THE WHITE, cheesy discharge and vulvovaginal itching characteristic of Candida vaginitis infection is a common experience among women, and a frequent topic of women's magazines. Although characteristic of C vaginitis, these symptoms occur with many other vaginal infections; neither clinical findings (e.g., self-diagnosis, itching, cheesy discharge) or laboratory tests (e.g., potassium hydroxide, Gram stain, vaginal pH) are sensitive and specific predictors of C vaginitis.1 For example, the sensitivity and specificity for self-diagnosis are 35% and 89%, with a positive predictive value of 62%. For KOH, the sensitivity and specificity are 61% and 77%, with a positive predictive value of 56%. Regardless, C vaginitis is often diagnosed over the telephone, or it is self-diagnosed and treated.2
Over-the-counter sales of antifungal vaginal medications were reported by industry sources to be approximately $269 million in 1995.3 Despite the apparent high frequency of infections, there have been few studies describing occurrence or risk factors. The National Health Care Survey does not separate C vaginitis from other vaginal infections, and C vaginitis was not assessed in either the Health Interview Survey or the National Health and Examination Survey. An unknown proportion of women have chronic, recurring C vaginitis, defined as four or more infections in a 12-month period. Although not documented, clinical impression is that these women account for the majority of physician visits for C vaginitis.
We are aware of only one population-based study describing the incidence of C vaginitis, which was calculated from a mailed survey of a random sample of women attending a midwestern University.4 By age 25 years, an estimated 54.7 percent of these women had a self-reported history of at least one presumed C vaginitis infection. The incidence of infection was higher among black women than among women of other racial or ethnic groups, and first infection was associated with age of coitarche. In a case-control study in the same population, history of C vaginitis in the previous year and engaging in oral sex were strong predictors of current infection.5
We estimated the 2-month incidence of presumed C vaginitis and the cumulative probability by age of having experienced at least one presumed episode of C vaginitis. Our data were obtained from a random digit-dialing survey of 2000 women representative of the United States population. We used these results, along with sociodemographic information, to estimate the social costs of presumed C vaginitis. We described elsewhere the frequency and occurrence of vaginal symptoms and C vaginitis during the previous 5 years, and health-seeking behavior in response to those symptoms.2
Random Digit-Dialing Survey
The telephone survey has been described in detail elsewhere.2 Briefly, a random digit-dialing sample of 29,754 US nonbusiness phone numbers was obtained from Survey Sampling, Inc. (Westport, CT). The sample included both listed and unlisted phone numbers. All interviewers were female. Attempts were made at different times of the day and on at least one weekend day. The woman 18 years or older with the most recent birthday was interviewed in all households reached in which there was an eligible woman. The cooperation rate (i.e., number of completed interviews/total number of completed calls + partially completed interviews and refusals, excluding known ineligibles) was 54.0%. Women were asked about their lifetime number of C vaginitis episodes, the date of their first and last episode of C vaginitis, who diagnosed their most recent episode of C vaginitis, and how the infection was treated. Sociodemographic information including age, education, employment status, marital status, and number of children was obtained. The C vaginitis battery included questions developed by Geiger et al.4
Definition of C vaginitis
A woman was considered to have had a C vaginitis if she answered yes to the question, “Has a healthcare professional such as a physician or nurse practitioner ever told you that you had a vaginal yeast infection, yeast vaginitis, candidiasis, or moniliasis?” For descriptive analyses, we used infections occurring in the previous 2 months to minimize the effects of recall. We present infections diagnosed by physical examination, over the phone, and self-diagnosis separately. Although the majority of women (98.7%) knew whether they had been diagnosed with C vaginitis, a substantial proportion (9.1%) of those with a previous diagnosis could not identify the month of the most recent episode. We conservatively assumed that women who did not know the date of their last infection, or if they ever had one, did not have an infection in the previous 2 months. Of the 129 women reporting recent infection, method of diagnosis was known in 127 (98.4%) of cases. We refer to infections as either self-reported or presumed because we cannot validate the accuracy of self-diagnosis, physician diagnosis, or self-report of physician diagnosis.
The social cost of a disease can be defined as the value of the additional goods and services not related to the disease that would be available in the absence of the disease. This cost has two elements: (1) direct costs, or the goods and services used in treatment or prevention valued before any insurance coverage; and (2) indirect costs, or the output lost through disability or premature death. Excluded from this definition are psychic costs of suffering (e.g., pain, discomfort, grief, anxiety).
We estimated the social cost of C vaginitis, and assumed that the indirect costs are negligible in the aggregate. The direct costs of the disease are not negligible, and can be divided into medical expenses (medications and clinic charges) and nonmedical expenses (costs of travel and time involved in obtaining treatment). The assumptions used in estimating these expenses are shown in Table 1.
