Urinary incontinence (UI) is a common chronic condition that affects people of all ages.1–4 Although UI increases with age, its prevalence varies widely. The median prevalence of any UI in women across 35 studies was 27.6% (with a range of 4.8–58.4%).5 Even when one accounts for differences in age distribution among studies by comparing studies with similar age groups, there is substantial variation in prevalence estimates.5,6 Although the case definition varies by study to some degree, it is difficult to separate the effect of the definition itself from other factors such as sample selection method, data collection technique, wording of questions, and other factors that could influence the prevalence of UI.6 Hence, it is difficult to fully understand how the case definition influences estimates of prevalence.
One way to understand the influence of case definition is to examine variation in UI prevalence in a single study by case definition while holding constant other factors such as type of UI and common risk factors. It is known that stress UI (loss of urine with cough, sneeze, lifting, or activity) and urge UI (loss of urine with a sudden urge to urinate that is difficult to defer) occur alone and together in an approximate 5:1:3 ratio among women.3,5,7 This ratio, however, differs by age, race, and other risk factors.
In women, common factors associated with UI include childbearing, age, body mass index, previous hysterectomy, and other comorbidities.8 Most prevalence studies report on the relationship between risk factors and any UI without providing information on variation in risk factors by subtypes of UI.9,10 However, stress and urge UI have different pathophysiologies and degrees of severity and, thus, potentially different predisposing risk factors.11 The goal of this study is to estimate variation in prevalence and risk factors by type and severity of UI, using data from the National Health and Nutrition Examination Survey (NHANES) 2001–2002.
MATERIALS AND METHODS
The National Health and Nutrition Examination Survey is a national survey of a representative sample of the U.S. population and comprises two parts: the home interview and the health examination. During the home interview, participants are asked questions about health status, disease history, and diet. The National Health and Nutrition Examination Survey, funded by the Centers for Disease Control and Prevention, releases public use data updates every 2 years.12 Hence, institutional review board approval was not needed. Every 2-year cycle includes a nationally representative sample. Single counties are used as sampling units, and small counties are combined. Within the individual counties, a group of households is selected. Demographic characteristics of the households are screened, and a representative sample is selected from which one or more persons are interviewed.
In the 2001–2002 cycle, the interview included questions about different subtypes of UI (NHANES 2001–2002). Of the 11,039 participants, only individuals 20 years of age and older (ie, 2,536 males and 2,875 females) were asked the UI questions. The analysis herein is limited to the 2,577 of 2,875 (90%) of the females, 20 years of age and older, who had complete data available. The NHANES survey included questions relevant to defining UI and related subtypes. The questions were as follows:
- “During the past 12 months, have you leaked or lost control of even a small amount of urine with an activity like coughing, lifting, or exercise?”
- “During the past 12 months, have you leaked or lost control of even a small amount of urine with an urge or pressure to urinate and you couldn’t get to the toilet fast enough?”
- “During the past 12 months, have you leaked or lost control of even a small amount of urine without an activity like coughing, lifting, or exercise or an urge to urinate?”
Participants who answered in the affirmative to any one question were subsequently asked how often the particular type of UI occurred (ie, “every day,” “a few times a week,” “a few times a month,” or “a few times a year”).
Prevalence of UI was estimated by using the following case definition: Stress UI was defined as an affirmative response to question 1 only, and urge UI was defined as an affirmative response to question 2 only. Mixed UI was defined as affirmative response to questions 1 and 2, and any UI was defined as affirmative response to one or more of the above three UI questions. Severe UI was defined as incontinence daily or a few times per week. Moderate and mild UI were defined as incontinence a few times per month and a few times per year, respectively. Severity of UI was defined solely based on frequency. The NHANES questionnaire did not include questions related to amount of urine lost in incontinent individuals.
