Lechtenberg, Richard J. MPH*†; Samuel, Michael C. DrPH†; Bernstein, Kyle T. PhD, ScM*‡; Lahiff, Maureen PhD*; Olson, Nicole MPH†; Bauer, Heidi M. MD, MS, MPH*†
Neisseria gonorrhoeae (NG) is the second most common reportable infectious disease in the United States.1 Untreated, NG infection in women can lead to pelvic inflammatory disease, which heightens the risk for ectopic pregnancy and infertility.2 In addition, untreated NG infection has been shown to facilitate the transmission and acquisition of HIV.3
Since NG became treatable in the 1940s, the development of antibiotic resistance has prompted the Centers for Disease Control and Prevention (CDC) to repeatedly update their treatment guidelines to ensure adequate treatment of infections and slow the emergence of continued resistance.4,5 Most recently, CDC removed orally administered cephalosporins such as cefixime as recommended first-line treatment of NG, given data suggesting declining antimicrobial susceptibility, leaving administration of the injectable cephalosporin ceftriaxone in combination with azithromycin or doxycycline as the sole remaining recommended first-line treatment option.6
Previous research has documented substantial variation by clinical practice setting in the quality of sexually transmitted disease (STD) care provided in California and nationally. Deficiencies in adherence to screening, treatment, and reporting guidelines have been found among county clinics,7 private physicians,8 and public hospital outpatient clinics.9 One limitation of the existing research on the quality of STD care has been a focus on relatively small geographic areas or a limited number of different clinical settings. Knowledge of how adherence to the NG treatment guidelines varies by clinical setting may identify potentially important targets for education and other interventions to reduce barriers to appropriate treatment. This could be particularly useful in California, which covers a substantial portion of the Pacific region from which antimicrobial resistance has historically spread to the rest of the United States.10 The present analysis investigates this variation in adherence using data collected from a variety of clinical settings throughout the state.
MATERIALS AND METHODS
The California Gonorrhea Surveillance System (CGSS) is a supplemental surveillance platform consisting of a geographically representative probability sample of NG cases. A random sample of reported NG infections is drawn weekly for CGSS and assigned for telephone interview by trained disease intervention specialists. Participants are eligible for interview if they are English or Spanish speaking, are older than 14 years, and have not had another NG diagnosis in the prior 30 days. Interviews are conducted both with participants and their providers. Because the sampling probability differs by local health jurisdiction (LHJ) and has varied from year to year, the data are weighted. For this analysis, the weights were calculated as the ratio of the number of participants reported in a given local health jurisdiction and year divided by the number of participants for whom follow-up occurred with both the participant and their provider. Additional details regarding methodologic aspects of the CGSS can be found elsewhere.11
Participants were included in this analysis if they were diagnosed as having confirmed, uncomplicated anogenital and/or pharyngeal NG in California between January 1, 2009, and December 31, 2011. They were excluded from this analysis if there was no documentation regarding treatment of NG. Participants were also excluded if they were diagnosed as having nongenital infection (e.g., gonococcal conjunctivitis) or complicated infection (e.g., pelvic inflammatory disease or disseminated gonococcal infection). Transgender participants were also ultimately excluded from the analysis because of small numbers. Participants from San Francisco were not included as the San Francisco Department of Public Health disseminated local guidelines differing from the CDC’s through the end of 2010 (K. T. Bernstein, 2013, personal communication).
