Indirect Estimation of Chlamydia Screening Coverage Using Public Health Surveillance Data

Transcription

1 American Journal of Epidemiology Copyright 2004 by the Johns Hopkins Bloomberg School of Public Health All rights reserved Vol. 160, No. 1 Printed in U.S.A. DOI: /aje/kwh162 Indirect Estimation of Chlamydia Screening Coverage Using Public Health Surveillance Data William C. Levine 1, Linda W. Dicker 1, Owen Devine 2, and Debra J. Mosure 1 1 Division of STD Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, Atlanta, GA. 2 National Center for Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, Atlanta, GA. Received for publication December 8, 2003; accepted for publication January 16, Although routine screening of all sexually active adolescent females for Chlamydia trachomatis infection is recommended at least annually in the United States, no national or state-specific population-based estimates of chlamydia screening coverage are known to exist. Conclusions regarding screening coverage have often been based on surveys of health care provider or facility screening practices, but such surveys do not consider persons who do not seek care at these facilities or who seek care at more than one facility. The authors developed a method to estimate the proportion of sexually active females aged years screened for chlamydia in 45 states and the District of Columbia by using national data on chlamydia positivity, estimates of sexual activity from the National Survey of Family Growth, and chlamydial infections reported to the Centers for Disease Control and Prevention. Because of uncertainty regarding these values and related assumptions, credibility intervals were calculated by using a Monte Carlo model. When this model was used, the median state-specific proportion of sexually active females aged years screened in 2000 was 60% (90% credibility interval: 55, 66). These results and this method should be evaluated for their utility in guiding implementation of national and state chlamydia control programs. adolescent; chlamydia; mass screening Abbreviations: CDC, Centers for Disease Control and Prevention; NSFG, National Survey of Family Growth. In 1993, the Centers for Disease Control and Prevention (CDC) recommended routine screening of all sexually active adolescent females for Chlamydia trachomatis infection at least annually (1). Estimates of screening coverage among adolescent females (i.e., the proportion of sexually active females aged years who have a chlamydia test during a calendar year) are important for measuring progress toward this goal and allocating resources for screening programs. Similar data on coverage of other preventive services have been important for monitoring progress in achieving public health objectives and assuring the quality of clinical services (2). However, to our knowledge, no national or state-specific population-based estimates of chlamydia screening coverage exist, and they are difficult to obtain. Although data on screening practices have been obtained from surveys or studies of health care facilities or providers (3 9), they supply information only on the practices of these providers and not on population-based coverage of screening services. Even if problems of provider recall and documentation could be overcome, such facility-based surveys do not address the issue that some sexually active teens may not enter services at all, which could result in an overestimate of screening coverage. Perhaps even more significant, many sexually active teens may seek care from more than one facility; for example, they may receive primary health care through a managed care organization but contraceptive services and screening for sexually transmitted infections from public family planning clinics or sexually transmitted disease clinics. Estimates derived from facility-based assessments of screening practices may in some instances substantially underestimate overall coverage. To develop an approach to estimating population-based chlamydia screening coverage, we used readily available national data on chlamydial infection to estimate state- and Correspondence to Dr. Debra J. Mosure, Epidemiology and Surveillance Branch, Division of STD Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, 1600 Clifton Road, NE (MS E02), Atlanta, GA ( djm1@cdc.gov). 91

2 92 Levine et al. region-specific coverage of chlamydia screening of sexually active females aged years in the United States. Although this method is subject to substantial uncertainty regarding some of its assumptions, it may serve as an initial model for population-based estimation of screening parameters when data on prevalence and case reports both exist. As our data and assumptions improve, so should the estimates that result from the use of this method. MATERIALS AND METHODS Estimation of screening coverage For this analysis, screening coverage was defined as the number of females years of age who were tested for chlamydia in each of 45 states and the District of Columbia during 2000 divided by the number of sexually active females aged years in each state. We assumed that the number of females tested would be similar to an estimate of the number of chlamydial infections diagnosed and