Curable sexually transmitted infections (STIs) are prevalent in the United States despite the availability of sensitive and noninvasive diagnostic screening tests.
Untreated Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (GC) infections can lead to pelvic inflammatory disease (PID) with sequelae of ectopic pregnancy, infertility, and chronic pelvic pain.1 Furthermore, prevalent CT, GC, and Trichomonas vaginalis (TV) infections all increase an individual’s susceptibility to HIV acquisition.2,3
Because most STI infections are asymptomatic, many infected individuals are not diagnosed and treated in a timely manner. To expand access to screening services, groups have evaluated the use of home sampling kits that individuals request via the Internet and submit to a laboratory via the US postal service.4,5 However, these evaluations did not collect longitudinal data to determine the impact on early detection and treatment of infection or include a standard care comparison group.
In 2012 to 2013, we conducted a demonstration project6 to determine the feasibility of a full-scale trial on home STI sample collection and e-prescriptions and to collect preliminary data on study outcomes. The demonstration project established the potential feasibility and effectiveness of the eSTI system. Using data from the demonstration project, we modeled the potential cost impact that may be seen in a future comparative effectiveness trial of eSTI versus referral to standard clinical care that includes e-prescriptions for treatment in the model.
MATERIALS AND METHODS
We constructed a decision tree (Supplemental Digital Content, http://links.lww.com/OLQ/A95) for a future comparative effectiveness research (CER) trial comparing 2 strategies for enrolling, testing, and treating women for CT, GC, and TV using nucleic acid amplification tests (NAATs): (1) eSTI with participants receiving a home collection kit for STI screening and an e-prescription for treatment versus (2) referral to standard clinic-based STI screening and treatment. The primary outcome was the number of STIs detected, and the secondary outcome was the number of STI tests completed.
In the planned comparative effectiveness trial, 2790 participants would be randomized to either the eSTI arm or the clinic referral arm. This number was chosen as the sample size required to show significant differences between arms given the findings of our demonstration project.6 To ensure inclusion of low literacy populations, participants would either enroll through the Internet or receive assistance enrolling from community health workers. Those in the clinic referral arm would be directed to a local clinic for STI testing and treatment. To ensure comparability of testing outcomes between the 2 arms, participants in the clinic referral arm would also self-collect specimens in the clinic setting using the same kit as those in the eSTI arm. Those in the eSTI arm would receive a home collection kit in the mail and would mail their sample to the laboratory for testing. They could access their test results from the eSTI online system or from community health workers by telephone and, if positive and asymptomatic, would have the option to have an electronic prescription sent to a local pharmacy or to receive a referral to a local clinic for treatment. Those with positive results who did not obtain their results within 3 days would be contacted by trial staff and assisted with obtaining treatment.
Our decision tree incorporated costs of screening, result notification, and treatment of positives. The analyses were conducted from 2 perspectives: the clinical trial perspective (CTP), which included only direct medical costs incurred by a clinical trial and the health care system perspective (HSP), which included only direct medical costs incurred by the health care system. Patient costs, such as travel time to the clinic, were not included. Costs were adjusted to 2013 US dollars using the medical care component of the consumer price index. Published literature was also used to estimate probabilities and costs for the referral to standard clinical care arm.
Probability estimates are provided in Table 1. We derived our participant testing rate estimates of 67% for the eSTI arm and 40% for the clinic referral arm as well as the ranges for sensitivity analyses from our primary data6 and the literature.5,7,8
Our STI prevalence estimate was derived from the Centers for Disease Control and Prevention chlamydia surveillance data from family planning clinics for the 3 regions that we plan to include in the future trial: 5.2% in California, 14.1% in Texas, and 8.3% in Maryland.9 Although the average prevalence in these 3 regions is 9.2%, we cannot be sure that an equal proportion of participants will be recruited from each region. There may be some dilution of STI prevalence if participants who enroll are at lower risk than the overall population from which they are drawn. Therefore, we conservatively estimated an 8% STI prevalence for the potential future trial. We chose to use CT prevalence as the primary outcome of interest, as this STI is much more prevalent than GC and regional prevalence data for TV are not readily available.