The primary outcome of this analysis was whether or not participants’ documented treatment of NG conformed to a CDC-recommended (first-line or alternative) treatment regimen appropriate to their specific known or suspected anatomical sites of infection; because the cases analyzed were diagnosed in calendar years 2009 to 2011, the applicable guidelines were either those released by CDC in August 2006 or those released in December 2010 (Table 1). The most notable changes to the CDC guidelines in 2010 were an increase in the dose of ceftriaxone to 250 mg and the addition of cotreatment with azithromycin or doxycycline. To account for the time needed for these changes to be implemented, a 60-day lag was incorporated into the analysis. Any treatment provided within 60 days of the 2010 update to the guidelines was considered guideline compliant as long as it adhered to the previous recommendations. Two sensitivity analyses were conducted, one with this lag shortened to 30 days and another with it lengthened to 90, to ensure that our results were not unduly influenced by the choice of lag. An additional sensitivity analysis was conducted to determine how robust our results were to the decision to classify as recommended those regimens consisting solely of azithromycin 2 g. Although the treatment guidelines specify azithromycin 2 g as appropriate treatment for persons who are allergic to cephalosporins, data on the presence of cephalosporin allergy were not available. A final sensitivity analyses excluded participants without reported treatment or with treatment more in line with a chlamydia-only diagnoses (i.e., azithromycin 1 g and/or doxycycline 100 mg for ≥7 days) because some of these participants may have received delayed treatment of NG that simply had not yet been given at the time of provider report.
All covariates were categorical, coded by sets of mutually exclusive indicator variables. Owing to small numbers, emergency department (ED) and urgent care were combined into a single category, as were hospital inpatient and hospital outpatient clinic. Other clinical types examined included military/Veterans’ Affairs (VA), correctional facility, private physician/health maintenance organization (HMO), family planning facility, community/public health clinic, HIV clinic, and STD clinic.
Variables expected to confound the association between clinical practice setting and treatment were pregnancy, HIV infection, or being a man who has sex with men (MSM). It was also expected that geography might confound the association between clinical setting and treatment. Geography was operationalized using 7 regions defined by the California Department of Public Health (CDPH) STD Control Branch (details can be found in the 2011 annual report12). With the exception of clinical setting and region, all covariates were reported by the patient.
The analytic data set was prepared in SAS 9.3 (Cary, NC) and analyzed in R version 18.104.22.168 Bivariate associations between treatment and individual predictors were assessed first; statistical significance was assessed using weighted χ2 tests for independence. Predictors exhibiting an association with treatment significant at the α = 0.20 level but not included among the set of a priori confounders were included as covariates in multivariable analyses.14
The measure of association examined in multivariable analyses was the cumulative incidence ratio—henceforth referred to as relative risk (RR)—of receiving non-recommended treatment. Relative risks were estimated by fitting weighted generalized linear models with a log link function and using the svyglm function in the Survey package of R.15 Estimation was first attempted with the error distribution specified as binomial. When the model failed to converge as commonly arises with binomial regression models (potentially due to the insufficiency of the statistical software package’s default convergence criterion or proximity of the maximum likelihood estimate to a boundary of the parameter space16,17), the Poisson distribution family was specified instead with robust SEs (svyglm’s default) estimated by Taylor series linearization.18,19 All variables showing a significant bivariate association with treatment were included as covariates in the initial model, in addition to clinical setting and all a priori confounders. Adjusted Wald tests were performed using the RegTermTest function in R’s Survey package to test the significance of sets of coefficients. Variables were dropped from the model if their independent associations with treatment were not significant at an α level of 0.10.
This research protocol received a determination of nonresearch from the Committee for the Protection of Human Subjects of the State of California’s Health and Human Services Agency and was approved by the Committee for the Protection of Human Subjects at the University of California, Berkeley.
Of the 78,369 cases of NG reported to CDPH between 2009 and 2011, 7735 (9.9%) were randomly selected for follow-up as part of the CGSS. For 1598 cases (20.7% of those sampled), the interview with the provider was not completed. For an additional 2516 (32.5%), the patient could not be contacted or refused to be interviewed. Among the 3621 participants for whom interviews were completed with both patient and the provider (46.8% of those sampled), 20 were excluded because of diagnosis with a complicated or nongenital infection. For an additional 359, laboratory data were not sufficiently complete to confirm the presence of anogenital or pharyngeal NG. For 57 of the remaining, treatment data were inadequate to establish concordance with the guidelines. Finally, 10 transgender participants were excluded because of small numbers. Altogether, data from 3178 (41.1%) of the originally sampled participants were available for the analysis.
