Vaccination may be economically and epidemiologically advantageous over frequent screening for gonorrhea prevention

Abstract

Background:

Gonorrhea’s rapid development of antimicrobial resistance underscores the importance of new prevention modalities. Recent evidence suggests that a serogroup B meningococcal vaccine may be partially effective against gonococcal infection. However, the viability of vaccination and the role it should play in gonorrhea prevention is an open question.

Methods:

We modeled the transmission of gonorrhea over a 10-year period in a heterosexual population to find optimal patterns of year-over-year investment of a fixed budget in vaccination and screening programs. Each year, resources could be allocated to vaccinating people or enrolling them in a quarterly screening program. Stratifying by mode (vaccination vs screening), sex (male vs female), and enrollment venue (background screening vs symptomatic visit) we consider eight different ways of controlling gonorrhea. We then found the year-over-year pattern of investment among those eight controls that most reduced the incidence of gonorrhea under different assumptions. A compartmental transmission model was parameterized from existing literature in the U.S. context.

Results:

Vaccinating men with recent symptomatic infection, which selected for higher sexual activity, was optimal for population-level gonorrhea control. Given a prevention budget of $3 per capita, 9.5% of infections could be averted ($299 per infection averted), decreasing gonorrhea sequelae and associated antimicrobial use by similar percentages. These results were consistent across sensitivity analyses that increased the budget, prioritized incidence or prevalence reductions in women, or lowered screening costs. Under a scenario where only screening was implemented, just 5.5% of infections were averted.

Conclusions

A currently available vaccine, though only modestly effective, may be superior to frequent testing for population-level gonorrhea control.

Short Summary:

A modeling study compared vaccination to screening for gonorrhea prevention, given an annual budget constraint, and found that a modestly effective vaccine was optimal compared to frequent screening.

1. Introduction

Reported gonorrhea case rates have increased an alarming 118% in the U.S. since 2009 [1]. Left untreated, gonorrhea can lead to serious health outcomes including pelvic inflammatory disease (PID), ectopic pregnancy, infertility, and increased risk of HIV acquisition [2]. At the same time, gonorrhea’s ability to rapidly develop antimicrobial resistance has left clinicians with few options for treatment and generated concerns about our ability to effectively treat gonorrhea in the future [3]. Currently, the only recommended regimen is a single 500 mg injection of ceftriaxone [4]. The US Preventive Services Task Force currently recommends screening for gonorrhea in women under the age of 25 years, or older women at increased risk for infection [5], and quarterly screening for men who have sex with men (MSM) who are prescribed antiretroviral preexposure prophylaxis (PrEP) [6]. The CDC also recommends screening every 3–6 months for MSM living with HIV or at increased risk of infection [4].

This context has increased the urgency for developing a vaccine that is effective at preventing Neisseria gonorrhoeae (NG) infections [7]. Recent observational studies have suggested that an outer membrane vesicle (OMV) serogroup B meningococcal vaccine may also be protective against NG infection. A study using sexual health clinic data in New Zealand found the vaccine to be 31% effective against gonorrhea [8], and a study using clinical records in two cities in the U.S. found the complete vaccination series was 40% effective against gonorrhea [9]. Some recent studies have modelled the potential impact of a gonorrhea vaccine under various assumptions of efficacy and duration and found meaningful impacts on prevalence and incidence [ 10, 11, 12].

Though recent evidence suggests a meningococcal B vaccine may be effective at preventing NG infections in real world settings, it is unknown how it compares to other preventive measures in terms of costs and effectiveness. Public health departments generally have fixed budgets, which have tended to decline in recent years [13]. Dynamic optimal control methods offer a framework for comparing the viability of multiple interventions given these realistic economic constraints, and they can yield practical insights on how to best approach infectious disease prevention [14].This paper analyzes the viability of adding vaccination or an enhanced screening program for gonorrhea control alongside standard care. We describe conditions under which one intervention, or a combination of interventions, were optimal.

2. Methods

We applied a dynamic optimal control approach to identify how to optimally allocate a fixed annual budget between two gonorrhea interventions, programmatic screening and vaccination, to minimize infections [15]. Beyond simply modeling what happens when an intervention is introduced, our approach calculates the optimal combination from a set of possible interventions each year, allowing the mix of interventions to adjust to evolving epidemiologic dynamics. We adapted a previously published compartmental transmission model of gonorrheal infection in a heterosexual population using data-informed parameter inputs in a U.S. context [12]. Our main objective criteria minimized cumulative gonorrhea incidence over a ten-year period, subject to transmission dynamics and an annual budget constraint.