Estimates for the proportion of eSTI arm participants obtaining test results online or receiving results from staff were derived from our primary data. As was done in the demonstration study, if a participant randomized to eSTI does not retrieve her positive result online within 3 days, the staff would make 3 telephone calls and send 3 e-mails to her to connect her to care. Infected participants with CT or GC that are unable to be reached would be referred to Disease Intervention Specialists (DISs) at health departments for assistance with treatment. All 8 of the infected women identified in the demonstration project were treated without DIS assistance. Given the small sample, estimates for the proportion of women who are successfully notified and treated were derived from the literature,10 and we made the assumption that 10% would require DIS assistance.
It is assumed that all participants with positive results who are randomized to the clinic arm would receive their treatment in a clinic or through public health department DIS field–delivered therapy. However, those with positive CT and/or TV results who are randomized to eSTI would have the option of treatment in a clinic or eRx, which is an electronic prescription sent to their pharmacy of choice. For pregnant women or those with any symptom suggestive of PID in the eSTI arm, clinic treatment would be recommended. Women with GC would be required to receive treatment in clinic because current Centers for Disease Control and Prevention treatment guidelines recommend first-line therapy include ceftriaxone, which must be administered by injection.11 Based on preliminary data6 and weighting for expected prevalence of each infection (see Appendix http://links.lww.com/OLQ/A94), we estimated that 64.2% of women would choose eRx.
Participants who are assigned to the eSTI arm who have negative test results can log in to the Web site to retrieve their result or call the trial hotline. Staff would only assist with providing a negative result if the participant initiated a call to staff. We assumed that approximately 100% of those who enroll online and obtain their negative result would do so online, rather than choosing the telephone option. To be conservative, we estimated a lower bound of 75% for the sensitivity analysis range. In the demonstration study, 1 woman of 7 who enrolled with staff assistance and had a negative test result called staff for her result. The remaining 6 viewed their negative results online. Consequently, we estimated that 14% of those who enroll with staff in the eSTI arm would call staff to obtain their negative result.
We based our estimate for proportion of participants in the clinic arm who receive a negative result on the experience of a home testing program that did not offer online results, but attempted to contact all participants with results (negative as well as positive).4 In that study, the proportion of women who called in for results before study staff contacted them was approximately 60% (M. R. Barnes, personal communication, July 22, 2013). The range for our estimate was derived from the proportion in our study that enrolled with a staff person and called in for result (14%) and the proportion in our study who obtained their negative result via telephone call or online (91%).6
We did not incorporate sensitivity and specificity of the STI tests into the model because both arms will use the same highly sensitive (≥90%) and highly specific (≥99%)12 diagnostic test, but rather made the assumption that all infections would be detected if the test is completed. The primary goal of this decision analysis was to determine which strategy produces greater test completion and more STI diagnoses regardless of which diagnostic test is used.
All costs are provided in Table 2 as “per participant” costs in 2013 dollars. Estimates for the Web site server and maintenance costs were supplied by Zerolag and N-Tonic (D. Calebresi, personal communication, February, 8, 2013), the companies that provided these services for the demonstration study.6 Staff time required to enroll participants was derived from primary data and then multiplied by the Bureau of Labor Statistics hourly wage for a community health worker plus fringe benefits ($23.53).13,14
Kit costs and STI test processing costs for the eSTI arm are based on estimates supplied by the Johns Hopkins University (JHU), International STD, Respiratory, and Biothreat Research Laboratory, the laboratory that provided the kits and testing for the demonstration study. The JHU kit and STI test processing cost were also used as a conservative estimate for the clinic referral arm because these costs are lower than the Medicaid reimbursement rates for commercial NAATs (C. A. Gaydos and M. R. Barnes, personal communication, February 1, 2013).15 The upper limit for the range was based on the Medicaid reimbursement rate for HCPCS 87801: NAAT detection of multiple organisms ($96.49).15 The lower limit was based on the JHU kit and STI processing cost for CT and GC testing only ($55; C. A. Gaydos and M. R. Barnes, personal communication, February 1, 2013) plus the Medicaid reimbursement rate for HCPCS 87210: wet mount microscopic examination ($5.87)15 because most STD clinics still rely on saline microscopic detection for TV rather than a NAAT.
