Nevin, Remington L. MD, MPH*; Shuping, Eric E. MD, MPH†; Frick, Kevin D. MA, PhD‡; Gaydos, Joel C. MD, MPH§; Gaydos, Charlotte A. MPH, DrPH∥
GENITAL CHLAMYDIA TRACHOMATIS IS THE most common sexually transmitted infection in the United States, with 976,445 cases reported to the Centers for Disease Control and Prevention in 2005, corresponding to a rate of 332.5 cases per 100,000 population.1 Reported rates underestimate both the incidence and prevalence of infection because many women and men with Chlamydia are asymptomatic and do not seek healthcare unless notified of infection through screening.
Rates of Chlamydia prevalence in the United States vary significantly between demographic groups. A recent study suggests 2.2% of those aged 14 to 39 are positive for Chlamydia,2 though significantly higher prevalences were found among certain economic and behavioral subgroups, including those without a high school education, those of low socioeconomic class, and those with an early age of sexual debut. Previous studies of Chlamydia among both male and female US military recruits reflect the prevalences of these higher risk groups. One study reported a Chlamydia prevalence of 9.5% in a cross sectional study of 23,010 asymptomatic female Army recruits between 1996 and 1999.3 Using a similar study design another study found a prevalence of 5.3% among 2245 male Army recruits between 1998 and 1999.4
A high economic burden is associated with Chlamydia infection because of high rates of related sequelae arising from asymptomatic infection, including pelvic inflammatory disease (PID) and chronic pelvic pain (CP). Annual direct medical expenditures for PID and CP have been estimated to exceed $1.06 billion and $188 million, respectively.5 These high costs have prompted many professional societies and government agencies to recommend population-based female Chlamydia screening.6–8 Because of the high direct healthcare costs associated with treatment of the sequelae of asymptomatic Chlamydia (AC) infection, universal population-based screening of female Army recruits has been shown to be cost-effective from a healthcare payer perspective.9
Males comprise the majority of US military recruits,10 and represent an ideal population in which to achieve identification and interruption of sexually transmitted infection. With advances in and wide availability of nucleic acid amplification tests (NAAT),11 male recruits can now be accurately and rapidly screened with a noninvasive urine test. Using static models, investigators have determined that screening of males using NAAT is cost-effective in specific settings such as detention centers.12,13 Despite a number of studies employing static modeling and demonstrating cost-effectiveness of male screening in opportunistic settings,14 the United States Preventive Service Task Force has concluded that the evidence is currently insufficient (category I) to recommend for or against screening males in the general population for AC infection.7,8
The objective of this study was to determine whether population-based screening of male Army recruits would be either cost-saving or cost-effective by permitting the subsequent notification and treatment of infected female partners who would otherwise not present for care. Screening of males would be cost-saving if the discounted savings from averted long-term costs of untreated female infection exceeded the sum of the costs of male screening and the short-term costs of increased female healthcare usage. Screening of males would be cost-effective if the total healthcare payer costs per case of CP or PID averted were comparable to those of other feasible interventions. In our study, we chose to limit analysis to the costs associated with chronic pain and PID only, and to defer an analysis of averted male costs and costs associated with other potential female complications.
Although the limitations of static models in modeling population outcomes have been well described,14–16 in specific settings where a newly segregated cohort experiences an intervention in tandem, static modeling may be used to isolate the effects of the cohort amidst dynamic population processes. The screening of male recruits, with the potential early treatment of infected female partners through subsequent notification, represents one such scenario.
In our study, we have built upon the efforts of previous investigators4,12,15,17–30 in determining the prevalence, natural history, costs, and outcomes of Chlamydia infection (including progression to PID and CP), and the effectiveness and outcomes of notification strategies, to model the expected healthcare payer cost savings and reductions in select female adverse outcomes obtainable through screening of male recruits.
We developed a static decision tree incorporating Markov processes using TreeAge Pro Healthcare software (TreeAge Pro 1.6, Williamstown, MA, 2005) to model the expected differences in direct healthcare costs and cases of PID and CP averted among partners of male recruits through the implementation of recruit screening policies incorporating attempted female partner notification.
