any or severe) urinary incontinence at follow-up,” while adjusting for history of (any or severe) urinary incontinence at the prior visit, and permitted retention of all women in the analysis. Because participants contributed multiple reports over time, this approach also enabled us to account for repeated measures of urinary incontinence per woman.15 We adjusted for additional correlation using generalized estimating equations with a working independence correlation structure. All models were also adjusted for presence of diabetes to account for the supplemental sampling of women with diabetes from the Diabetes Registry.

We compared our results from the transitional logistic regression models with those obtained using conventional logistic regression in stratified analyses. The stratified analyses evaluated the occurrence of any urinary incontinence at follow-up in the subset of women without urinary incontinence at the prior visit and the occurrence of severe urinary incontinence at follow-up among women without severe urinary incontinence at the prior visit.

Finally, we built a prediction model for each outcome using transitional logistic models, adjusted for urinary incontinence at the prior visit and age and diabetes status and included all exposure variables from the individual models in Table 2 that were significantly associated with the presence of incontinence (P<.1). Using a backwards elimination procedure, we then sequentially removed variables, with a P value for model inclusion of 0.1. This procedure was performed with and without physical function score from the Medical Outcomes Study Short Form 36 (SF-36) in the full model to avoid initial overadjustment, because this variable aggregates information from other potential exposures of interest, such as diabetes. Once prediction models were obtained, we evaluated interaction terms between prior report of any or severe urinary incontinence and each factor included in the multivariable models, adjusting for the other variables in the models.

Table 2
Table 2:
Sample Characteristics and Risk Factors for Urinary Incontinence Among Postmenopausal Women
Table 2
Table 2:
Sample Characteristics and Risk Factors for Urinary Incontinence Among Postmenopausal Women (continued)


A total of 1,017 women were eligible and agreed to participate in the study, including 799 women without diabetes and 218 women with diabetes (65 randomly recruited from the Group Health Cooperative enrollment file plus 153 randomly selected from the diabetes registry). Twelve- and 24-month follow-up visits were completed by 87% and 81% of participants, respectively. The rates of participation are detailed in prior publications.12,16

Participants were predominantly white with a mean age of 64 years, and 21% had diabetes. One third of participants had had hysterectomies, and most took oral estrogen (Table 2). Relative to baseline, women following up at the 12- and 24-month follow-up visits, were less likely to report oral estrogen use in the past month or vaginal symptoms, and were more likely to be vaginally colonized with Escherichia coli. They also reported less diabetic peripheral neuropathy and less diabetic retinopathy (Table 2).

A detailed report of the urinary incontinence prevalence data for this cohort has been previously published,10 with 66% prevalence at baseline of any incontinence, and 8% severe incontinence, in the past year (Table 2). Among the 345 women without urinary incontinence at baseline, 65 (19%) at 1 year and 66 (19%) at 2 years reported any incontinence. Ninety-two of 672 (14%) and 96 of 672 (14%) women with urinary incontinence at baseline reported no incontinence at years 1 and 2. Conditional upon whether they reported incontinence at baseline, the occurrence of any or severe incontinence at 1- and 2-year follow-up visits was similar among women with and without diabetes, except that among women reporting any incontinence at baseline, women with diabetes more frequently reported severe urinary incontinence at follow-up (P=.03 for 1-year follow-up) (Table 1). Only one woman with no urinary incontinence at baseline reported severe urinary incontinence at 1- and 2-year follow-up, whereas approximately one half of women with severe urinary incontinence at baseline continued to report it during follow-up.

We used transitional logistic regression to assess associations between potential risk factors and the likelihood of any or severe urinary incontinence at follow-up visits (Table 2). The strongest factor was a report of urinary incontinence at the prior visit, for which the odds ratios were 24.7 for any incontinence and 23.7 for severe incontinence. White race, low physical function score on the SF-36, any vaginal symptom, and lifetime number of urinary tract infections were factors positively associated with both any and severe incontinence at follow-up. Estrogen cream use (past month), self-reported atrophic vaginitis, and diabetic peripheral neuropathy were associated with any continence only, and prior hysterectomy, problems with constipation, and oral estrogen use in the past month were associated with severe incontinence only (Table 2).

Exposures that were evaluated, but were not associated with either outcome measure included income, education, sexual activity in the past month, parity (0, 1–3, 4 or more), smoking history, dyspareunia, any urinary tract infection in the past year, glucose control (no diabetes, HgbA1C≤7.5%, 7.6–8.5%, >8.5%), diabetes duration (no diabetes, <10 years, ≤10 years), and postvoid residual bladder volume (<50 mL, 50–100 mL, >100 mL).

