The modeled age categories were 15 to 24 years and 25 to 39 years, with the cut point chosen to align with the current recommendation of routine annual screening for N. gonorrhoeae in all sexually active females younger than 25 years.16 For each age-sex-subpopulation stratum, we assigned 10% of the population to a higher sexual activity group, characterized by elevated annual rates of partner acquisition relative to the rest of the subpopulation. Rates of partner change differed by subpopulation (see Appendix for details, http://links.lww.com/OLQ/A278). Sexual partnerships were formed within and across subpopulations. Mixing between MSM and heterosexual subpopulations was assumed to occur via MSM forming sexual partnerships with females.
Gonorrhea Natural History
Gonorrhea natural history in the absence of antimicrobial resistance was modeled using the approach of Garnett et al11 and is described in detail in the Appendix. Individuals with symptomatic infections were assumed to seek treatment, with a delay between infection onset and receipt of treatment. Asymptomatic cases could be identified and treated via screening. This would include individuals undergoing opportunistic screening and those seeking testing due to perceived risk (eg, partner with identified infection). We also modeled the reporting process, recognizing that not all treated (symptomatic and asymptomatic) cases will be captured in the surveillance data. After treatment or natural recovery, individuals returned to the susceptible state.
We calibrated parameters describing sexual mixing, gonorrhea natural history, and screening rates using an adaptive Metropolis-Hastings MCMC algorithm implemented in R.17 This method uses a Bayesian approach to estimate probability distributions for uncertain parameters, given the model and available data. The adaptive procedure optimizes the proposal distribution by first adapting the size of the covariance matrix to achieve an optimal acceptance rate, and then adapting the shape of the covariance matrix.18
Prior parameter distributions were guided by the available data, using point estimates and plausible ranges from the biomedical literature where possible, or expert opinion and assumption when estimates were unavailable (Tables 1 and 2). When information about parameters was scarce (eg, sexual mixing coefficients), we assumed broad priors.
The periods covered by the data sources used for calibration varied, but overall, the model described gonorrhea transmission between 2000 and 2015. Calibration targets were based on National Health and Nutrition Examination Survey (NHANES) prevalence data,19,20 national gonorrhea case reports,21 case characteristics as reported by the STD Surveillance Network,2 and National Survey of Family Growth sexual behavior data.15
Model Outputs and Analysis
To evaluate the impact of gonorrhea screening we compared our fitted model (“base case”) to a scenario using the same parameters, but with screening removed (ie, only symptomatic cases treated). We also compared the base case to 3 alternative screening scenarios: (i) full adherence to screening guidelines (“guidelines”),16 (ii) annual screening for all age and racial/ethnic groups (“universal”), or (iii) enhanced screening in groups with highest incidence in the base case (“enhanced”) (Table 3). All of these scenarios were applied retrospectively to the period 2000 to 2015.
Key model outputs included: total incident gonorrhea infections over the 16-year period, infections averted relative to the base case, actual and reported incidence rates, true and reported incidence rate ratios by race/ethnic group, proportion of male infections occurring in MSM, and number needed to screen to avert an infection. Further details are provided in the Appendix. We calculated mean values and 95% credible intervals (CrI) based on 1000 draws from the parameter posterior distributions, with intervention effect relative to the base case compared within each parameter set draw.
In the main analysis, we allowed for differential reporting of symptomatic cases by race/ethnicity and sex. We repeated the model calibration and analysis assuming a single reporting rate for all symptomatic cases.
The transmission model reproduced trends in reported cases over time and the observed disparities in the burden of gonorrhea in the US population by age, sex, and race/ethnicity (Fig. 2). It also fit well to other calibration targets (S1 Fig, http://links.lww.com/OLQ/A272).
Fitting the model to both prevalence and reported case data required a large proportion of unreported symptomatic cases in males (S2 Fig, http://links.lww.com/OLQ/A273). The risk of symptomatic reporting, relative to the reporting rate in asymptomatic cases, in nonblack males (0.09; 95% CrI, 0.04–0.15) was estimated to be lower than in black males (0.62; 95% CrI, 0.40–0.83). Overall, reporting for female symptomatic cases was higher, but a similar trend of higher relative risks of reporting for black females was seen (0.61; 95% CrI, 0.37–0.83 for nonblack females; 0.92; 95% CrI, 0.83–0.97 for black females). Estimates of asymptomatic screening and treatment rates were consistent with current screening guidelines,16 with lower rates in males and higher rates in MSM and females.
Estimated Gonorrhea Burden Between 2000 and 2015
We estimated that approximately 21 million (95% CrI, 1.7 × 107−2.6 × 107) incident gonorrhea cases occurred over the 16-year period, with a trend of stable or slightly increasing incidence in males and declining incidence in females (Fig. 3). For comparison, 4,931,200 cases were reported nationally over this period.21 Incidence was higher in males than females, with a mean male/female ratio of 1.9 (95% CrI, 1.3–2.7). Infection burden was concentrated in MSM, who comprised 67% (95% CrI, 56–76%) of total infections in males.