reported among females years of age divided by an estimate of the number of prevalent chlamydial infections among all sexually active females aged years in these states. Data on the total number of chlamydial infections diagnosed and reported in 2000 in the United States were obtained from sexually transmitted disease surveillance data consisting of case report information provided to CDC. These data include all cases of notifiable sexually transmitted diseases reported to the CDC by state and local health departments (10). The total number of females aged years with diagnosed and reported chlamydial infections in each state was assumed to be the number of chlamydial infections in females aged years reported to CDC, minus a 10 percent adjustment for repeat positive test results for persons within the calendar year. This adjustment for repeat infections was based on an analysis of repeat positive test results in the state of Washington (11). To estimate the number of prevalent chlamydial infections among all sexually active females aged years, we used data on chlamydia positivity (the number of positive tests divided by the total number of tests) among females years of age screened in 2,373 family planning clinics in These data were obtained from the Public Health Service Regional Infertility Prevention Projects, and positivity was assumed to be similar to clinic prevalence, based on prior analyses of this data source (12). The Regional Infertility Prevention Projects collaborate with CDC on monitoring prevalence of chlamydial infection among women routinely screened in selected family planning clinics and other providers in all states (13). To account for females who were not screened in family planning clinics, we applied an estimated ratio of chlamydia prevalence among sexually active females who did and did not attend family planning clinics to state-specific data on chlamydia positivity among females screened in family planning clinics. The ratios applied were from a study that estimated proportional differences in chlamydia positivity among females who did and did not attend family planning clinics based on the differential distribution of risk factors for chlamydial infection and family planning clinic attendance (14). We used the most recent available postcensal population estimates (1999), based on Bureau of the Census data from the 1990 US Census (15), to obtain the numbers of females years of age in each state. We then estimated the number of those females who were sexually active by using data from the National Survey of Family Growth (NSFG) (16). In this survey, personal interviews were administered to a nationally representative sample of 10,847 females aged years in The survey collected data on factors affecting women s reproductive health, including sexual activity. Because state-specific estimates of sexual activity cannot be obtained from this survey, we used the NSFG region-specific estimates (Northeast, South, Midwest, West). To estimate the number of sexually active females aged years in each state, we multiplied the intercensal population estimate for the state by the regional estimate of the proportion of females aged years who reported being sexually active. We estimated the proportion of sexually active females aged years who use family planning clinics in each state by dividing the number of females aged years who used family planning clinics in 1999 (obtained from federal Title X family planning program grantees) (17) by our estimate of the number of sexually active adolescent females in each state. We did not have precise data available to estimate screening coverage for each state. A Monte Carlo model was used to account for the uncertainty associated with several of the parameters used to estimate screening coverage in each state (18). We modeled the uncertainty associated with the estimates of the proportion of females aged years who were sexually active by using the standard errors associated with these estimates provided through the NSFG. To do this, we assumed that the range of possible values for the true state-specific proportion of sexually active females in this age category was normally distributed with a mean equal to the survey estimate and a standard error equivalent to that associated with the estimate. A discrete distribution was constructed to reflect uncertainty concerning the true ratio of the chlamydia positivity among females years of age who did not visit a family planning clinic to the positivity among those who did. Previous estimates of the prevalence of chlamydial infection among females aged years in US Public Health Service Region X who visited family planning clinics ranged from 10.6 percent to 12.6 percent (14). Positivity estimates from the same region for females years of age who did not visit family planning clinics ranged from 9.9 percent to 11.5 percent. Using a Monte Carlo approach, we drew 5,000 independent samples from two normal distributions constructed to coincide with these observed positivity ranges to derive a sample of possible values for the non-familyplanning-clinic users/family planning clinic users positivity ratio. On the basis of the histogram produced by repeated estimation of this ratio, we estimated that the probability that the true positivity ratio for family planning clinic nonusers to users is 0.8 equals 0.1, the probability that this ratio is 0.9 equals 0.6, the probability that the ratio is 1.0 equals 0.28,