There were two interventions, vaccination and programmatic screening, available for gonorrhea prevention. Male or female (M/F) patients could be enrolled into either (or both) interventions during one of two clinic visit types: a routine background screening visit or a symptomatic, treatment-seeking visit to a clinic. This resulted in 8 total controls: combinations of vaccination or programmatic screening, male or female, and enrollment from background (non-symptomatic) or symptomatic visits. These 8 controls represent the levers a public health program could have at their disposal for gonorrhea prevention. Enrollment into the interventions was in addition to the standard care already occurring, i.e., background screening continued to occur at the same rate for those not enrolled in programmatic screening. Protection from vaccination was assumed to be 40% with a 1-year duration on average. We assumed vaccines were administered at the time of enrollment. Programmatic screening involved an average of four express screening visits per year in which the patient was tested for gonorrhea and treated if positive. A person remained enrolled in the screening program until they dropped out at a fixed rate. If a person developed a symptomatic infection while they were in screening program, they sought out treatment at the same rate as symptomatic infected persons (i.e., they do not wait till their next screening visit). To represent additional administrative limitations faced by public health decision-makers, expenditures across all 8 controls were bound by an annual budget and control allocations were determined annually. The solution to the optimization problem is known as a dynamic optimal control policy (DOCP), which specifies the optimal amount of each control in each year of the study.

2.1. Transmission Model

We modeled a susceptible-infectious-susceptible (SIS) heterosexual transmission system. There was no mortality induced by infection and the model-entry and -exit rates were constant for all states such that the population size was fixed and there were always exactly 50% men and 50% women. Individuals moved between high and low sexual activity states, where the activity level denoted the rate of sexual contact, and the high-activity proportion of the population was constant. A mixing parameter governed the frequency of contact between individuals of the same activity category. A fraction of incident infections were symptomatic, which caused screening and treatment at an elevated rate relative to people with an asymptomatic infection. Both susceptible and asymptomatically infected people received background screening at the same rate, though women at a higher rate than men, resulting in (successful) treatment for those who were infected. States in the model were categorized by infectious status (S/I), symptomatic or asymptomatic (S/A), sex (M/F), activity level (H/L), and intervention enrollment status (vaccinated (V), programmatic screening (X), neither (Z), both (XV)). We define our “primary” scenario as one that uses our preferred parameter values, presented in Table (1). Our primary scenario was parameterized such that the starting equilibrium prevalence was 1%. More details and model equations are presented in the supplemental digital content.

Table 1:

Primary model parameters.

Symbol	Description	Value (M/F)	Reference
γB	Background screening + treatment (per year)	0.15 / 0.3	[16]
γ I	Asymptomatic recovery (Natural clearance rate) (per year)	8 / 4	[16]
γST	Symptomatic treatment rate (per year)	22 / 11	[16]
γAT	Programmatic screening + treatment rate (per year)	4	Model choice
πA	Probability that infection is asymptomatic	0.41 / 0.68	[16]
ψ V	Protection due to vaccination (1=full protection)	0.4	[8, 9]
${\delta }_V$	Annual rate of vaccine protection loss	1	[8, 9]
${\delta }_X$	Annual rate of drop-out from screening	0.5	Model choice
${\pi }_H$	Proportion in the high sexual activity group	0.1	[12]
ω	Mean duration high risk [years]	5	[12]
${\gamma }_H={\omega }^{-1}$	Movement from low to high [year]		Derived
$\gamma L=\frac{1-\pi H}{\omega \pi H}$	Movement from high to low [year]		Derived
	— Auxiliary parameters —
$\tau$	Ratio of high to low contact rate	28.5	[12]
${\chi }_L$	Contacts, low activity	1	Model choice
$\psi$	Proportion assortative mixing	0	Model choice
${\beta }_M$	Transmission probability M to F	0.7	[12]
${\beta }_F$	Transmission probability F to M	0.42	[12]
$\alpha$	Overall entry rate (per year)	5,000	Model choice
$\mu$	Removal rate (per year)	20−1	Model choice
N	Total population	100,000	Model choice

M/F denotes values specific to males/females.

2.2. Budget Constraint

Costs associated with the interventions were constrained to be within an annual budget. Any costs directly related to activities from vaccination or programmatic screening beyond standard clinic operations were assigned to the program and labelled “internal” costs. The costs of vaccination included personnel costs from an initial and follow-up appointment and two OMV serogroup B meningococcal vaccine dose administrations. For simplicity, we assumed all vaccination costs were realized at the initial visit. For programmatic screening, there was an initial enrollment cost and then personnel, testing, and treatment (if infected) costs that accrued at each subsequent screening visit.