Policy options modeled included the status quo involving no male screening, a policy of universal male recruit screening, and a policy of selective screening of males aged 24 years and younger. Screening was assumed to occur in tandem among a population of males separated from their female partners and between whom no further disease transmission would occur. The natural history of Chlamydia infection in females was assumed to unfold following male screening with no subsequent opportunities for infection during the modeled period. Perfect sensitivity and specificity of the screening test was not assumed, permitting false positive and false negative screening results. The number of female partners per male was modeled to vary according to infection status. Uninfected males were assumed to have no infected female partners, and the population of female partners was assumed to be large enough such that no 2 males shared a female partner (Fig. 1).
Among female partners of infected males, health states were modeled using Markov processes with a period length of 6 months. Modeled health states included AC, symptomatic Chlamydia, silent PID (SP), overt PID (OP), prechronic pain, CP, and uninfected and asymptomatic (UA). Transitions permitted in the Markov model reflected the natural history of Chlamydia infection in females resulting from preexisting infection or transmitted from the infected male partner. Health states CP and UA were accumulator states that permitted no transitions (Fig. 2).
Distinct health state transition probabilities were defined for both notified and unnotified female partners of infected males. Notification was assumed to occur (or not occur) immediately after testing. Health state transition probabilities varied between notified and unnotified female partners only in the asymptomatic health state of AC in which notification of potential infection could lead to a modification of the natural history of infection or the accumulation of direct healthcare costs. Among unnotified partners, community screening for Chlamydia infection was presumed to occur, and the results of screening were permitted to create a distinct set of health state transition probabilities within each relevant period. Additionally, the model permitted partner notification to affect health state transition probabilities only during the first 2 Markov periods. Partners notified of potential infection but not pursuing immediate follow-up were presumed to have health state transition probabilities independent of community screening. Root probabilities were calculated by assuming sexual contact occurred 6 months before screening and modeling a single period transition from an initial uninfected or asymptomatic health state.
A healthcare payer perspective was used that included both the direct costs of male screening and relevant direct female healthcare costs. A 10-year time horizon with future cost discounting at 3% per annum was employed. Indirect costs and nonhealthcare direct costs were not modeled. Specific costs included in the model were the direct costs of male screening, the direct healthcare costs arising through Chlamydia-related healthcare usage among female partners, and the costs of healthcare from treatment of Chlamydia-related PID and CP. Costs of notification were assumed to be minimal and were not modeled, and no other direct costs were included.
Female costs were modeled at the individual level and multiplied by the average number of female partners per male to obtain total costs and outcomes per male recruit. Costs were tabulated so as to permit cost analysis and cost-effectiveness analysis if the least expensive strategy were not the most effective; therefore, only direct healthcare costs that potentially varied between policy options were included. Differences between outcome measures thus represented the expected direct cost savings (or expenditures) per male recruit obtained through implementation of screening.
As measures of effectiveness, the expected number of females ever experiencing PID, and the expected number of females in the CP state at the end of the 10 year period, per male recruit, were calculated. Incremental cost-effectiveness figures were calculated that reflected the cost per case of CP and case of PID averted when using a more costly screening strategy.
Parameter estimates for the model were obtained from available literature.4,12,15,17–30 Where significant uncertainty existed for a parameter, a range of plausible values were constructed from available sources (Table 1). Health state transition probabilities were defined by formulas incorporating multiple parameter estimates (Table 2). Given the significant uncertainties in parameters defining health state transition probabilities, the Markov model defining the natural history of Chlamydia infection was used to calibrate the model by means of variation of key parameters. Parameter estimates which most accurately predicted the natural history of untreated female Chlamydia infection were selected for base-case analysis. The calibrated model simulated the natural clearance (Fig. 3) within 4% of published estimates17 at 1 and 2 years following infection. Additionally, progression of untreated Chlamydia to PID towards an asymptotic prevalence of 34% was consistent with published estimates.31,32
Cost-effectiveness analyses were performed by inverting incremental effectiveness measures with a minimum effectiveness of zero. Threshold analyses for PID cost-effectiveness were performed on all parameters under an assumption of high nonlinearity. To assess sensitivity of the base-case to changes in parameter estimates, analyses were repeated under 3 model scenarios: a high cost scenario, in which all cost parameters were maximized; a slow progression scenario, in which parameters favoring a slow progression of Chlamydia to PID and CP were maximized or minimized as appropriate; and an effective notification and screening scenario, in which the likelihood of successful notification and screening were maximized.