Covariates considered in the analysis to explain variation in prevalence included age (20–29, 30–39, … 70–79, 80+), body mass index (BMI less than 18.5 kg/m2 [underweight], BMI=18.5–24 kg/m2 [normal], BMI=25–29 kg/m2 [overweight], BMI greater than or equal to 30 kg/m2 [obese]), ethnicity (white, black, Hispanic, other), underlying diseases using individual binary covariates (diabetes, asthma, arthritis, congestive heart failure, stroke, thyroid disease), parity (nulliparous , parous [one or more deliveries]), previous hysterectomy, hormone replacement (any, none), and smoking (not at all, some days, every day).
Data analysis was performed with a SAS 8.01 (SAS Institute, Inc., Cary, NC). Analysis of covariates was done separately for severe, moderate, and mild UI. This was done for all four types of UI (stress, urge, mixed, and any). Univariable correlations between UI status and the covariates were examined with the χ2 test for categorical variables and the t test for continuous variables. To compare adjusted estimates between incontinence severity and types, multivariate models were adjusted for all predictor variables. Crude and adjusted odds ratios (and 95% confidence intervals) were calculated with a survey logistic model (proc surveylogistic in SAS). Because the NHANES sample is not proportionally allocated to the sampling strata, sample weights provided by the Centers for Disease Control and Prevention were used. Additionally, sampling errors were calculated after accounting for survey stratum and primary sampling unit. Categorical and continuous versions of variables were examined when applicable.
Because the sample size was limited to available data, post hoc power analysis was conducted to determine the detectable prevalence ratios, which depend on the prevalence of urinary incontinence and the prevalence of the risk factors. This study had 80% power to detect prevalence ratios between 1.3 (when prevalence of urinary incontinence and risk factor were relatively large) and 4.0 (when prevalence of urinary incontinence and risk factor were relatively small).
Demographic characteristics as well as UI subtype and severity levels are shown in Table 1. The NHANES 20+ female sample consisted of 55% white, 18% black, 24% Hispanic, and 3% other, with a mean and median age of 51 and 48 years, respectively. A total of 33% had a normal BMI or were underweight, 32% were overweight, and 34% were obese. Twenty-four percent had a high school education, 30% had less than a high school education, and 46% had more than a high school education. The overall prevalence of the remaining risk factors was diabetes (10%), asthma (11%), arthritis (30%), congestive heart failure (3%), stroke (4%), thyroid problem (13%), parity (84%), hysterectomy (36%), hormones (33%), and smoking (45%). The overall prevalence of stress, urge, mixed, and any UI was 23.7%, 9.9%, 14.5%, and 49.2%, respectively, when all degrees (mild, moderate, and severe) of incontinence severity were included. Mild UI was the most common severity level in women with stress-only or urge-only incontinence. On the other hand, most women with mixed UI had severe disease (Table 1). Figures 1 through 4 represent the prevalence of stress, urge, mixed, and any UI, respectively, across all age groups and broken down into the mild, moderate, and severe categories.
For mild UI, risk factor analysis showed that age was a significant independent risk factor for stress, urge, and any UI. White women had a higher prevalence of all types of UI compared with blacks and a higher prevalence of mixed UI compared with Hispanics. Other significant risk factors included BMI for mixed and any UI, thyroid problem for stress UI, and smoking for any UI. History of diabetes and stroke were protective for stress UI (Table 2). Age was a significant risk factor for moderate and severe UI across all types. Additionally, for moderate UI, BMI and white ethnicity were risk factors for stress, mixed, any UI but not for urge UI. Other risk factors included parity for stress and any UI and previous hysterectomy and thyroid problem for mixed UI. History of asthma was protective against urge UI (Table 3). Finally, for severe UI, BMI and previous hysterectomy were risk factors across all types of UI, except urge UI. White women had a higher prevalence of stress and any UI compared with blacks and Hispanics. Parity was a risk factor for stress and any UI, arthritis was a risk factor across all types except stress UI, and thyroid problem was a risk factor for mixed UI (Table 4).