Personnel costs were estimated using 2022 nationwide non-facility prices from the CMS physician fee schedule (Table S2, Supplemental Digital Content) [17]. Physician fee levels were chosen to be consistent with recent estimates of STD clinic visit costs [18], but they do not necessarily reflect how a clinic would bill for each service type. Fees billed to Medicare may be significantly lower than when billing private insurance and may not fully account for administrative and other overhead costs. By anchoring the personnel costs to the CMS schedule, we could systematically reflect the intensity of a visit by changing the level. Costs for screening visits and vaccination visits were based on a level 3 established patient office visit (HCPCS code 99213), and symptomatic visits were based on level 4 new patient office visits (HCPCS code 99204) due to the need for more extensive patient evaluation and counseling. Enrollment costs were assumed to be the difference between levels 4 and 5 new patient visits. Gonorrhea test costs came from the 2022 national rate of the CMS laboratory fee schedule (HCPCS code 87591), and the cost for each (Bexsero) vaccine dose came from the 2022 CDC vaccine price list [19]. Cost estimates for a “low screening cost” scenario were derived from Earnest 2021 [20]. We also tracked external costs, defined as gonorrhea-related costs to the health system beyond those specific to the program. This includes the costs of background screening visits (costs equivalent to programmatic screening visits) and symptomatic visits, which assumed personnel costs equal to a new patient office visit (HCPCS code 99204).

2.3. Scenario Descriptions

We first ran single-control benchmarks, where the entire budget was allocated to only one of the 8 possible controls, to assess the relative effectiveness of each control alone and to benchmark comparisons to the optimal control scenarios that allow simultaneous co-investment in multiple intervention modes and enrollment modalities. The single-control benchmarks and “primary” optimal control scenarios used the parameters and costs listed in Table 1 and Table S2 (Supplemental Digital Content), respectively. Next, we computed dynamic optimal control policies (DOCPs) for scenarios with alternative parameter values, control characteristics, and objective criteria, as described in Table 2.

Table 2:

Alternative Scenarios.

Scenario	Modifications to the primary analysis
Alternative parameterizations
Low background screening	γB (M/F) changed to 0.1/0.2. Equilibrium prevalence was 1.19%
High background screening	γB (M/F) changed to 0.25/0.5. Equilibrium prevalence was 0.69%
High contact rate	χ (L/H) changed to 1.01/28.54. Equilibrium prevalence was 1.19%
Low contact rate	χ (L/H) changed to 0.98/27.69. Equilibrium prevalence was 0.69%
Positive assortative mixing	ψ changed from 0 to 0.04. Equilibrium prevalence was 1.19%
High prevalence	χ (L/H) changed to 2.5/50. Equilibrium prevalence was 10%
Alternative control characteristics
No vaccination	Vaccination controls turned off.
No vaccination, no screening drop-out	Vaccination controls turned off, δX was 0.
Low screening cost	Programmatic screening cost lowered from $127.14 to $73.40.
Short vaccine duration	δV increased from 1 to 2.
Alternative objective criteria
Minimizing incidence in women	Cumulative incidence in women minimized rather than incidence in both men and women
Minimizing prevalence in women	Cumulative prevalence in women minimized rather than cumulative incidence
Minimizing cumulative prevalence	Cumulative prevalence minimized rather than cumulative incidence
High budget	Budget increased from $3 to $10 per capita.

Beyond the modifications noted here, all other model parameters are the same as in Table 1.

For the alternative parameterizations, we first used evidence from the National Survey of Family Growth (NSFG) to produce low and high background screening rate scenarios. Next, we varied the contact rate parameter to produce low and high contact-rate scenarios that matched the equilibrium prevalence in the high and low background screening scenarios, respectively. Another scenario increased the proportion of assortative mixing to the point that equilibrium prevalence equaled that of the low background screening scenario. In terms of alternative control characteristics, we considered the absence of vaccination, no screening drop-out, reduced cost of screening, and decreased duration of vaccine protection. Finally, for alternative objective criteria, we minimized incidence in women rather than total incidence, minimized prevalence rather than incidence, and increased the budget from $3 to $10 per capita.