Probabilistic sensitivity analysis was performed on all parameters sampling from triangular distributions over the range with the base-case estimate as the most likely value. The range of expected values for direct costs, expected cases of PID and CP, and incremental cost-effectiveness ratios (ICERs) were calculated from 10,000 runs of second-order simulation sampling all parameters simultaneously.
A policy of no male screening resulted in the lowest expected total healthcare payer direct costs per male recruit, and this cost was used as a reference for incremental analysis. Including the costs of male screening, selective male screening of recruits aged 24 and younger resulted in an additional $10.30 in total direct costs per male recruit. Universal screening of all male recruits added an additional $1.60 in total direct costs above selective screening or $11.90 above a policy of no screening. Examining only the direct costs of female healthcare, both screening policies reduced the total expected direct costs by $1.10 per recruit.
A policy of no male screening resulted in the highest number of expected cases of PID and CP among female contacts, and these values were used as a reference for incremental analysis. Selective screening prevented an additional 281 cases of PID and 140 cases of CP per 100,000 males screened above a policy of no male screening. Universal screening prevented 19 additional cases of PID and an additional 10 cases of CP per 100,000 males screened, or 299 and 150 cases of PID and CP, respectively per 100,000 males screened above a policy of no screening (Table 3).
In cost-effectiveness analysis, no policy option clearly dominated another. A policy of selective male screening yielded an ICER of $3.7K per case of PID averted, and $7.3K per case of CP averted. Universal screening yielded an ICER of $8.2K per additional case of PID and $16.4K per additional case of CP averted beyond selective screening.
Results of cost and cost-effectiveness analysis were not qualitatively sensitive to changes in parameter estimates, and no threshold values were detected over the plausible range of each parameter. Across all modeled scenarios, the ICERs were well within an order of magnitude of base-case estimates. A scenario of effective notification and screening maximized averted cases of PID and CP and minimized ICERs. Under this scenario, a policy of selective male screening yielded an ICER of $2.3K per case of PID averted, and $4.5K per case of CP averted. Universal screening yielded an ICER of $5.3K per additional case of PID and $10.5K per additional case of CP averted beyond selective screening.
Among other parameters, in one-way sensitivity analysis from the base-case scenario, increasing the 6-month probability of female community screening from 0.209 to 0.275 increased the ICER of both male screening policies. Under selective screening, the ICER increased to $4.6K per case of PID averted, and $9.2K per case of CP averted. Under universal screening, the ICER increased to $10.0K per case of PID and $19.9K per case of CP averted. To assess the sensitivity of our strategies to the rate of preexisting female Chlamydia infection, we increased the probability of infection to 0.50. Under selective screening, the ICER increased to $6.7K per case of PID averted, and $13.4K per case of CP averted. Under universal screening, the ICER increased to $13.5K per case of PID and $27.0K per case of CP averted.
Results of probabilistic sensitivity analysis were qualitatively consistent with base-case analysis (Fig. 4). Under selective screening, the mean ICER was $1.9K per case of PID averted, and $3.6K per case of CP averted (Table 4). Under universal screening, the ICER was $4.4K per case of PID and $8.2K per case of CP averted. The ranges incorporating the middle 95% of the expected values under probabilistic sensitivity analysis included outcomes saving money although averting more cases of PID under both selective and universal screening strategies.
Although a number of studies have examined the feasibility and cost-effectiveness of screening for Chlamydia in military cohorts,4,9,23,33 ours is the first study to use a calibrated Markov model to investigate the potential healthcare payer cost savings from averted female complications of Chlamydia infection through notification of screened male military partners. Examining only female direct costs, our analyses suggest that male Chlamydia screening programs lead to higher short-term healthcare costs among notified females, but that these costs are exceeded by the expected discounted long-term savings from later averted sequelae of untreated Chlamydia infection. Screening of male recruits is effective at preventing a significant number of adverse outcomes among female partners, and this benefit is achieved at reasonable incremental cost-effectiveness under both selective and universal screening policies. Our results suggest the principal direct costs of these policies are the costs of the male NAAT screening. With continued improvements in the convenience, availability and marginal costs of NAAT,11 the incremental cost-effectiveness of either policy will decrease.