When compared with the results provided by conventional stratified logistic regression analyses, generally the same risk factors were identified; however, the transitional logistic models identified a greater number of statistically significant risk factors but with smaller odds ratios.

In multivariable models, independent predictors of any urinary incontinence at follow-up were white race, estrogen cream use (past month), vaginal dryness, vaginal discharge, lifetime number of urinary tract infections, and diabetic peripheral nephropathy (borderline significance). A history of hysterectomy and report of any vaginal symptom (borderline significance) predicted severe incontinence (Table 3). When physical function score from the SF-36 was included in the models, the lowest category(0–50) was significantly associated with any urinary incontinence (OR 1.5, CI 1.0–2.2) and severe urinary incontinence (OR 2.4, CI 1.3–4.3), and the effect size of diabetic peripheral neuropathy was diminished and not statistically significant for any urinary incontinence. Similarly, for the severe urinary incontinence outcome, inclusion of physical function score diminished the effect size of any vaginal symptom.

Table 3
Table 3:
Predictors of Urinary Incontinence Among Postmenopausal Women

Significant interaction was found between use of estrogen cream and report of urinary incontinence at prior visit for the any urinary incontinence outcome. For those reporting urinary incontinence at the prior visit, the use of estrogen cream was highly associated with report of incontinence at follow-up (OR 7.6, CI 1.1–54.1). Significant interaction was also found between diabetes treatment type and report of severe urinary incontinence at prior visit for the severe incontinence outcome. For those without a report of severe urinary incontinence at the prior visit, diabetes treatment with pills was associated with a report of severe incontinence at follow-up (OR 2.4 CI 1.1–5.0). This association was not found in those reporting urinary incontinence at a prior visit. Instead, for this group it seemed that those treating their diabetes through diet had a higher odds of reporting severe incontinence (OR 8.2, CI 1.0 – 67.6).


This large, population-based study indicates that urinary incontinence is not only highly prevalent among generally healthy postmenopausal women, but also that substantial numbers transition between continence and incontinence annually. Sixty-six percent of these women reported experiencing any urinary incontinence at baseline, while 8% had severe urinary incontinence. Each year of the study, 19% of continent women developed urinary incontinence, while 14% of women with urinary incontinence became continent. White race, vaginal estrogen cream, vaginal symptoms, history of urinary tract infections, and diabetic peripheral neuropathy were independently associated with development of urinary incontinence. Incontinent women with diabetes were more likely to develop severe urinary incontinence during follow-up, while prior hysterectomy and vaginal symptoms predicted new-onset severe urinary incontinence.

Our findings regarding the occurrence and remission of urinary incontinence are generally consistent with the broad ranges described by prior studies.1–8 Incidence of urinary incontinence that is greater than remission is consistent with prior data among postmenopausal women suggesting that the prevalence of urinary incontinence progressively increases into old age,1,3,5,17 concurrent with declining physical function.5,18–20

Previously reported risk factors for urinary incontinence among postmenopausal women have been based primarily on prevalence data, while longitudinal studies have not presented adjusted independent predictors of urinary incontinence. A primary contribution of this study is the inclusion of a wide variety of risk factor data with adjustment for multiple conditions affecting the incidence of urinary incontinence. We found, as have others, that white race was associated with increased risk of urinary incontinence.19,21,22 While physiologic genitourinary differences between white and black women have been reported,19 our study did not address this. Our data demonstrate that lower physical function, as measured by the SF-36, is associated with greater occurrence of incontinence, and confounds the relationships between incontinence and risk factors such as vaginal symptoms and diabetic peripheral neuropathy. Prior hysterectomy was associated with severe incontinence during follow-up, a finding that supports other analyses suggesting that as women age, those who have undergone hysterectomy have an increased risk of urinary incontinence.6,23 Six or more lifetime urinary tract infections predicted urinary incontinence at follow-up, echoing prior data associating recurrent urinary tract infection and incontinence.24 The prospective nature of this study indicates that, even after adjustment for prior urinary incontinence, a history of recurrent urinary tract infection places a woman at higher risk for future urinary incontinence. Anatomic, genetic, or immune factors specific to these women may contribute to this finding.