Urinary incontinence remains a prevalent condition that affects women of all ages. Although the prevalence of any UI typically increases with age, this trend varies according to type of UI. In this study, prevalence of stress UI peaked in the fifth decade of life and gradually decreased with older age. Prevalence of urge UI showed a small peak in the fourth decade followed by a slight decline in the fifth decade, and then trended up with age. In contrast, prevalence of mixed UI showed a gradual increase throughout life. These trends in prevalence across decades of life are in agreement with other large population-based epidemiologic studies.3,4,6,13,14
However, actual prevalence estimates vary from one study to another. One major reason for this variation is the way UI is defined (“do you leak urine now?” or “during the last 12 months, have you lost control of your urine?”) as well as frequency of UI episodes (one or more episodes over the past 12 months, three or more episodes over the last month, several times per week, daily, and others).5,7 The more restrictive the question is, the less prevalent cases are captured. Moreover, it has been shown that the answer to the same question of UI may vary based on the way the problem is presented to the patient.15
In the Norwegian EPINCONT (Epidemiology of Incontinence in the County of Nord-Trøndelag) study, cases were defined as presently having UI.3 Their definition was more restrictive than the NHANES definition of any urine loss in the past 12 months because it does not capture those who may have been incontinent in the past year but are currently on medical treatment or had surgical correction and thus are no longer incontinent. As such, the EPINCONT prevalence of UI of 25% (13% stress UI, 3% urge UI, and 9% mixed UI) was lower than our findings.3 In another population-based study in the United States, in which UI was defined as loss of urine at least once per month, the prevalence of UI was shown to be 45%.9 Case definitions also vary across studies according to severity of UI. Studies distinguishing between mild, moderate, or severe UI using a validated severity index16 have demonstrated that, in general, mild and/or moderate UI are more prevalent in the younger age groups, whereas severe UI becomes more prevalent in later years.3,9 Similar trends were observed in our study. However, this generalization did not hold true when each subtype of UI was analyzed separately. Mild UI was the most common severity type in stress and urge UI, severe UI was the most common severity type in mixed and any UI, and moderate UI was somewhere in between for all types.
There are at least two interpretations (neither of which is exclusive) for the variation in prevalence ratios by case severity. First, the underlying stress and urge pathophysiology independently contribute to overall severity of UI. That is, if only one or the other condition is present, UI will tend toward a mild state rather than a severe state when both are present together. We do not know if mild UI is a transient nonpathologic phenomenon or a precursor of moderate and severe UI. Second, variation could represent nondifferential misclassification. Specifically, the false-positive and false-negative case rates could differ by severity. The false-positive rate could be higher for a loose definition of UI (loss of urine in the past 12 months). On the other hand, a more stringent definition could lead to a higher false-negative rate. Therefore, caseness may be linked not only to the type of UI, but also to its severity. In one study, when survey questions were compared with the clinical diagnosis made by a gynecologist after bladder testing, stress UI was underreported, and mixed UI was overreported, whereas urge UI remained unchanged.17
The most frequent and strongest risk factor associations were observed for severe UI subtypes as compared with mild or moderate UI. Age was as an independent risk factor across most types and severities of UI. In addition, adjusted prevalence ratios were generally weak or absent for mild UI in the various age groups. Age was a much more robust predictor for mixed and any UI in cases of severe UI as compared with moderate UI, although the opposite was true for stress and urge UI. This seemingly paradoxical pattern may be the product of disease progression. As individuals age, they are more likely to have both stress and urge UI coexisting together. Consequently, mixed UI is more likely to be expressed as a severe condition.