3 Estimation of Chlamydia Screening Coverage 93 and the probability that the ratio is 1.1 equals The sum of the probabilities associated with these discrete possible values for the ratio equals one, implying that the true, yet unknown positivity ratio falls in this interval (0.8 to 1.1). To reflect the uncertainty, a Monte Carlo model was run 1,000 times and included uncertainty estimates for the regional estimates of sexual activity provided in the NSFG, an estimate of uncertainty regarding the ratio of chlamydia prevalence among females attending family planning clinics to those who did not, and an estimate of uncertainty about the proportion of sexually active females who used these clinics. Estimates of screening coverage, and a 90 percent credibility interval for those estimates, were generated for each state. The credibility interval was defined as the interval separating the 5th and 95th percentiles for the screening coverage estimate generated by the Monte Carlo model. TABLE 1. Median state-specific chlamydia positivity among females aged years screened in family planning clinics, and median chlamydia screening coverage and 90% credibility interval, by US Census region, 2000 Geographic region Median chlamydia positivity (%) Estimated median chlamydia screening coverage % 90% credibility interval Northeast , 53.7 South , 66.3 Midwest , 77.0 West , 59.7 All regions , 65.7 Exclusions New York State was excluded from this analysis because genital C. trachomatis infections were not reportable until mid Four states (Connecticut, Maine, Minnesota, Rhode Island) were excluded because their 2000 Regional Infertility Prevention Project data were incomplete. Comparison of estimates with local data and perceptions To assess the face validity (19) of assumptions and the credibility of the state-specific coverage estimates, a written summary of the assumptions and estimates was provided to the 10 Regional Infertility Prevention Project coordinators, who then distributed them to chlamydia screening program staff in each of the 45 states for which coverage estimates were made. Responses from each state were collated by the coordinators, who provided written responses and discussed this information with us. RESULTS The median state-specific chlamydia positivity among females aged years screened in family planning clinics in 45 states and the District of Columbia in 2000 was 5.7 percent (table 1). The estimated screening coverage among sexually active females aged years was 60 percent (90 percent credibility interval: 55, 66). Screening coverage was highest in the Midwest (70 percent), followed by the South (61 percent), the West (54 percent), and the Northeast (49 percent). State-specific estimates of screening coverage ranged from 27 percent to more than 100 percent (table 2). Comparison of these estimates with local data and perceptions revealed two major concerns. First, public health representatives from many states thought that the NSFG regional estimates of sexual activity did not accurately reflect levels of sexual activity among adolescent females in their state. Some of these representatives thought that the estimates of sexual activity were too high, while others thought that they were lower than they expected. Second, representatives from several states also thought that the chlamydia positivity among adolescent females who were screened at family planning clinics was not representative of the overall chlamydia positivity among females aged years in their state. In some instances, the family planning clinics were thought to be located in areas of notably high prevalence; in others, the family planning attendees were thought to have a lower prevalence than others because they were being routinely screened. Representatives from the states with very high estimates of screening coverage thought that high coverage in certain subgroups (and possibly multiple tests performed within the calendar year) could account for the high estimate of screening coverage (e.g., >100 percent) for their state. For example, perceptions were that there were substantial disparities in coverage between urban and rural areas or among race/ethnicity groups that may have skewed the assumptions and estimates. However, representatives from the states with relatively low screening coverage estimates agreed that the coverage for adolescent females in their state was most likely low. DISCUSSION This analysis indicates that an estimated percent of sexually active adolescent females aged years in 45 states and the District of Columbia were screened for chlamydial infection in 2000, compared with national CDC (1) and US Preventive Services Task Force (2) guidelines that recommend screening all of these females at least annually. These proportions correspond to a lack of diagnosis and treatment of percent of prevalent chlamydial infections in this population. The estimate of chlamydia screening coverage was less than 40 percent in several states, suggesting that over 60 percent of chlamydial infections in sexually active adolescent females had not been diagnosed and treated. Screening coverage among these females was highest in the Midwest and lowest in the Northeast. These figures could be overestimates of true screening coverage if the prevalence of chlamydia among females aged years was substantially higher than the estimated prev-