In 2015, gonorrhea incidence rate ratios were estimated to be 1.8 (95% CrI, 1.6–2.2) and 2.9 (95% CrI, 2.4–3.3) in black males and females, respectively, relative to rates in the overall population for a given sex (Fig. 3). Among black males, the rate ratio for incidence was smaller than the rate ratio for reported cases (3.9; 95% CrI, 3.4–4.5). By contrast, for black females, the rate ratio for reported cases was not significantly different from that for incidence.
Quantifying the Impact of Screening and Treatment
We compared our base-case model with a counterfactual scenario that assumed a complete absence of screening and treatment of asymptomatic infections. Model comparison suggested that 30% (95% CrI, 18–44%) of total infections were averted by the screening undertaken between 2000 and 2015, with a larger effect observed in females (40% of infections averted; 95% CrI, 27–54%) than males (23%; 95% CrI, 11–37%). However, in MSM, screening was estimated to have averted a negligible number of infections (mean of 0%; 95% CrI, −6% to 8%). In 2015, screening only modestly reduced disparities in incidence in the black population, with most of the effect concentrated in males (incidence rate ratio of 2.5 (95% CrI, 2.0–2.9) without screening and 1.8 (95% CrI, 1.6–2.2) with screening).
Impact of Alternative Screening Approaches
We compared our base-case model estimates to alternative screening approaches that might have been used between 2000 and 2015 (Fig. 4). We estimated that perfect adherence to guidelines (Guidelines) would have averted 51% (95% CrI, 23–75%) of gonorrhea infections. Uncertainty in this scenario was due in part to parameter combinations that markedly reduced transmission in MSM, who experience the most disease.
Annual screening for the entire sexually active population (Universal) was estimated to have a similar effect to following guidelines but was less effective for reducing infection burden in MSM, as screening frequency in this group was reduced compared with guideline recommendations. Enhanced screening (Enhanced) was estimated to most effectively reduce both overall infection burden and racial disparities in incidence (Fig. 5). Although transmission was significantly reduced, gonorrhea persisted in most simulations, primarily in MSM.
Adherence to screening guidelines (Guidelines) was identified as the most efficient strategy for averting gonorrhea infections, whereas universal screening was the least efficient, requiring over 5 times as many screening tests as with the guidelines to avert a single infection (Fig. 6).
Sensitivity of Model Results to the Assumption of Differential Reporting
Repeating calibration without differential reporting of symptomatic cases by race/ethnicity or sex resulted in a low estimated risk of reporting of symptomatic cases relative to asymptomatic cases (0.27; 95% CrI, 0.16–0.41) (S3 Fig and S4 Fig, http://links.lww.com/OLQ/A274, http://links.lww.com/OLQ/A275). The relative burden of infection in MSM was estimated to be lower than that in the main analysis, and consequently, there was a smaller difference in incidence between males and females (S5 Fig, http://links.lww.com/OLQ/A276). Screening was estimated to have been more impactful, averting 38% (95% CrI, 20–55%) of incident cases over the period, relative to no screening.
With the exception of the guidelines strategy, all alternate screening approaches were expected to be as, or more, effective than was estimated in the main analysis (S6 Fig, http://links.lww.com/OLQ/A277). The enhanced screening approach dramatically reduced the total number of new infections to ~7% of what was estimated with screening at base-case rates. By contrast, adherence to guidelines was as, or less, effective for reducing incidence, compared with our findings with differential reporting.
Without differential reporting, racial/ethnic disparities in males were larger than estimated in the main analysis and estimates of disparities using reported cases were more reflective of true underlying differences in incidence in males (S6 Fig, http://links.lww.com/OLQ/A277). Despite the changes in the estimated impact of the different screening approaches on various population groups without differential reporting, the relative efficiencies of the strategies did not change, with adherence to guidelines remaining the most efficient approach for reducing overall gonorrhea incidence in the population (S6 Fig, http://links.lww.com/OLQ/A277).
We developed a dynamic mathematical model that describes gonorrhea transmission in the United States, including observed disparities in infection burden. This novel model provides a platform for estimating the burden of both nonreported and reported gonorrhea infections, as well as the impact of current and alternative screening approaches, in a way that captures population characteristics relevant for prevention.