The presence of vaginal symptoms predicted any and severe urinary incontinence among postmenopausal women, specifically dryness and discharge for any urinary incontinence. Vaginal bacterial colonization and estrogen deficiency could plausibly affect vaginal symptoms. Vaginal colonization with E coli was univariately associated with severe urinary incontinence at follow-up (Table 2). Self-report of atrophic vaginitis, a condition related to estrogen deficiency, was predictive of any urinary incontinence at follow-up. (Table 2). However, women in our study who used exogenous estrogen, in oral or vaginal form, actually reported more frequent and severe incontinence. Oral estrogen replacement has been associated with increased urinary incontinence in postmenopausal women in multiple large epidemiologic studies,2,8,9,25 and we found it to be univariately associated with greater odds of severe urinary incontinence at follow-up (Table 2). Women in this study who used vaginal estrogen cream had greater odds of any urinary incontinence at follow-up (Table 3), but this may have reflected confounding by indication, because vaginal estrogen was formerly used to treat urinary incontinence. Additional study of the relationship between vaginal estrogen and urinary incontinence is warranted, because it is used to treat vaginal symptoms associated with menopause and atrophic vaginitis.

Our earlier analysis of data from this cohort showed that the prevalence of severe urinary incontinence was greater in women with diabetes than those without diabetes. These prospective data now suggest that, over time, women with diabetes and any urinary incontinence are more likely to develop severe urinary incontinence than women without diabetes (Table 1). From the analysis of cross-sectional data we also found that prevalent severe incontinence was associated with the duration of diabetes, treatment type (diet, pill, or insulin), peripheral neuropathy, and retinopathy.10 In the current prospective analysis, diabetic peripheral neuropathy was borderline independently predictive of urinary incontinence at follow-up. Although we do not have a direct measure of bladder function in this study, the higher odds of urinary incontinence in association with diabetic peripheral neuropathy suggests that associated autonomic neuropathy may contribute to the development of urinary incontinence.

This study has several strengths that include a large, population-based sample, a high proportion of participant retention during the 2 years of follow-up, a wide variety of measured risk factors for urinary incontinence including diabetes characteristics, and microbiologic data regarding urinary tract infection and vaginal flora. The ability to adjust for multiple potential confounding factors enabled identification of risk factors that were independently predictive of urinary incontinence for postmenopausal women. Due to a wide variety of urinary incontinence outcome measures in the literature, we used an approximation to a validated measure of incontinence severity, the Sandvik index,13 so that our results may be more easily compared across studies. Potential limitations of this study include limited validity of self-report of incontinence, possibly limited generalizability to women who do not receive care in a setting such as a health maintenance organization, are less healthy, and who have different demographic characteristics, and probable confounding by indication for vaginal estrogen use. Prevalence and measures of incontinence occurrence in this cohort, recruited from a study of urinary tract infection with supplemental enrollment of women with diabetes, may be higher than among other populations. It is also possible that differentially greater loss to follow-up of women more ill with diabetes could have limited our ability to detect stronger associations between urinary incontinence and this metabolic disorder.

We found that urinary incontinence frequently affects postmenopausal women without preexisting bladder or neurologic dysfunction. Independent risk factors differed by severity of incontinence. Potential areas for future prospective study suggested by these data include the roles of vaginal symptoms, diabetic peripheral neuropathy, and hysterectomy in relation to the occurrence of urinary incontinence in postmenopausal women.