The assumptions were conventional, except in the case of the before-tax wage ($Q), which serves as an indicator of the output (paid or unpaid) lost in each hour of obtaining treatment. In disease cost analysis, this wage is often assumed to be a fixed multiple of the legal minimum wage; however, we used a separate wage for each patient. Because our random digit-dialing survey did not include questions regarding wages, we estimated each patient's wage on the basis of her age, education, marital status, and number of children. The estimating formula was based on a regression analysis relating the four predictors to the actual hourly wages of women with paid jobs, using data from a household survey of income and expenditure.8 Implicit in this procedure are the assumptions that a woman without a paid job produces unpaid output (e.g., housework), and that the value of this unpaid output is indicated by the wage she would receive if in a paid job.
We describe the 2-month incidence by age, history of a prior episode of C vaginitis, history of four or more episodes of C vaginitis during a 1-year period, and racial or ethnic group. To estimate the annual number of women who had at least one episode of C vaginitis nationally, we multiplied the age- and race-specific 1-year incidence rate estimates by the 1995 age-race distribution of women in the United States.6
We calculated the cumulative probability of C vaginitis by age by applying methods of survival analysis to the data on age of first episode of C vaginitis. Women reporting no lifetime episodes of C vaginitis were considered to be censored at their current age. We used Kaplan-Meier estimates and calculated Hall-Wellner 95 percent confidence bands.10 Differences between ethnic groups in the cumulative probability curve were tested for significance using the log-rank test. Information on year of first episode of C vaginitis was available for 666 of 911 women with a prior episode.
Overall, 6.5 percent of women aged 18 and older reported at least one presumed episode of C vaginitis during the previous 2 months (95% CI, 5.4-7.5%). Women reporting one or more prior episodes of C vaginitis accounted for the majority (90.2%) of those reporting infections during the previous 2 months. In every age interval, the majority of women with presumed C vaginitis during the previous 2 months reported a history of at least one previous physiciandiagnosed infection. The incidence of C vaginitis among women with a history of infection (10.7%) was 12 times higher than that observed among women with no history of infection (0.9%; odds ratio, 12.8; 95% CI, 6.4-26.3).
The 2-month incidence of presumed C vaginitis was highest among women reporting four or more episodes in a 1-year period (31.6%; 95% CI, 24.2-39.0%). Women reporting a 1-year period with four or more episodes of C vaginitis comprised 8.0% of the sample, but accounted for 48 of 129 women (37.2%) reporting C vaginitis during the previous 2 months. The 2-month incidence of C vaginitis among women reporting a 1-year period with four or more episodes of C vaginitis was highest in the youngest age group (Figure 1). C vaginitis among those without a history of four or more infections during a 1-year period occurred at a fairly constant rate throughout the life course.
Women reporting C vaginitis in the previous 2 months were as likely to self-diagnose their most recent infection (48.0%) as to contact a physician by phone (16.5%) or for a physical examination (35.4%). This pattern of diagnosis was observed across all age groups (data not shown). Women with a history of four or more infections were more likely to self-diagnose (68.8%) than to contact a physician by phone (12.5%) or for physical examination (18.8%). Women who had a physician visit were less likely to purchase medication (72.5%) than those diagnosed over the phone (95.2%) or those who self-diagnosed (78.3%), possibly because they used already-purchased medications, or because their physician provided them with medication without charge.
Women self-identifying as black reported almost three times the number of C vaginitis episodes in the previous 2 months (17.4%; 95% CI, 11.2-23.5%) as white women (5.8%; 95% CI, 4.7-6.9%) or women in other racial or ethnic groups (4.8%; 95% CI, 0.7-8.8%). The increased incidence was primarily among women 18 to 44 years (Figure 2). However, black women were less likely than women of other racial or ethnic groups to report a 1-year period with four or more episodes of C vaginitis (blacks, 5.8%; other ethnicities, 8.3%), but were more likely to report a history of at least one physician-diagnosed episode of C vaginitis (blacks, 50.7%; other ethnicities, 44.0%).
Although black women were as likely as other women to treat their episode of C vaginitis during the previous 2 months with medication (80.0% versus 75.5%), black women were more likely to seek a physician's diagnosis. Almost two thirds (62.5%) of black women underwent a physical examination for the diagnosis of C vaginitis during the previous 2 months, compared with 29.9% of white women and 20% of women in other racial or ethnic groups. Only one quarter of black women, compared with half (51.5%) of other women, self-diagnosed C vaginitis during the previous 2 months. Further, black women were as likely to report physician diagnosis regardless of number of prior infections. By contrast, each additional prior episode of C vaginitis among other women reduced the odds of physician diagnosis by 28% (data not shown).