Costs of female community screening were not modeled in our analysis because the incremental difference in expected healthcare payer costs of screening would not differ substantially across male screening policies. In sensitivity analysis, we found that increasing community screening in concert with an increased probability of notification decreased the ICER of male screening policies, but increasing community screening in isolation significantly increased the ICER. Increased detection of asymptomatic infection through female community screening would lead to fewer cases detected solely as a result of notification from male screening, hence increasing the relative cost-effectiveness of screening policies targeted solely at males. For this reason, programs that either increase opportunistic female community screening, such as in urban emergency departments,34 or that increase systematic female community screening in other healthcare settings22,31,35 may be a more cost-effective allocation of healthcare funding than programs targeted specifically at males. Research has demonstrated rates of screening in emergency departments are relatively low as compared with other healthcare settings.36 Others have argued for caution in increasing female screening at the expense of male screening, arguing that gender inequalities in screening carry hidden emotion costs.37
Our model has a number of important limitations. Our Markov model may misestimate the expected rate of transition from Chlamydia infection and PID to CP, thereby misestimating the potential costs, effectiveness, and cost-effectiveness of male recruit screening policies. In our model, significant uncertainty exists in the parameters used to model transitions between sequelae health states SP, OP, prechronic pain, and CP. Our Markov model was calibrated using data from one observational study, but other observational studies have noted fewer cases of PID developing than predicted by our calibration model.18 We noted moderate sensitivity of our effectiveness and cost-effectiveness measures to uncertainty in these parameters.
We did not model the health states of subsequent female partners of infected males, limiting our model to prior female partners at the time of screening. Our model incorporated assumptions about the behavior of these female partners, which permitted no subsequent opportunities for infection during the modeled period. These assumptions permitted static modeling. Although these assumptions are unrealistic, assuming both a decrease in the number of infections among subsequent female partners, and a nondifferential distribution of subsequent infections and adverse outcomes among prior female partners from unmodeled sources would likely preserve or improve the qualitative cost-savings and cost-effectiveness differences between screening policies consistent with the results of our sensitivity analysis. Dynamic modeling,14–16 although superior, is challenging to implement, and we believe static modeling will continue to play a role in addressing issues involving mass tandem screening.
Our findings are in consonance with previous studies employing static modeling. A study examining school-based Chlamydia screening among both males and female found an ICER of $1.5K per case of PID prevented38 above a policy of community screening only. Another study39 found that the universal population-based use of NAAT among males would result in 346 cases of PID per 100,000 males screened, at an ICER of $21.7K over a policy of no screening. Importantly, our methods differ substantially from those used in both models yet arrive at similar conclusions.
Male screening using NAAT tests is acceptable and accurate,11,40 and our study suggests that even modest rates of successful partner notification result in cost-effective identification and treatment of female contacts. Given our findings, consideration should be given to a policy of universal screening of male recruits for Chlamydia infection, linked to comprehensive educational programs.33
2. Datta SD, Sternberg M, Johnson RE, et al. Gonorrhea and chlamydia in the United States among persons 14 to 39 years of age, 1999 to 2002. Ann Intern Med 2007; 147:89–96.
3. Gaydos CA, Howell MR, Quinn TC, et al. Sustained high prevalence of Chlamydia trachomatis
infections in female army recruits. Sex Transm Dis 2003; 30:539–544.
4. Cecil JA, Howell MR, Tawes JJ, et al. Features of Chlamydia trachomatis
and Neisseria gonorrhoeae
infection in male army recruits. J Infect Dis 2001; 184:1216–1219.
5. Rein DB, Kassler WJ, Irwin KL, et al. Direct medical cost of pelvic inflammatory disease and its sequelae: Decreasing, but still substantial. Obstet Gynecol 2000; 95:397–402.
6. Horsburgh CR Jr, Douglas JM, LaForce FM. Preventive strategies in sexually transmitted diseases for the primary care physician. US Preventive Services Task Force. JAMA 1987; 258:814–821.
7. US Preventive Services Task Force. Screening for chlamydial infection: Recommendations and rationale. Am J Prev Med 2001; 20(3 suppl):90–94.