Compared with blacks and Hispanics, white women were at a significantly higher risk of having UI of any severity (primarily stress, mixed, or any UI). These findings agree with the work of other investigators who have shown that being black is generally protective against UI, in particular stress UI.9,18–20 Some studies show that blacks are at a higher risk of having urge UI.18–20 Our data did not confirm these findings, possibly due to small numbers in the urge UI group. For mild UI, other risk factors included thyroid problem for stress UI, BMI for mixed and any UI, and smoking for any UI. These could be chance associations as it is expected to have a 5–10% chance occurrence of significance when in fact there is no real association. This suggests that, at least for risk factors, mild UI may not be much different from no incontinence.
Body mass index was a risk factor for all types (except urge UI) of moderate and severe UI. In the large Norwegian population study, BMI was strongly correlated with severe UI.10 Parity was another risk factor for moderate and severe stress and any UI. Most investigators agree that the effect of pregnancy and delivery is to weaken the pelvic floor support structures, which in turn leads to increased prevalence of stress UI.11 Hormone replacement therapy, asthma, congestive heart failure, stroke, and diabetes have been identified in other studies as risk factors for UI,5,8,9,21,22 but they failed to show any significance in our study. In fact, stroke was protective for mild stress UI. One explanation could be that women with stroke are less mobile and hence are less prone to activities that lead to stress UI. It is not clear why diabetes was a protective factor for mild stress and any UI. Although it also appears that asthma is a protective factor for moderate urge UI, the wide range of 95% confidence intervals indicates that this may simply be a chance occurrence.
Arthritis was a risk factor for severe urge, mixed, and any UI. Arthritis patients usually have mobility problems and may have difficulty getting to the bathroom on time. Hysterectomy is reported as a risk factor for UI in some studies and not others.5 In our study, previous hysterectomy was an independent risk factor for moderate mixed UI and severe stress, mixed, and any UI. Finally, history of thyroid problems was an independent risk factor for moderate and severe mixed UI. This was an incidental finding in the NHANES data analysis, and it has not been reported in prior studies to be associated with UI. Because hypothyroidism is much more common than hyperthyroidism, one would assume that UI is associated with the former. This interaction needs further research to understand the underlying pathophysiology of UI in women with thyroid problems.
Based on these findings, it is not clear if mild UI represents a pathophysiologic precursor state of more severe forms of UI. Longitudinal data would be required to determine if mild UI is a transient condition or a true precursor of progression to a more severe state. Hence, including all levels of severity as cases for clinically significant UI has a “dampening” effect on the association between prevalence of UI and its risk factors. Women designated as having mild UI and some with moderate UI may be more likely false-positive cases. The optimal case definition would require a threshold below the severe group used in this study, but greater than the mild UI criteria. Additionally, a valid case definition may need to include not only the frequency of urine loss but an estimate of the amount of urine loss as well.16 Unfortunately, the NHANES questionnaire did not include questions related to amount of urine lost in incontinent individuals.
A limitation of the NHANES data are that the definition of any loss of urine in the past 12 months does not necessarily mean that an individual is presently complaining of or bothered by symptoms of UI. Also, the cutoffs for UI severity were based on the authors’ judgment of what is clinically relevant, and they need to be validated. Another weakness is that our study is a cross-sectional analysis with no follow-up data available. It is well established that UI is a dynamic condition whereby significant incidence rates are associated with equally significant remission rates.23,24 Finally, some UI subtype sample sizes were relatively small (eg, mild mixed UI [n=47] and moderate urge UI [n=58]). Hence, making a good estimate for rare outcomes would be difficult, but it is the best one can do with the existing data.
Despite these weaknesses, we believe that the NHANES data present the best available estimates of the true national prevalence of UI and its subtypes in women over the age of 20. Another unique feature of this study is that it demonstrates how risk factors of UI vary based on its definition and severity level. In summary, UI remains a highly prevalent health condition and is associated with several risk factors that vary based on the case definition. The ultimate validation of criteria will require longitudinal studies that will allow investigators to refer back to baseline criteria to distinguish a true-positive or persistent case from a false-positive or transitional case.