Using these incidences and the 1995 age-race distribution of women in the United States,11 we estimate that 7.9 million women had their last 1995 episode of C vaginitis diagnosed by a physician (including a physical examination), and that 11.4 million women self-diagnosed the infection or were diagnosed by phone consultation. Over-the-counter and prescription medications were used by an estimated 10.2 and 5.7 million women, respectively.
Cumulative Probability of C vaginitis by Age and Race
Overall, 55.7% of women will experience at least one episode of C vaginitis in their lifetime (95% CI, 49.6-62.6%) (Figure 3). The cumulative probability rises faster, and is higher for black women than for other women; lifetime risk of C vaginitis for black women is 62.8% versus 55.0% for other women (log rank P = 0.005).
The annual cost of C vaginitis in 1995 is shown in Table 2. Because it became apparent that the per-capita cost varied substantially between black women and other women, the cost estimates are shown separately for these two groups. For the two groups combined, the cost of the disease in 1995 was $1.8 billion. Approximately half of this amount consisted of charges for doctor visits.
The total annual cost of $414 million among the population of 12.1 million black women implies a per-capita cost of $34, which is more than twice as much as the per-capita cost of $16 for other women. The difference is largely due to (1) a higher incidence of the disease among black women; and (2) a greater tendency for black women to seek (relatively expensive) treatment in person at clinics versus buying over-the-counter medications or getting prescriptions over the phone. It should be noted that the 2:1 ratio in per-capita costs represents a ratio of social costs, and not necessarily a ratio of out-of-pocket expenses after allowance for insurance coverage.
Several of the assumptions of the cost analysis might be questioned, and some sensitivity analysis may therefore be in order. The information used for the cost estimates is presented in Tables 1 and 2 in such a way that the consequences of alternative assumptions can be readily calculated. Some examples of sensitivity analysis are as follows:
- Because of sampling error, the 95% CI for the proportion of women who had a C vaginitis episode in 1995 (PY) ranged from 0.23 to 0.38 among black women and from 0.16 to 0.19 among other women. These intervals imply a range of $1,572 million to $2,055 million for the total cost of all episodes in 1995, in comparison with our point estimate of $1,812 million.
- The cost estimates shown in Table 2 assume that office and lab charges per doctor visit ($DV) were at their median levels. If instead they were at the 90th percentiles.9 the total cost of all episodes in 1995 becomes $2,298 million.
- The cost estimates shown in Table 2 assume that drug prices ($P and $C) are at the mean values of widely used products. $P (for prescription drugs) ranges from $6.63 to $26.52 and $C (over-the-counter drugs) from $2.39 to $15.97. If the lower ends of the two ranges are preferred, the total cost of all C vaginitis episodes in 1995 falls to $1,547 million. If the upper ends are preferred, the total rises to $2,046 million.
- If the values of output per hour ($Q) are halved, the total annual cost of the disease falls from $1,812 million to $1,628 million.
We conclude that even large changes-to the margins of plausibility-in the specific parameters selected do not significantly change the order of magnitude of our disease cost estimate.
The methods used for estimating costs in 1995 can also generate estimates for years beyond. In making such projections, we assume that the number of episodes would grow 0.8% annually, in line with the expected growth rate of the US population as a whole; and that wages ($Q), travel and childcare expenses ($TC), clinic charges ($DV), and medication prices ($P and $C) would all grow 2% annually. The assumed growth rates of the monetary parameters are expressed in real terms (i.e., after adjustment for general inflation). For purposes of our projection, the other parameters shown in Table 1 are assumed to stay at their 1995 values.
Based on these assumptions, the annual cost of C vaginitis rises from $1.8 billion in 1995 to $3.1 billion in 2014. The total cost over the twenty years is $47.8 billion, if we give equal weight to a dollar spent in 1995 and one spent in 1996 or later (no discounting of future costs). If the costs occurring after 1995 are discounted 3% annually, as recommended by the panel of Cost-Effectiveness in Health and Medicine,12 the total 20-year cost has a present value of $34.6 billion; at a 5% discount rate, the total cost is $28.5 billion.
Self-reported history of physician-diagnosed C vaginitis is common. In this random digit-dialing survey of the United States, the strongest predictors of a C vaginitis episode in the previous 2 months were a history of one or more episodes of C vaginitis in the past and black ethnicity. An association with black ethnicity has been reported previously in a college population.4,5,13 In that study, after controlling for history of a previous episode of C vaginitis, the increased risk could not be explained by any of the medical or sexual-history variables studied. Ethnic differences in history of self-reported C vaginitis in our survey also were not explained by differences in sexual history, recent sexual practices, birth-control method, menopausal status, estrogen use, douching, or yogurt consumption. The strongest predictor of C vaginitis in the previous 2 months was history of a previous C vaginitis, which was more common among black women (50.7% versus 44.0% for other women).