8. US Preventive Services Task Force. Screening for chlamydial infection: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2007; 147:128–134.
9. Howell MR, Gaydos JC, McKee KT Jr, et al. Control of Chlamydia trachomatis
infections in female army recruits: Cost-effective screening and treatment in training cohorts to prevent pelvic inflammatory disease. Sex Transm Dis 1999; 26:519–526.
10. Department of Defense. Population Representation in the Military Services: Fiscal Year 2002; 29th Annual Department of Defense (DoD) Report on Social Representation in the U.S. Military Services. Accessed August 26, 2007.
11. Gaydos CA. Nucleic acid amplification tests for gonorrhea and chlamydia: Practice and applications. Infect Dis Clin North Am 2005; 18:55–66.
12. Blake DR, Gaydos CA, Quinn TC. Cost-effectiveness analysis of screening adolescent males for chlamydia on admission to detention. Sex Transm Dis 2004; 31:85–95.
13. Gift TL, Lincoln T, Tuthill R, et al. A cost-effectiveness evaluation of a jail-based chlamydia screening program for men and its impact on their partners in the community. Sex Transm Dis 2006; 33(10 suppl):S103–S110.
14. Roberts TE, Robinson S, Barton P, et al. Screening for Chlamydia trachomatis
: A systematic review of the economic evaluations and modeling. Sex Transm Infect 2006; 82:193–200.
15. Welte R, Postma M, Leidl R, et al. Costs and effects of chlamydial screening: Dynamic versus static modeling. Sex Transm Dis 2005; 32:474–483.
16. de Vries R, van Bergen JE, de Jong-van den Berg, et al. Systematic screening for Chlamydia trachomatis
: Estimating cost-effectiveness using dynamic modeling and Dutch data. Value Health 2006; 9:1–11.
17. Molano M, Meijer CJ, Weiderpass E, et al. The natural course of Chlamydia trachomatis
infection in asymptomatic Columbian women: A 5-year follow-up study. J Infect Dis 2005; 191:907–916.
18. Morre SA, van den Brule AJ, Rozendaal L, et al. The natural course of asymptomatic Chlamydia trachomatis
infections: 45% clearance and no development of clinical PID after one-year follow-up. Int J STD AIDS 2002; 13(suppl 2):12–18.
19. Smith KJ, Ness RB, Roberts MS. Hospitalization for pelvic inflammatory disease: A cost-effectiveness analysis. Sex Transm Dis 2007; 34:108–112.
20. Hu D, Hook EW, Goldie SJ. The impact of natural history parameters on the cost-effectiveness of Chlamydia trachomatis
screening strategies. Sex Transm Dis 2006; 33:428–436.
21. Sutton MY, Sternberg M, Zaidi A, et al. Trends in pelvic inflammatory disease hospital discharges and ambulatory visits, United States, 1985–2001. Sex Transm Dis 2005; 32:778–784.
22. Hu D, Hook EW III, Goldie SJ. Screening for Chlamydia trachomatis
in women 15 to 29 years of age: A cost-effectiveness analysis. Ann Intern Med 2004; 141:501–513.
23. Sutton TL, Martinko T, Hale S, et al. Prevalence and high rate of asymptomatic infection of Chlamydia trachomatis
in male college Reserve Officer Training Corps cadets. Sex Transm Dis 2003; 30:901–904.
24. Zimmerman-Rogers H, Potterat JJ, Muth SQ, et al. Establishing efficient partner notification periods for patients with chlamydia. Sex Transm Dis 1999; 26:49–54.
25. Low N, McCarthy A, Roberts TE, et al. Partner notification of chlamydia infection in primary care: Randomised controlled trial and analysis of resource use. BMJ 2006; 332:14–19.
26. Tao G, Tian LH, Peterman TA. Estimating Chlamydia screening rates by using reported sexually transmitted disease tests for sexually active women aged 16 to 25 years in the United States. Sex Transm Dis 2007; 34:180–182.
27. Joesoef MR, Mosure DJ. Prevalence trends in chlamydial infections among young women entering the national job training program, 1998–2004. Sex Transm Dis 2006; 33:571–575.
28. Yeh JM, Hook EW, Goldie SJ. A refined estimate of the average lifetime cost of pelvic inflammatory disease. Sex Transm Dis 2003; 30:369–378.