2.4. Optimization Methods

To compute the DOCPs, the direct multiple shooting method was used. We considered 10-year intervention periods each divided into one-year-long intervals. For each interval, the initial value problem was solved using a standard ODE solver. The continuity of the solution was ensured by introducing a matching penalty that, when minimized, eliminates any discontinuity in the state variables between intervals. The overall optimization problem was then solved by means of the numerical constrained minimization algorithm SQP (sequential quadratic programming) [21]. Like any non-convex optimization problem, the proper choice of the initial guess is essential for finding a good solution. To this end, the optimization routine was run 50 times with random initial guesses over a broad range of parameter values. The best 5 results were further used to generate a new set of local initial guesses. Finally, the best solution was retained.

2.5. Cost-Effectiveness

For each scenario, we calculated total number of cases averted, by sex, and the incremental cost-effectiveness ratios (ICERs) over 10- and 15-year periods, from both the program and health system perspectives. Within each scenario (defined by Tables 1 and 2), ICERs compared the costs and infections under the DOCP to the counterfactual costs and infections in the absence of any intervention. ICERS were defined as the difference in cost divided by the difference in infections (i.e., infections averted). The health system perspective subtracted external costs averted from the program expenditures when calculating the difference in cost. Courses of antibiotics administered were estimated from prevalence and screening/treatment rates.

3. Results

Vaccinating men with symptomatic infection was the most effective single-control benchmark, averting 9.5% of all infections (Figure 1). Female vaccination performed second best, averting 8.9% of infections (Figure 1; Table S3, Supplemental Digital Content). Screening prevented fewer infections than vaccination, but female screening outperformed male screening. Enrollment from symptomatic visits averted more infections than enrollment from background visits for both vaccination and screening.

Figure 1:

Percent of infections averted over 10 years, under each single-control benchmark using the primary parameterization. Under the counterfactual (absence of any control), there were 46,000 male infections, 39,300 female infections, and 85,300 total infections.

In the primary optimal control scenario, other than some investment in female vaccination in year 4, the DOCP was total investment in vaccination of men that were seeking treatment for symptomatic infection. Enrollment of patients from background visits was zero for both sexes and interventions, so those controls are not shown in Figure 2. Epidemiologically, the interventions prevented infection acquisition, and as a result, spared antibiotic use (Table 3). At equilibrium without either intervention operating, incidence was slightly higher in men than in women, but prevalence was much higher in women than in men (Figure S3, Supplemental Digital Content). In general, interventions in this system had a larger effect on incidence in men compared to women (Figure S1, Supplemental Digital Content). Infections averted per month increased asymptotically to approximately 50 male (i.e., 1 infection per 1,000 men per month) and 40 female monthly infections averted by year 10. Over ten years, the decrease in infections resulted in 2,900 courses of antibiotics spared (2.9 courses per 1,000 people per year). Under the primary DOCP, 8,100 infections were averted (4,500 in men and 3,600 in women; 8.1 infections per 1,000 people per year) compared to a counterfactual with no interventions (Table 3). With an annual budget of $3 per capita, only approximately 0.6% of the population had vaccine protection at any given time (Figure S2, Supplemental Digital Content). The 10-year cost per infection averted from the program perspective was $370 in the DOCP and decreased to $299 under a wider health system perspective.

Figure 2:

Proportion of budget dedicated to vaccination and screening programs, by sex, under selected optimal control scenarios. All vaccination and screening program enrollment occurred through symptomatic visits.

Table 3:

Infections averted and cost-effectiveness under selected optimal control scenarios.

Scenario	Sex	Counterfactual incidence	Percent of infections Averted	Antibiotic courses spared	ICER: program, 10 years	ICER: health system, 10 years	ICER: program, 15 years
Primary	Total	85,300	9.5%	2,900	370	299	286
	Male	46,000	9.8%
	Female	39,300	9.2%
High prevalence	Total	826,500	1.1%	3,500	326	251	303
	Male	456,500	1.2%
	Female	370,000	0.9%
Screening only	Total	85,300	5.5%	1,800	632	516	556
	Male	46,000	5.7%
	Female	39,300	5.3%
Screening only, no drop-out	Total	85,300	3.6%	1,200	943	799	925
	Male	46,000	3.9%
	Female	39,300	3.6%
Minimize incidence in women	Total	85,300	9.4%	2,900	375	304	288
	Male	46,000	9.6%
	Female	39,300	9.2%
Minimize prevalence	Total	85,300	9.5%	2,900	370	295	332
	Male	46,000	9.8%
	Female	39,300	9.2%
Lower screening cost	Total	85,300	10.0%	3,100	353	273	307
	Male	46,000	10.2%
	Female	39,300	9.7%
Shorter vaccine duration	Total	85,300	6.7%	2,100	526	446	459
	Male	46,000	7.0%
	Female	39,300	6.6%
High Budget	Total	85,300	29.0%	8,900	405	331	312
	Male	46,000	29.1%
	Female	39,300	28.5%	2,900