4 94 Levine et al. TABLE 2. Chlamydia positivity among females aged years screened in family planning clinics, and estimated chlamydia screening coverage and 90% credibility interval, by US state, 2000 Geographic region and state Chlamydia positivity (%) Estimated chlamydia screening coverage (%) 90% credibility interval Northeast Massachusetts , 59.2 New Hampshire , 40.4 New Jersey , 41.8 Pennsylvania , 62.4 Vermont , 59.9 South Alabama , 83.3 Arkansas , 61.3 Delaware , District of Columbia , Florida , 85.2 Georgia , 74.4 Kentucky , 92.7 Louisiana , 62.7 Maryland , 88.2 Mississippi , 43.1 North Carolina , 75.6 Oklahoma , 81.2 South Carolina , 49.5 Tennessee , Texas , 55.9 Virginia , 62.1 West Virginia , 84.8 Table continues TABLE 2. Continued Geographic region and state Chlamydia positivity (%) Estimated chlamydia screening coverage (%) 90% credibility interval Midwest Illinois , 60.8 Indiana , 82.5 Iowa , 74.3 Kansas , 88.6 Michigan , Missouri , Nebraska , 77.1 North Dakota , 78.1 Ohio , 80.2 South Dakota , Wisconsin , 72.1 West Alaska , Arizona , 79.8 California , 57.4 Colorado , 71.1 Hawaii , 86.3 Idaho , 56.5 Montana , 58.9 Nevada , 59.3 New Mexico , 75.1 Oregon , 95.5 Utah , 31.1 Washington , 60.0 Wyoming , 66.7 alence or if the proportion of sexually active females was higher than that estimated by the NSFG, even after uncertainty was factored into the Monte Carlo model. Conversely, the figures for coverage could be underestimates if the actual number of females aged years diagnosed with chlamydial infection was substantially higher than the numbers reported by clinicians and laboratories to state health departments. One household survey of chlamydial infection among persons years of age in the San Francisco Bay Area of California demonstrated substantial variation (both higher and lower) in chlamydia positivity compared with screening data from three local family planning clinics (20); however, that comparison did not consider potential differences in prevalence among persons who did and did not attend family planning clinics, as we attempted to do. One source of error that may not be adequately accounted for by the credibility intervals is the heterogeneity of coverage among populations within a state. In those states reported to have very high coverage, it is possible that certain subpopulations are being screened multiple times during a year, even while coverage may be low in other subgroups. For example, analysis of data by race/ethnicity from one state found that among females aged years in the other race/ethnicity group (most likely Native Americans), coverage was estimated to be 150 percent. In that same state, there were lower coverage estimates for Hispanics (38 percent) and African Americans (32 percent) (unpublished data, CDC). We are also aware of geographic differences in screening coverage. For example, chlamydia screening coverage among sexually active females years of age in one city was estimated to be 84 percent, while the estimate for the rest of the state outside of that city was 54 percent; for another city, coverage was estimated to be 93 percent and for the remainder of the state 50 percent (unpublished data, CDC). These examples demonstrate the importance of adjusting for demographic and geographic variables (e.g., race/ethnicity, age, and place of residence) that may be related to prevalence when comparing family planning clinic data, case reports, and census data. Another possible source of error arises from the lack of data on test type used for reported cases of chlamydial infection. We know that approximately 20 percent of females

5 Estimation of Chlamydia Screening Coverage 95 attending the family planning clinics considered for our prevalence estimates were tested by using nucleic acid amplification methods, which are more sensitive than other methods. If such tests were used in higher proportions for females attending family planning clinics than for females outside of these settings, our coverage figures would be underestimates. These state-specific estimates should be considered starting points that can be compared with other data on coverage that may be available from other sources, for example, surveys of health care providers chlamydia screening practices (3 5) or systematic review of medical records of persons attending health care facilities (6 9). The Health Employer Data Information Set (HEDIS) measure for chlamydia screening of sexually active adolescents is based on record review of participating managed care organizations (7). In 2000, this data set estimated that the chlamydia screening coverage for sexually active females aged years was 27.5 percent (8). A report on chlamydia screening by health care providers participating in California s statefunded family planning program during fiscal year found that 68 percent of eligible females aged