1. Thomas TM, Plymat KR, Blannin J, Meade TW. Prevalence of urinary incontinence. Br Med J 1980;281:1243–5.
2. Thom D. Variation in estimates of urinary incontinence prevalence in the community: Effects of differences in definition, population characteristics, and study type. J Am Geriatr Soc 1998;46:473–80.
3. Hannestad YS, Rortveit G, Sandvik H, Hunskaar S. A community-based epidemiological survey of female urinary incontinence: The Norwegian EPINCONT study. J Clin Epidemiol 2000;53:1150–7.
4. Milsom I. The prevalence of urinary incontinence. Acta Obstet Gynecol Scand 2000;79:1056–9.
5. Minassian VA, Drutz HP, Al-Badr A. Urinary incontinence as a worldwide problem. Int J Gynaecol and Obstet 2003;82:327–38.
6. Hunskaar S, Burgio K, Diokno A, Herzog AR, Hjalmas K, Lapitan MC. Epidemiology and natural history of urinary incontinence in women. Urology 2003 62 (suppl 4A):16–23.
7. Hunskaar S, Lose G, Sykes D, Voss S. The prevalence of urinary incontinence in women in four European countries. BJU Int 2004;93:324–30.
8. Bump RC, Norton PA. Epidemiology and natural history of pelvic floor dysfunction. Obstet Gynecol Clin North Am 1998;25:723–46.
9. Melville JL, Katon W, Delaney K, Newton K. Urinary incontinence in US women: a population-based study. Arch Intern Med 2005;165:537–42.
10. Hannestad YS, Rortveit G, Daltveit AK, Hunskaar S. Are smoking and other life style factors associated with female urinary incontinence? The Norwegian EPINCONT study. BJOG 2003;110:247–54.
11. Nygaard IE, Heit M. Stress urinary incontinence. Obstet Gynecol 2004;104:607–20.
12. National Center for Health Statistics. National Health and Nutrition Examination Survey (NHANES). Available at: www.cdc.gov/nchs/nhanes.htm
. Retrieved November 19, 2007.
13. Stewart WF, Van Rooyen JB, Cundiff GW, Abrams P, Herzog AR, Corey R, et al. Prevalence and burden of overactive bladder in the United States. World J Urol 2003;20:327–36.
14. Milsom I, Abrams P, Cardozo L, Roberts RG, Thuroff J, Wein AJ. How widespread are the symptoms of an overactive bladder and how are they managed? A population-based prevalence study [published erratum appears in BJU Int 2001;88:807]. BJU Int 2001;87:760–6.
15. Fultz NH, Herzog HR. Prevalence of urinary incontinence in middle-aged and older women: a survey-based methodological experiment. J Aging Health 2000;12:459–69.
16. 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.
17. Sandvik H, Hunskaar S, Vanvik A, Bratt H, Seim A, Hermstad R. Diagnostic classification of female urinary incontinence: an epidemiologic survey corrected for validity. J Clin Epidemiol 1995;48:339–45.
18. Bump RC. Racial comparisons and contrasts in urinary incontinence and pelvic organ prolapse. Obstet Gynecol 1993;81:421–5.
19. Sze EH, Jones WP, Ferguson JL, Barker CD, Dolezal JM. Prevalence of urinary incontinence symptoms among black, white, and Hispanic women. Obstet Gynecol 2002;99:572–5.
20. Burgio KL, Matthews KA, Engel BT. Prevalence, incidence and correlates of urinary incontinence in healthy, middle-aged women. J Urol 1991;146:1255–9.
21. Grady D, Brown JS, Vittinghoff E, Applegate W, Varner E, Snyder T; The HERS Research Group. Postmenopausal hormones and incontinence: the Heart and Estrogen/Progestin Replacement Study. Obstet Gynecol 2001;97:116–20.
22. 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.
23. Nygaard IE, Lemke JH. Urinary incontinence in rural older women: prevalence, incidence and remission. J Am Geriatr Soc 1996;44:1049–54.
24. 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.