Incidence, infections averted, and antibiotic courses rounded to nearest 100 to avoid highlighting very small differences that could be due to numerical noise. Counterfactual incidence represents the number of infections over 10 years in the absence of any controls. Without controls, under the primary parameters, 29,900 courses of antibiotics were used. ICER: Incremental cost-effectiveness ratio, the additional cost per infection averted (difference in costs divided by difference in infections). Within a given scenario, ICERs compared the costs and infections under the DOCP to the counterfactual costs and infections in the absence of any intervention. A more detailed cost break-down is available in Table S4.

The allocation towards vaccination of men was robust to alternative parameterizations, control characteristics, and objective criteria, even when the objective decreased incidence or prevalence in women alone (Figure 2C, Table 3). When screening did occur, women were prioritized over men (Figure 2B,D–F). In the screening-only scenario, just 5.5% of infections were averted, compared to 9.5% in the primary scenario.

4. Discussion

The optimal allocation of a fixed budget between programmatic screening and vaccination was mostly comprised of vaccinating men presenting for care with urogenital symptoms. Despite having relatively low individual-level efficacy, male vaccination was at least as efficient as all other single-control benchmarks or optimal control policies for population level gonorrhea control. When screening did occur, screening in women was optimal compared to screening in men, which is consistent with current screening recommendations in heterosexuals. Given the budget in the primary analysis, vaccinating less than 1% of the population led to a relatively large 9.5% reduction in infections.

The objective of our primary analysis was to minimize total incidence, thus equally weighting infections in men and women. However, direct medical costs per infection are over 3 times higher in women than in men due to the development of sequelae, notably PID [22]. In addition, it is unclear whether PID pathogenesis is more related to the absolute number of infections or duration of time infected—i.e., whether preventing gonorrheal incidence or prevalence is more important for PID prevention. To explore the robustness of our primary analysis, we examined both possibilities through alternative objective criteria, where either incidence in women or prevalence in women were minimized (Figure 2C, Table 3). In these scenarios similar DOCPs were identified, demonstrating robust evidence that vaccination in men is the optimal control policy even when preventing PID is the primary goal.

Sex-specific parameter values in our model impacted the favorability of each sex as control targets for either vaccination or screening. Screening was sometimes identified as an efficient intervention in women, but never in men. Screening likely identified more infections in women than men because asymptomatic infections in women persisted nearly twice as long compared to men. Additionally, men were more likely to develop symptomatic infection, making their infections even less likely to be detected by screening. Another important implication of higher symptomatic infection in men is that symptoms, while obviously a marker for potential infection, are also a marker of having higher sexual activity. Because our model assumed episodic sexual behavior, having higher sexual activity was not a permanent state. Thus, presenting with symptomatic infection marked not only infections but also people currently with higher sexually activity. Effectively selecting for this group made enrollment though treatment-seeking in symptomatically infected people universally more efficient than enrollment through background screening.

To understand why vaccination was optimal compared to screening, it is helpful to consider how screening controls infection at the population level. Screening prevents transmission by diagnosing asymptomatic infections and, with treatment, reducing the duration of infection. However, the fast natural clearance of gonorrhea in men means that even quarterly screening was not frequent enough to diagnose many infections, thus leaving the average duration of infection with and without screening largely unchanged in men. Further, there was an economic trade-off between screening frequency and cost of participation in the screening program: reducing the programmatic screening frequency lowered the per person cost of enrollment in the screening program but further decreased its potential impact on interrupting population level transmission. We assumed that vaccination had an immediate effect upon enrollment, while programmatic screening didn’t potentially affect transmission dynamics until subsequent screening visits occurred.