years attending 586 public clinics were screened for chlamydia, as were 53 percent of those attending 1,301 participating private clinics (9). These data do not differ greatly from our coverage estimate for California of 50 percent (90 percent credibility interval: 44, 57). It is important to keep in mind that these sources of data focus on the provider or the persons attending specific facilities and do not account for adolescents who do not attend these services or those who attend both these and other services where they may be screened. The estimates developed by using the method presented here are independent of these factors. National population-based data on community prevalence of chlamydial infection are now being obtained through the National Health and Nutrition Examination Survey (21). In the future, these data may be used to develop more accurate national and regional estimates of chlamydia coverage using the method described here. With the data from the National Health and Nutrition Examination Survey, it will not be necessary to derive estimates of prevalence based on family planning clinic data, which we used here because it is currently the only known available source of state-specific data on prevalence of chlamydial infection. Although our estimates of screening coverage are subject to the limitations discussed above, they are consistent with other sources of information demonstrating that among sexually active adolescent females in the United States, the burden of undiagnosed and unreported chlamydial infections remains high. These data should be discussed by national, state, and local partners in chlamydia prevention programs and compared with local information on chlamydia screening coverage to guide expansion of screening programs to prevent and treat this disease. REFERENCES 1. Recommendations for the prevention and management of Chlamydia trachomatis infections, Centers for Disease Control and Prevention. MMWR Recomm Rep 1993;42(RR- 12): US Preventive Services Task Force. Guide to clinical preventive services. 2nd ed. Baltimore, MD: Williams & Wilkins, Chlamydia prevalence and screening practices San Diego County, California, MMWR Morb Mortal Wkly Rep 1994;43:366 9, Chlamydia screening practices of primary-care providers Wake County, North Carolina, MMWR Morb Mortal Wkly Rep 1997;46: St Lawrence JS, Montano DE, Kasprzyk D, et al. STD screening, testing, case reporting, and clinical and partner notification practices: a national survey of US physicians. Am J Public Health 2002;92: Shafer MB, Tebb KP, Pantell PH, et al. Effect of a clinical practice improvement intervention on chlamydial screening among adolescent girls. JAMA 2002;288: National Committee for Quality Assurance (NCQA). HEDIS 2000: technical specifications. Washington, DC: NCQA, 1999: 68 70, National Committee for Quality Assurance (NCQA). The state of managed care quality, Chlamydia screening in women. ( 2001_CHL.html#fn1). (Accessed October 13, 2003). 9. Chow JM, Bradsberry M, Treat J, et al. Chlamydia screening and test type utilization in California publicly funded family planning services National STD Prevention Conference Abstracts. Abstract B5E. ( dstd/2002confabstracts/2002confaboralb 2.htm). (Accessed October 13, 2003). 10. Sexually transmitted disease surveillance, Atlanta, GA: Division of STD Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, September ( Survtext2000.pdf). 11. Xu F, Schillinger JA, Markowitz LE, et al. Repeat Chlamydia trachomatis infection in women: analysis through a surveillance case registry in Washington State, Am J Epidemiol 2000;152: Dicker LW, Mosure DJ, Levine WC. Chlamydia positivity versus prevalence. What s the difference? Sex Transm Dis 1998; 25: Sexually transmitted disease surveillance 2000 supplement. Chlamydia Prevalence Monitoring Project annual report Atlanta, GA: Division of STD Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention, November ( chlamydia2000/ct2000.pdf). 14. Mosure DJ. Adolescents, family planning, chlamydia and their relationships. Unpublished doctoral dissertation. Emory University, Atlanta, GA, US Bureau of the Census estimates of the population of counties by age, sex, and race/hispanic origin: 1990 to Washington, DC: US Bureau of the Census. (Machinereadable data files). 16. Abma JC, Chandra A, Mosher WD, et al. Fertility, family planning, and women s health: new data from the 1995 National Survey of Family Growth. Vital Health Stat ;May: Alan Guttmacher Institute. Family planning annual report: 1999 summary, part 1. New York, NY: Alan Guttmacher Institute, Morgan MG, Henrion M. Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis. New York,

6 96 Levine et al. NY: Cambridge University Press, Krowinski WJ, Steiber SR. Measuring and managing patient satisfaction. Chicago, IL: American Hospital Publishing, Inc, 1996: Klausner JD, McFarland W, Bolan G, et al. Knock-knock: a population-based survey of risk behavior, health care access, and Chlamydia trachomatis infection among low-income women in the San Francisco Bay area. J Infect Dis 2001;183: Mertz KJ, McQuillan GM, Levine WC, et al. A pilot study of the prevalence of chlamydial infection in a national household survey. Sex Transm Dis 1998;25:225 8.