We demonstrated that screening likely reduced gonorrhea incidence in the population over the years 2000 to 2015, although this impact was primarily seen in heterosexual men and women. Alternate screening strategies could have further reduced gonorrhea burden in the population, but with differential impacts on the subpopulations of interest. In particular, adhering to guidelines had the potential to reduce gonorrhea transmission in MSM, whereas universal and enhanced screening approaches were more effective at reducing racial/ethnic disparities. As we compared gonorrhea trends under our best estimates of actual screening coverage to what might have been achieved in an ideal world, it is notable that even the most intensive screening strategy was not expected to eliminate gonorrhea transmission. This is an important finding that contrasts with previously published models.10,11 Our results suggest that a shift in programmatic focus from elimination to other outcomes (eg, improved case management, reduction in repeat infections, or reducing disparities) may be appropriate.22
Although our aim was to estimate gonorrhea burden and disparities, and identify programmatic alternatives, the exercise of model calibration to available data also provided important insights into heterogeneities in both gonorrhea epidemiology and surveillance in the United States. The calibrated model estimates a low probability of reporting of symptomatic infections that varies by race/ethnicity and gender. This low level of reporting may appear inconsistent with improvements in electronic and automated reporting systems that have been implemented in many jurisdictions.23,24 Reasons for this finding could include individuals not seeking treatment, self-treatment25 or cases receiving presumptive treatment without laboratory testing. Our calibration results suggested a higher reporting probability in black Americans than the rest of the population, which could reflect differential health care access and utilization patterns by race/ethnicity,4,26–29 which in turn may impact reporting.8 The implication of differential case reporting by race/ethnicity is that underlying relative disparities in gonorrhea incidence may be exaggerated in the reported case data. When we repeated our model calibration without allowing for differential reporting of symptomatic cases, we did observe some differences in the estimated impact of different screening strategies, with the major divergence relating to the estimated impact of adhering to current screening guidelines. In the absence of differential reporting, adherence to guidelines was estimated to enhance relative disparities but possibly reduce absolute disparities in gonorrhea burden.
Like any mathematical model, ours has limitations. We had nonoverlapping time series for several of the data sources used for model fitting and limited data on changes in screening and reporting over time. The model also had a large number of parameters, some of which were informed by limited data or relied on expert opinion or assumption. Although the calibration approach allowed us to account for the uncertainty associated with our data sources and input parameters and to propagate that uncertainty in our model estimates, issues of parameter identifiability were a concern. However, we did not use the model fitting process to attempt to infer the true values of individual parameters; rather we used this approach to identify combinations of parameters that reproduced trends in the data. Our model was able to replicate the surveillance data with reasonable fidelity, but did not capture the increase in male cases aged 25 to 39 years that has been observed since 2011.1 Given this limitation, and the overall challenges associated with fitting the model to multiple data sources, we focused our analysis on the period between 2000 and 2015, rather than forecasting future trends. Of necessity our model included a number of simplifying assumptions: we did not model different anatomical sites of infection or screening. Our approach to modeling sexual behavior resulted in varying levels of partner change in different subpopulations, such that mapping of modeled interventions onto current screening guidelines was an approximation. An alternate modeling approach, such as an agent-based model, would be better able to apply screening to individuals with specific risk factors, at the expense of added model complexity. Our model does not account for emergence of antimicrobial-resistant gonorrhea,30 which could attenuate the impact of any screening program. Because this was a retrospective analysis, we reasoned that the impact of resistant strains on transmission was likely to be relatively minor. Nonetheless, our finding that screening has the potential to reduce gonorrhea incidence must be interpreted with caution, because modeling studies have demonstrated that increased treatment of gonorrhea has the potential to increase the spread of resistance (31 s, 32 s). Given that screening for asymptomatic cases may play a key role for limiting the spread of antimicrobial resistant strains, the inclusion of resistance in models projecting future trends will be critical.
Using a mathematical model calibrated to multiple data sources, we have shown that screening has likely reduced the gonorrhea burden in the US population and can be used strategically to further control infection spread. It is important to note that we find that gonorrhea was likely to have persisted in United States over the period modeled, regardless of the screening strategy used.
For further references, please see “Supplemental References,” http://links.lww.com/OLQ/A279.
1. Centers for Disease Control and Prevention. Sexually transmitted disease surveillance 2015. Accessed 21 Feb 2017: http://www.cdc.gov/std/stats
2. Newman LM, Dowell D, Bernstein K, et al. A tale of two gonorrhea epidemics: Results from the STD surveillance network. Public Health Rep 2012; 127:282–292.
3. National Center for Health Statistics. Vintage 2014 postcensal estimates of the resident population of the United States (April 1, 2010, Jul 1 2010-July 1, 2014), by year, county, single-year of age (0, 1, 2, 85 years and over), bridged race, Hispanic origin, and sex Prepared under a collaborative arrangement with the US Census Bureau. Available from: http://www.cdc.gov/nchs/nvss/bridged_race.htm
as of June 30, 2015, following release by the U.S. Census Bureau of the unbridged Vintage 2014 postcensal estimates by 5-year age group on June 25, 2015.