1. Herzog AR, Diokno AC, Brown MB, Normolle DP, Brock BM. Two-year incidence, remission, and change patterns of urinary incontinence in noninstitutionalized older adults. J Gerontol 1990;45:M67–74.
2. Grodstein F, Lifford K, Resnick NM, Curhan GC. Postmenopausal hormone therapy and risk of developing urinary incontinence. Obstet Gynecol 2004;103:254–60.
3. Holtedahl K, Hunskaar S. Prevalence, 1-year incidence and factors associated with urinary incontinence: a population based study of women 50-74 years of age in primary care. Maturitas 1998;28:205–11.
4. Moller LA, Lose G, Jorgensen T. Incidence and remission rates of lower urinary tract symptoms at one year in women aged 40-60: longitudinal study. BMJ 2000;320:1429–32.
5. Nygaard IE, Lemke JH. Urinary incontinence in rural older women: prevalence, incidence and remission. J Am Geriatr Soc 1996;44:1049–54.
6. Sherburn M, Guthrie JR, Dudley EC, O'Connell HE, Dennerstein L. Is incontinence associated with menopause? Obstet Gynecol 2001;98:628–33.
7. Dallosso HM, McGrother CW, Matthews RJ, Donaldson MM, Leicestershire MRC Incontinence Study Group. The association of diet and other lifestyle factors with overactive bladder and stress incontinence: a longitudinal study in women. BJU Int 2003;92:69–77.
8. Samuelsson EC, Victor FT, Svardsudd KF. Five-year incidence and remission rates of female urinary incontinence in a Swedish population less than 65 years old. Am J Obstet Gynecol 2000;183:568–74.
9. Hendrix SL, Cochrane BB, Nygaard IE, Handa VL, Barnabei VM, Iglesia C, et al. Effects of estrogen with and without progestin on urinary incontinence. JAMA 2005;293:935–48.
10. Jackson SL, Scholes D, Boyko EJ, Abraham L, Fihn SD. Urinary incontinence and diabetes in postmenopausal women. Diabetes Care 2005;28:1730–8.
11. Gupta K, Stapleton AE, Hooton TM, Roberts PL, Fennell CL, Stamm WE. Inverse association of H2O2-producing lactobacilli and vaginal Escherichia coli colonization in women with recurrent urinary tract infections. J Infect Dis 1998;178: 446–50.
12. Boyko EJ, Fihn SD, Scholes D, Abraham L, Monsey B. Risk of urinary tract infection and asymptomatic bacteriuria among diabetic and nondiabetic postmenopausal women. Am J Epidemiol 2005;161:557–64.
13. Sandvik H, Seim A, Vanvik A, Hunskaar S. A severity index for epidemiological surveys of female urinary incontinence: comparison with 48-hour pad-weighing tests. Neurourol Urodyn 2000;19:137–45.
14. Brown JS, Nyberg LM, Kusek JW, Burgio KL, Diokno AC, Foldspang A, et al. Proceedings of the National Institute of Diabetes and Digestive and Kidney Diseases International Symposium on Epidemiologic Issues in Urinary Incontinence in Women. Am J Obstet Gynecol 2003;188:S77–88.
15. Diggle PH, Heagerty P, Liang K-Y, Zeger SL. Analysis of longitudinal data. New York (NY): Oxford University Press; 2002.
16. Jackson SL, Boyko EJ, Scholes D, Abraham L, Gupta K, Fihn SD. Predictors of urinary tract infection after menopause: a prospective study. Am J Med 2004;117:903–11.
17. Hannestad YS, Rortveit G, Sandvik H, Hunskaar S, Norwegian EPINCONT study. Epidemiology of Incontinence in the County of Nord-Trondelag. A community-based epidemiological survey of female urinary incontinence: the Norwegian EPINCONT study. Epidemiology of Incontinence in the County of Nord-Trondelag. J Clin Epidemiol 2000;53:1150–7.
18. Diokno AC, Brock BM, Herzog AR, Bromberg J. Medical correlates of urinary incontinence in the elderly. Urology 1990;36:129–38.
19. Jackson RA, Vittinghoff E, Kanaya AM, Miles TP, Resnick HE, Kritchevsky SB, et al. Urinary incontinence in elderly women: findings from the Health, Aging, and Body Composition Study. Obstet Gynecol 2004;104:301–7.
20. Wetle T, Scherr P, Branch LG, Resnick NM, Harris T, Evans D, et al. Difficulty with holding urine among older persons in a geographically defined community: prevalence and correlates. J Am Geriatr Soc 1995;43:349–55.
21. Sampselle CM, Harlow SD, Skurnick J, Brubaker L, Bondarenko I. Urinary incontinence predictors and life impact in ethnically diverse perimenopausal women. Obstet Gynecol 2002;100:1230–8.
22. Grodstein F, Fretts R, Lifford K, Resnick N, Curhan G. Association of age, race, and obstetric history with urinary symptoms among women in the Nurses' Health Study. Am J Obstet Gynecol 2003;189:428–34.
23. Brown JS, Sawaya G, Thom DH, Grady D. Hysterectomy and urinary incontinence: a systematic review. Lancet 2000;356: 535–9.
24. Holroyd-Leduc JM, Straus SE. Management of urinary incontinence in women: scientific review. JAMA 2004;291:986–95.
25. Grady D, Brown JS, Vittinghoff E, Applegate W, Varner E, Snyder T, et al. Postmenopausal hormones and incontinence: the Heart and Estrogen/Progestin Replacement Study. Obstet Gynecol 2001;97:116–20.

Cited By

This article has been cited 1 time(s).

Obstetrics & Gynecology
Establishing a Mean Postvoid Residual Volume in Asymptomatic Perimenopausal and Postmenopausal Women
Gehrich, A; Stany, MP; Fischer, JR; Buller, J; Zahn, CM
Obstetrics & Gynecology, 110(4): 827-832.
PDF (185) | CrossRef
© 2006 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.