The sex, age, and region of these participants were compared with that of those participants excluded from analysis on the basis of the above criteria as well as completeness of the clinical setting variable (Table 2). No statistically significant differences were found by sex or age. Small but statistically significant differences were seen for all regions but Sacramento, however; the largest differences were for the Central Coast and Central Valley regions, whose participants were somewhat overrepresented and underrepresented in our sample, respectively. Unlike sex, age, and region, which could be determined from the laboratory reports that usually trigger a case investigation in the first place, race/ethnicity was self-reported and so missing for most of the excluded participants because so many lacked a patient interview. However, no statistically significant differences were found in comparing the distribution of race among participants included in the analysis on the one hand, and, on the other, those excluded from the analysis but for whom data on race were available (P = 0.44).
Overall, 14.9% of reported and interviewed participants received non-recommended treatment (95% confidence interval [CI], 13.2–16.7). In bivariate analyses of the unadjusted association of treatment with various characteristics (Table 3), receipt of non-recommended treatment was found to be significantly associated with sex, age, race/ethnicity, sexual orientation, and clinical setting, but not with region or pregnancy status; the association with HIV status approached significance. Figure 1 shows counts of sampled cases by clinical setting and whether treatment adhered to CDC guidelines; among those who received non-recommended treatment (shown in black), the largest proportions were treated at private physicians’ offices/HMOs (34.7% of those who received non-recommended treatment), family planning facilities (22.3%), and EDs/urgent care centers (12.8%). Smaller proportions of participants receiving non-recommended treatment were treated at STD clinics (7.3%), community/public health clinics (6.3%), military/VA facility (3.0%), HIV clinic (2.8%), correctional facility (2.6%), hospital (2.2%), or other facility (5.9%).
In multivariable analyses, region, pregnancy status, and HIV status were not found to be independently associated with receipt of non-recommended treatment and so were not included in the final model (P = 0.57). Results of the final model are shown in Figure 2. As compared with participants treated at an STD clinic and independent of age, race, or sexual orientation, statistically significant RRs for receiving non-recommended treatment were detected for the following settings: military/VA facility (RR, 6.9; 95% CI, 3.4–14.3), hospital (4.3; 1.7–10.9), correctional facility (4.2; 1.8–10.0), ED/urgent care (4.1; 2.2–7.6), private physician/HMO (3.1; 1.8–5.6), family planning facility (2.7; 1.5–4.9), community/public health clinics (2.0; 1.0–4.0), and other settings (3.1; 1.5–6.5). Men who have sex with men were less likely (RR, 0.70; 0.48–1.0) to receive non-recommended treatment than other men, although the RR only approached significance. In addition, blacks and Hispanics were both less likely (0.67 [0.49–0.92] and 0.72 [0.53–0.99], respectively) than whites to receive non-recommended treatment. Statistical associations with age were detected but were not monotonic.
We conducted sensitivity analyses to assess the robustness of these results with regard to lag (the period of time after the release of the 2010 guidelines during which we coded treatments as adherent if they adhered to either the former or the new guidelines). These results did not differ substantially when the lag was shortened to 30 days or lengthened to 90 days (see Fig. 1, Supplemental Digital Content 1, http://links.lww.com/OLQ/A83, illustrating the results of all sensitivity analyses). When regimens consisting solely of azithromycin 2 g were reclassified as nonrecommended, results were largely similar albeit attenuated, with the RRs for community/public health clinic, family planning facility, and hospital losing statistical significance but the RR for HIV clinic gaining significance. When participants without reported treatment or with chlamydia-only treatment were excluded, RRs were generally further from the null but followed the same general pattern as previously noted.