4. Hogben M, Leichliter JS. Social determinants and sexually transmitted disease disparities. Sex Transm Dis 2008; 35(12 Suppl):S13–S18.
5. Aral SO, Adimora AA, Fenton KA. Understanding and responding to disparities in HIV and other sexually transmitted infections in African Americans. Lancet 2008; 372(9635):337–340.
6. Hamilton DT, Morris M. The racial disparities in STI in the U.S.: Concurrency, STI prevalence, and heterogeneity in partner selection. Epidemics 2015; 11:56–61.
7. Laumann EO, Youm Y. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: A network explanation. Sex Transm Dis 1999; 26:250–261.
8. Miller WC. Epidemiology of chlamydial infection: Are we losing ground? Sex Transm Infect 2008; 84:82–86.
9. Mayberry RM, Mili F, Ofili E. Racial and ethnic differences in access to medical care. Med Care Res Rev 2000; 57(Suppl 1):108–145.
10. Hethcote HW, Yorke JA. Gonorrhea transmission dynamics and control. New York: Springer-Verlag, 1984.
11. Garnett GP, Mertz KJ, Finelli L, et al. The transmission dynamics of gonorrhoea: Modelling the reported behaviour of infected patients from Newark, New Jersey. Philos T R Soc B 1999; 354:787–797.
12. Turner KME. Investigating ethnic inequalities in the incidence of sexually transmitted infections: mathematical modelling study. Sex Transm Infect 2004; 80:379–385.
13. Chen MI, Ghani AC. Edmunds WJ. A metapopulation modelling framework for gonorrhoea and other sexually transmitted infections in heterosexual populations. J R Soc Interface 2009; 6:775–791.
14. Chen MI, Ghani AC. Populations and partnerships: Insights from metapopulation and pair models into the epidemiology of gonorrhoea and other sexually transmitted infections. Sex Transm Infect 2010; 86:433–439.
16. Workowski KA, Bolan GA, Centers for Disease Control and Prevention. Sexually transmitted diseases treatment guidelines, 2015. MMWR Recomm Rep 2015; 64(RR-03):1–137.
18. Camacho A, Kucharski A, Aki-Sawyerr Y, et al. Temporal changes in Ebola transmission in Sierra Leone and implications for control requirements: A real-time modelling Study. PLoS Curr 2015; 7.
19. Satterwhite CL, Torrone E, Meites E, et al. Sexually transmitted infections among US women and men: Prevalence and incidence estimates 2008. Sex Transm Dis 2013; 40:187.
20. Torrone EA, Johnson RE, Tian LH, et al. Prevalence of Neisseria gonorrhoeae
among persons 14 to 39 years of age, United States 1999 to 2008. Sex Transm Dis 2013; 40:202–205.
22. Unemo M, Bradshaw CS, Hocking JS, et al. Sexually transmitted infections: Challenges ahead. Lancet Infect Dis 2017; 17:e235–e279.
23. Centers for Disease Control and Prevention. Progress in increasing electronic reporting of laboratory results to public health agencies—United States 2013. MMWR Morb Mortal Wkly Rep 2013; 62:797–799.
24. Samoff E, Fangman MT, Fleischauer AT, et al. Improvements in timeliness resulting from implementation of electronic laboratory reporting and an electronic disease surveillance system. Public Health Rep 2013; 128:393–398.
25. Gordon SM, Mosure DJ, Lewis J, et al. Prevalence of self-medication with antibiotics among patients attending a clinic for treatment of sexually transmitted diseases. Clin Infect Dis 1993; 17:462–465.
26. Pathela P, Klingler EJ, Guerry SL, et al. Sexually transmitted infection clinics as safety net providers: Exploring the role of categorical sexually transmitted infection clinics in an era of health care reform. Sex Transm Dis 2015; 42:286–293.
27. Rice RJ, Roberts PL, Handsfield HH, et al. Sociodemographic distribution of gonorrhea incidence: Implications for prevention and behavioral research. Am J Public Health 1991; 81:1252–1258.
28. Bonney LE, Cooper HL, Caliendo AM, et al. Access to health services and sexually transmitted infections in a cohort of relocating African American public housing residents: An association between travel time and infection. Sex Transm Dis 2012; 39:116–121.
29. Farley TA. Sexually transmitted diseases in the Southeastern United States: Location, race and social context.Sex Transm Dis 2006; 33(7 Suppl):S58–64.
30. Bolan GA, Sparling PF, Wasserheit JN. The emerging threat of untreatable gonococcal infection. N Engl J Med 2012; 366:485–487.
Supplemental Digital Content
© Copyright 2018 American Sexually Transmitted Diseases Association