We estimated incidence and cumulative probability of C vaginitis by age using self-reported history of C vaginitis among 2000 women participating in an anonymous, random digit-dialing telephone survey. It is possible that women with C vaginitis might have been more likely to participate, leading to an overestimate of the incidence. The study was described to potential participants as regarding “an important public health problem among women,” and more particularly, “about vaginal yeast infections.” Our findings are similar to those found in student populations and in a follow-up survey of women identified from a student health survey at a gynecology clinic; however, both studies had a similar potential to select for women with a history of C vaginitis.4,14 However, C vaginitis is so common that it seems likely that most women know someone who has had C vaginitis or have had one themselves and that the effect of a selection bias, if it occurred, would be small.
Diagnosis of C vaginitis is often inaccurate.15 Thus, it is likely that many of the cases of C vaginitis reported here-even those diagnosed by a physician following a physical examination-may not, in fact, be C vaginitis. If so, our data would overestimate the true incidence in the population. Whether accurately diagnosed, the majority of these infections were treated with medications. The frequent and often inappropriate use of antifungals may result in the emergence of resistant strains or other Candida species that may be more difficult to treat, should vaginal overgrowth occur.
The symptoms of C vaginitis are often mild, and women may be less likely to recall infections not occurring in the recent past. The cumulative lifetime probability of C vaginitis may be underestimated due to such underreporting, particularly in older age groups in which infections may have occurred some time ago. In our data, when we compared incidence estimates from infections in the previous 2 months to the incidence for 3 and 4 months before the survey, there was a considerable drop in the number reported. Thus, we chose to report incidence for the previous 2 months. C vaginitis is not life threatening. Although it interferes with quality of life, women with presumed C vaginitis most likely maintain most of their usual activities (although sexual activity is probably curtailed). However, the personal suffering involved, particularly in severe infections, is considerable. The enormous number of women suffering annually from presumed C vaginitis and the propensity for recurrence underscore the need for a better understanding of its epidemiology and pathogenesis, and stress the need for the development of more accurate, rapid diagnostics and effective treatments.
1. Abbott J. Clinical and microscopic diagnosis of vaginal yeast infection: a prospective analysis. Ann Emerg Med 1995; 25:587–591.
2. Foxman B, Marsh JV, Gillespie B, Sobel JD. Frequency and response to vaginal symptoms among white and black women: results of a random digit dialing survey. J Women's Health 1998; 7:1167–1174.
3. Stock Checker. Feminine health care: a changing market share. Drug Topics 1996;42–47.
4. Geiger AM, Foxman B, Gillespie BW. Epidemiology of vulvovaginal candidiasis among university students. Am J Public Health 1995; 85:1146–1148.
5. Geiger AM, Foxman B. Risk factors for vulvovaginal candidiasis: a case-control study among university students. Epidemiology 1996; 7:182–187.
6. US Bureau of the Census. Resident population of the United States: estimates, by age and sex. Available at: http://www.census.gov/population/estimates/nation/intfile2–1.txt
7. Medical Economics Co. 1995 Drug Topics Red Book. Montvale NJ: Medical Economics Co., 1995.
8. Institute for Social Research, University of Michigan. Panel Study of Income Dynamics, 1992. Available at: http://www.isr.umich.edu/src/psid
9. Practice Management Information Corporation. Physician Fees: A Comprehensive Guide for Fee Schedule Review and Management. Los Angeles: Practice Management Information Corp., 1998.
10. Harris EK, Albert A. Survivorship Analysis for Clinical Studies. New York: Marcel Dekker, Inc., 1991
11. Day, JC. Population Projections of the United States by Age, Sex, Race, and Hispanic Origin: 1995 to 2050 (US Bureau of the Census, Current Population Report no. P25–1130). Washington DC: US Government Printing Office, 1996.
12. United States Department of Health and Human Services. Cost-Effectiveness in Health and Medicine. Washington DC: United States Department of Health and Human Services, 1996.
13. Geiger AM. Vulvovaginal candidiasis: epidemiology and risk factors (dissertation). Ann Arbor: University of Michigan, 1995.
14. McCormack WM, Zinner SH, McCormack WM. The incidence of genitourinary infections in a cohort of health women. Sex Transm Dis 1994; 21:63–64.
15. Sobel JD, Faro S, Force RW, Foxman B, Ledger WJ, Nyirjesy PR, et al. Vulvovaginal candidiasis: epidemiologic, diagnostic, and therapeutic considerations. Am J Obstet Gynecol 1998; 178:203–11.