Among those receiving non-recommended treatment, the relative frequencies of different regimens were inspected both for the period before the release of the 2010 guidelines as well as the period beginning 60 days after the update and lasting through the end of 2011. The most common nonadherent regimen during the first period was azithromycin 1 g and/or doxycycline 100 mg for 7 or more days (44.4%), followed by no treatment (28.5%), and then treatment with a fluoroquinalone whether with or without noncephalosporin cotreatment (18.3%). During the second period, the most common nonadherent regimen was 250 mg or more of ceftriaxone without cotreatment (53.6%), followed by an inadequate dose of ceftriaxone whether with or without cotreatment with azithromycin or doxycycline (12.6%), and then azithromycin 1 g and/or doxycycline 100 mg for 7 or more days (10.7%).
Our analysis of documented treatment of uncomplicated NG diagnosed from 2009 to 2011 showed variability in compliance with the applicable CDC treatment guidelines across a variety of clinical settings and by demographic characteristics of patients. These results suggest a need for interventions aimed at increasing adherence to the guidelines in many health care settings and, furthermore, that barriers may be greatest at military/VA settings. This mirrors the results of a previous study, which found low compliance with recommendations in other domains of STD care (specifically reporting and secondary screening) at naval medical facilities, albeit in an entirely different part of the United States.20 However, we would be remiss not to note that, in these data, the absolute and relative differences in nonadherence suggest quite opposing conclusions: although military/VA and correctional facilities seem most likely to stray from the CDC’s NG treatment recommendations, they also report relatively few cases, which would very likely diminish the overall impact of interventions targeted at them. Among those who received nonrecommended treatment, more than a third were treated at private physicians’ offices or HMOs and about another third were treated at family planning facilities or ED/urgent care facilities. Thus, interventions in these settings may be most fruitful for disease control. Notably, according to a recent analysis, ED/urgent care settings were the slowest to adapt to revisions of national guidelines eliminating fluoroquinolones as recommended treatment.21 A major strength of this analysis was that it used data collected from a large, geographically representative sample of cases reported from a range of providers in nearly all regions of California, which not only covers a large area but also lies in a region from which antimicrobial resistance has historically spread to the rest of the United States.10
Although the overall rates of adherence seen here leave room for improvement, the higher proportion of proper treatment of NG seen among MSM and among blacks and Hispanics is encouraging because, in recent history, antimicrobial-resistant NG has first taken hold among MSM before gaining ground among heterosexuals,22–25 and because all these groups have born a disproportionate NG burden. However, the independent effects of sexual orientation and race/ethnicity on treatment seen here are surprising. First of all, although previous research has found racial disparities in other treatment domains, these disparities have often gone in the opposite direction.26,27 Second, on the assumption that any effects of these demographic variables on treatment would be largely mediated by clinical setting, the inclusion of setting in our model was expected to control away their effects on treatment. One possible explanation for this unexpected result is that these effects reflect differences in the demographic characteristics of the patients seen by specific providers within categories of clinical setting. Those providers who see more NG patients may be more likely be familiar with and adhere to the treatment guidelines28; so, given that the incidence of NG is relatively high among MSM, blacks, and Hispanics, members of these groups would tend to receive more appropriate NG treatment insofar as they tend to cluster at certain specific providers.
Monotherapy with ceftriaxone, as opposed to dual therapy, which would also include azithromycin or doxycycline, constituted more than half of all non-recommended treatment regimens provided following the release of the 2010 guidelines. Assuming this does not reflect a problem with data collection, this suggests that providers may not always realize that the additional gram of azithromycin specified in the treatment guidelines is not for codiagnoses with chlamydia, for which azithromycin 1 g is the recommended treatment, but rather additional treatment of NG above and beyond the ceftriaxone.5 In any case, the guidelines specify that all diagnosed cases of NG should be presumptively treated for coinfection with chlamydia, given their frequent co-occurrence.5 Thus, the prevalence of NG treatment with ceftriaxone 250 mg alone may reflect a lack of knowledge regarding the CDC STD treatment guidelines more generally than a misunderstanding of azithromycin’s role as co-treatment of NG.