29. Scholes D, Stergachis A, Heidrich FE, et al. Prevention of pelvic inflammatory disease by screening for cervical chlamydial infection. N Engl J Med 1996; 334:1362–1366.
30. Lau CY, Qureshi AK. Azithromycin versus doxycycline for genital chlamydial infections: A meta-analysis of randomized clinical trials. Sex Transm Dis 2002; 29:497–502.
31. Howell MR, Quinn TC, Gaydos CA. Screening for Chlamydia trachomatis
in asymptomatic women attending family planning clinics. A cost-effectiveness analysis of three strategies. Ann Intern Med 1998; 128:277–284.
32. Howell MR, Quinn TC, Brathwaite W, et al. Screening for Chlamydia trachomatis
in family planning clinics. Sex Transm Dis 1998; 25:274–280.
33. Arcari CM, Gaydos JC, Howell MR, et al. Feasibility and short-term impact of linked education and urine screening interventions for Chlamydia and gonorrhea in male army recruits. Sex Transm Dis 2004; 31:443–447.
34. Silva A, Glick NR, Lyss SB, et al. Implementing an HIV and sexually transmitted disease screening program in an emergency department. Ann Emerg Med 2007; 49:564–572.
35. Tao G, Abban BK, Gift TL, et al. Applying a mixed-integer program to model re-screening women who test positive for C. trachomatis
infection. Health Care Manag Sci 2004; 7:135–144.
36. Gift TL, Hogben M. Emergency department sexually transmitted disease and human immunodeficiency virus screening: Findings from a national survey. Acad Emerg Med 2006; 13:993–996.
37. Duncan B, Hart G. Sexuality and health: The hidden costs of screening for Chlamydia trachomatis
. BMJ 1999; 318:931–933.
38. Wang LY, Burstein GR, Cohen DA. An economic evaluation of a school-based sexually transmitted disease screening program. Sex Transm Dis 2002; 29:737–745.
39. Ginocchio RH, Veenstra DL, Connell FA, et al. The clinical and economic consequences of screening young men for genital chlamydial infection. Sex Transm Dis 2003; 30:99–106.
40. Marrazzo JM, Ellen JM, Kent C, et al. Acceptability of urine-based screening for Chlamydia trachomatis
to asymptomatic young men and their providers. Sex Transm Dis 2007; 34:147–153.
The following notes apply to the parameters listed in Table 1:
Note 1: (c_op × (1 − p_ip)) + (c_ip × p_ip).
Note 2: Among women aged 20 to 29 from Sutton et al.21
Note 3: Period costs discounted at 3% and assuming 10 year time horizon for lifetime costs from Smith et al.19
Note 4: (p_ctu25 × p_u25) + (p_cto25 × (1 − p_u25)).
Note 5: Among recruits aged 18 to 24 from Sutton et al.23
Note 6: Range ± 10% of point estimate percentage.
Note 7: Using parameters for 2 months before screening from Sutton et al.23
Note 8: Range ± 5% parameter estimate.
Note 9: Based on notification of 927 of 1744 local contacts in Zimmerman-Rogers et al.24
Note 10: Range obtained probabilities from 1 sex partner versus multiple sex partners in Tao et al.26; adjusted to a 6 month period.
Note 11: Point estimate from women aged 15 to 19 in Joesoef et al.27
Note 12: Upper limit of range is approximate from Scholes et al.29
Note 13: To calculate the proportion of nonclearing infections transitioning to PID, we used data from Scholes et al.29 that showed 9 of 44 treated infections transitioned to OP over 12 months. Assuming 19% spontaneous clearing in 12 months (Yeh et al.28) of the 36 treated cases that were nonspontaneously clearing, 9 (25%) transitioned to OP. Ranges calculated from binomial distribution were used in model calibration to select parameter estimate.
Note 14: Point estimate and ranges calculated using the formula 1 − √(1 − p_opidy).
Note 15: Point estimate calculated using the formula p_opid /p_io. Range calculated from range of point estimate for p_opid.
Note 16: This is also the proportion of uncured symptomatic cases remaining symptomatic.
Note 17: Calculated using the formula 1 − √(1 − p_cpy).