We observed three types of DOCPs: transient (temporary investment in an alternative control), switching (the DOCP switches between two controls), and persistent (long-term co-investment in multiple controls). Generally speaking, transient DOCPs like that observed from the primary optimal control scenario, where there was a single-year investment in vaccinating women, occur when the marginal utility of two or more interventions are very close to one another and the DOCP method can find true-but-trivial trade-offs between those interventions. Outcomes between the primary DOCP and the single-control male vaccination benchmark were practically indistinguishable, with any differences likely attributable to numerical noise. In several scenarios, we see an apparent switch from one control to another as the end of the intervention period is reached. As the end of the intervention period approaches, the DOCP adopts an increasingly short-term view, leading to a preference for short-term effectiveness. For example, when incidence reduction in women was prioritized, the DOCP switches to vaccinating women in the final two years of the intervention period. This reduced female incidence at the cost of decreasing the total number of cases averted, but, because vaccinating women produces an effect faster than the more powerful but indirect effect from vaccinating men, it is optimal only at the end of the intervention period. Similar switching controls are seen in the Low Screening Cost and Short Vaccine Duration scenarios in Figure 2. In those scenarios, screening in women produces a short-term reduction in incidence by effectively shortening the period of asymptomatic infection. These tradeoffs occur because the benefit of vaccination persists in the population even after new vaccination stops. To test this idea, we ran an alternative version of the low screening cost scenario where interventions were optimized over 15 years instead of 10 (Figure S5, Supplemental Digital Content). The switch from vaccination to screening occurred approximately 5 years later when the intervention period was increased by 5 years, confirming that screening was only optimal when short-term gains were prioritized.

There may be other practical reasons why screening investment might be favored. There are conditions where decision-makers may want to be more short-sighted than in our primary intervention period of ten years. During a disease outbreak, reducing infections quickly might be of foremost priority. Alternatively, when dealing with a cohort that ages out of being at high-risk of infection, more short-sighted policies may be preferable towards the end of the risky period. Finally, in the high budget scenario, we see an example of a persistent optimal control policy where long-term co-investment in multiple intervention strategies and in different target sub-populations leads to a better public health outcome than simple investment in a single strategy.

Determining the economic viability of a program requires weighing costs of implementation against the value of infections averted. The cost-effectiveness of averting an infection depends on one’s perspective. A program may only care about its own direct expenditures related to gonorrhea control, while a decision-maker with a broader perspective might also consider reductions in health costs elsewhere (health system perspective). From the program’s perspective, the incremental cost per infection averted over the ten-year baseline period was $370. Thus, a program should be willing to implement the intervention if they regard one prevented NG infection to be worth at least $370. Ten-year health system perspective ICERs included cost savings from reduced infections more broadly and thus led to lower ICERs ($299 under the primary scenario). We also calculated cost per case averted ratios over a 15-year period, because there remain downstream benefits after the controls are turned off (after year 10) despite patients no longer being enrolled into interventions. The 15-year ratios demonstrate these benefits, having lower cost-per-case averted ratios than the 10-year ratios.

Our modeling approach attempted to reflect real-world decision-making constraints while maintaining mathematical tractability by making simplifying assumptions. We found robust quantitative optimal control policies over a range of sensitivity analyses; however, real-world implementation can differ in ways that we did not consider. We model the costs and effects of vaccination as occurring at the initial visit, even though there is a lag between the costs and benefits of receiving the second dose of the vaccine. Further, vaccination prices may change if the vaccine is approved for gonorrhea prevention. Likewise, we do not assume any costs of keeping patients enrolled in the screening program. Coming into the clinic quarterly for screening may be a large burden for patients, making it more difficult to enroll and maintain people in the program. On the other hand, in real-world setting it is more efficient to test patients for a range of STIs (including chlamydia, syphilis, and HIV) rather than just for gonorrhea. This type of screening program would have additional benefits, but modeling additional pathogens in our framework was not feasible. Similarly, the vaccine we model in this paper has additional unmeasured benefits through meningococcal B protection. Finally, we compare vaccination and screening in an isolated heterosexual population. There are a disproportionate number of NG infections among MSM [1], so this population is also important to consider. Optimal policies may differ in this case; for example, screening at extragenital sites may be more effective in MSM than urogenital screening in heterosexual men.

Our study presents compelling evidence that vaccination may be a beneficial alternative for gonorrhea prevention and control. Even a moderately effective and somewhat expensive vaccine was optimal compared to screening in a wide range of scenarios considered. Particularly in men, because the natural clearance rate of infection (1.5 months) was shorter than the time between screening visits, the population level benefits of programmatic screening were limited. At the same time, vaccination of highly sexually active men effectively reduces not only personal acquisition but also downstream transmission, which is critical for control. Our analysis demonstrates the potential multipurpose value of the OMV serogroup B meningococcal vaccine and may inform future vaccination and screening recommendations. Considering a long-run perspective, vaccination for gonorrhea prevention and control is likely a cost-effective intervention.