This analysis had several limitations. Although the population of cases from which CGSS cases are sampled is unlikely to omit a substantial proportion of all diagnosed NG cases because cases are reported to CDPH through two independent mechanisms (by both laboratories and providers), not all sampled participants could be interviewed, nor could providers always be reached. Although the demographic characteristics of participants interviewed and not interviewed were similar, providers who are most likely to cooperate with CDPH in completing the CGSS interview may also be most likely to keep abreast of the treatment guidelines and make special efforts to comply. These findings then likely underestimate the proportion of NG cases that received non-recommended treatment. Furthermore, on the assumption that such providers may be differentially found at STD clinics, this would have most likely biased our RRs toward the null. Our results also likely overstate overall adherence because of our inability to assess compliance with the recommendation that patients treated with azithromycin 2 g alone receive a subsequent test of cure. If a recent analysis of adherence to repeat-testing recommendations for chlamydia is any guide,29 such repeat-testing rates may have been quite low. In addition, this analysis classified regimens as compliant or not on the basis of only known or suspected sites of infection, and a number of pharyngeal infections—for which the treatment guidelines are more stringent—may have been missed. It is also unfortunate that the data set combined private physicians with those practicing in HMOs because this precluded observing differences between them with regard to NG treatment patterns: adherence to treatment guidelines on the part of HMO providers may be substantially higher than for privately practicing physicians, given the institutional support structure in which the former group is enmeshed.30 Lastly, we can neither rule out inaccuracies in the treatment data, as it was not verified by chart review, nor in the classification of treatments as adherent or not, although inaccuracies of the latter sort are likely very minimal because classification was manually verified.
Future studies should aim at identifying the specific barriers that exist to providing guideline-compliant treatment and contribute to the variation in adherence found herein. In some settings, divergence from the treatment guidelines may stem largely from a lack of awareness of what they are, which may be exacerbated by their complexity and the frequency with which CDC has changed them. Insofar as this is true, future studies might also do well to identify effective means of informing providers of updates to the treatment guidelines. In other settings, structural barriers to the administration of the recommended treatments may play a role. This may be particularly true of ceftriaxone because it must be administered by injection, which poses more logistical challenges than oral regimens. Insofar as such structural barriers do, in fact, lie at the root of much of the nonadherence seen here, methods of overcoming them will need to be identified because ceftriaxone is the only remaining recommended first-line treatment of NG,6 and proper treatment is critical not only to halting continued transmission by those infected but also to slowing the emergence of continued resistance.
1. Centers for Disease Control and Prevention. Sexually Transmitted Disease Surveillance, 2010. Atlanta, GA US Dep. Heal. Hum. Serv 2011. Atlanta, GA: CDC; 2011.
2. Walker CK, Sweet RL. Gonorrhea infection in women: Prevalence, effects, screening, and management. Int J Womens. Health 2011; 3: 197–206.
3. Fleming DT, Wasserheit JN. From epidemiological synergy to public health policy and practice: The contribution of other sexually transmitted diseases to sexual transmission of HIV infection. Sex Transm Infect 1999; 75: 3–17.
4. Newman LM, Moran JS, Workowski K. Update on the management of gonorrhea in adults in the United States. Clin Infect Dis 2007; 44 (suppl 3): S84–S101.
5. Centers for Disease Control and Prevention. Sexually Transmitted Diseases Treatment Guidelines, 2010. Morb. Mortal. Wkly. Rep. 59th ed. Atlanta, GA: Office of Surveillance, Epidemiology, and Laboratory Services, Centers for Disease Control and Prevention (CDC), U.S. Department of Health and Human Services; 2010; 59.
6. Kirkcaldy RD. Morbidity and Mortality Weekly Report (MMWR) update to CDC’s sexually transmitted diseases treatment guidelines, 2010: Oral cephalosporins no longer a recommended treatment for gonococcal infections. Morb Mortal Wkly Rep 2012; 61: 590–594.
7. Asch SM, Sa, adah MG, Lopez R, et al. Comparing quality of care for sexually transmitted diseases in specialized and general clinics. Public Health Rep 2002; 117: 157–163.
8. St Lawrence JS, Montaño 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: 1784–1788.
9. Shekelle PG, Kosecoff J. Evaluating the treatment of sexually transmitted diseases at an urban public hospital outpatient clinic. Am J Public Health 1992; 82: 115–117.
10. Barry PM, Klausner JD. The use of cephalosporins for gonorrhea: The impending problem of resistance. Exper Opin Pharmacother 2009; 10: 555–577.
11. Samuel M, Mocello A. California Gonorrhea Surveillance System: Methodologic aspects and key results of a sample-based system. Public Health 2009; 124: 87–97.
12. California Department of Public Health SCB. Sexually Transmitted Diseases in California. 2011.
13. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria; 2012.
14. Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol 1989; 129: 125–137.
15. Lumley T. Analysis of complex survey samples. J Stat Softw 2004; 9: 1–19.
16. Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005; 162: 199–200.
17. McNutt L-A, Wu C, Xue X, et al. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 2003; 157: 940–943.
18. Zou G. A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004; 159: 702–706.
19. 19. Lumley T. Package “survey” [Internet]. [The Compr. R Arch. Netw. Web site]. Available at: cran.r-project.org/web/packages/survey/survey.pdf. Cited December 19, 2013.
20. Bond MM, Yates SW. Sexually transmitted disease screening and reporting practices in a military medical center. Mil Med 2000; 165: 470–472.
21. Dowell D, Tian LH, Ja S, et al. Changes in fluoroquinolone use for gonorrhea following publication of revised treatment guidelines. Am J Public Health 2012; 102: 148–155.
22. Forsyth S, Penney P, Rooney G. Cefixime-resistant Neisseria gonorrhoeae
in the UK: A time to reflect on practice and recommendations. Int J STD AIDS 2011; 22: 296–297.
23. Kirkcaldy RD. Cephalosporin susceptibility among Neisseria gonorrhoeae
isolates—United States, 2000–2010. Morb Mortal Wkly Rep 2011; 60: 873–877.
24. de Vries HJ, van der Helm JJ, Schim van der Loeff MFS, et al. Multidrug-resistant Neiserria gonorrhoeae with reduced ceforaxime susceptibility is increasingly common in men who have sex with men, Amsterdam, the Netherlands. Eurosurveillance 2009; 14: 1–6.
25. Goldstein E, Kirkcaldy RD, Reshef D, et al. Factors related to increasing prevalence of resistance to ciprofl oxacin and other antimicrobial drugs in Neisseria
. Emerg Infect Dis 2013; 18: 1290–1297.
26. Mehta JL, Bursac Z, Mehta P, et al. Racial disparities in prescriptions for cardioprotective drugs and cardiac outcomes in Veterans Affairs Hospitals. Am J Cardiol 2010; 105: 1019–1023.
27. Mills AM, Shofer FS, Boulis AK, et al. Racial disparity in analgesic treatment for ED patients with abdominal or back pain. Am J Emerg Med 2011; 29: 752–756.
28. Anschuetz G, Asbel L, Salmon ME, et al. Use of first-line treatment for Neisseria gonorrhoeae
after treatment guideline changes. Sex Transm Dis 2014; 41: 64–6.
29. Hoover KW, Tao G, Nye MB, et al. Suboptimal adherence to repeat testing recommendations for men and women with positive chlamydia tests in the United States, 2008–2010. Clin Infect Dis 2013; 56: 51–57.
30. Magid DJ, Stiffman M, Anderson LA, et al. Adherence to CDC STD guideline recommendations for the treatment of Chlamydia trachomatis
infection in two managed care organizations. Sex Transm Dis 2003; 30: 30–32.