The cost of a clinic screening visit for HSP was estimated using Bureau of Labor Statistics hourly wage plus fringe benefits for a clinician (physician assistant or nurse practitioner) for a 30-minute visit.14,16,17 We used a lower range of 20 minutes and an upper range of 40 minutes for the visit. For CTP, we used the mean Medicaid E&M reimbursement rate for a level 2 initial visit ($59) in the 3 states, where the future comparative effectiveness trial is proposed. The Medicaid reimbursement rate was used as an estimate of what the trial would be expected to reimburse a participating clinic for providing the screening services. The upper end of the range was based on the E&M rate for a level 3 ($85) initial visit, and the lower end was based on the typical public STD clinic charge ($25), where the demonstration project was conducted (V. Levy, personal communication, July, 30, 2014).
Staff time required to provide participants with their positive test results (10 minutes) is based on data from one study participant and from personal experience of one of the authors (F.S.). We estimated that 3 unsuccessful telephone call attempts to notify an infected participant would require 5 minutes and that a telephone call to provide negative test results would require 2 minutes. Staff time to assure treatment of positives or refer to DISs was estimated at 15 minutes. We estimated that DIS involvement in treatment would require 1 hour.10
The cost of a clinic treatment visit for the HSP was estimated using clinician salary plus fringe benefits for a 30-minute visit if the participant was screened via eSTI and a 15-minute visit if the participant was screened in the clinic.14,16,17 We assumed that a return clinic visit for treatment would require less time than an initial visit for treatment. The medication costs to treat CT, TV, and GC were estimated using the wholesale acquisition costs,18 California formulary pricing,19 340B pricing, and the San Mateo Medical Center costs to purchase (G. Horne, personal communication, August 17, 2012). The costs from these 3 sources were averaged and weighted for expected prevalence of each infection (see Appendix http://links.lww.com/OLQ/A94). For the CTP, we incorporated only the costs of assisting participants with obtaining treatment, but not the cost of treatment itself because medication was self-pay in the demonstration project for participants who chose eRx and otherwise was provided through publicly funded clinics.
Analyses were conducted using TreeAge Pro decision analysis software (TreeAge Software, Williamstown, MA). Incremental costs and incremental cases of STIs detected were calculated using clinic referral STI screening as the comparator strategy. With few exceptions, incremental cost-effectiveness ratios were not calculated because the clinic strategy was either weakly or strongly dominated by the eSTI strategy for almost all cases. One-way and best-case/worst-case sensitivity analyses were conducted, using ranges presented in Tables 1 and 2, for parameter estimates that were less certain.
Baseline Cost Analysis
Table 3 displays results of the baseline cost analysis. The eSTI strategy detected 75 infections and the Clinic referral strategy detected 45 infections
Health Care System Perspective
The total cost of the eSTI strategy was $96,088 ($1281 per STI detected), and total cost of the clinic referral strategy was $71,668 ($1593 per STI detected).
Clinical Trial Perspective
The total cost of the eSTI strategy was $94,938 ($1266 per STI detected), and total cost of the clinic referral strategy was $87,367 ($1941 per STI detected. Results for the secondary outcome (STI tests completed) are included in Table 3.
One-Way Sensitivity Analyses
Table 4 displays results of the 1-way sensitivity analyses.
Proportion in Clinic Referral Strategy Visiting STD Clinic for Testing
For both HSP and CTP, at a clinic visit rate of 32%, the clinic strategy cost less than the eSTI strategy, but detected half as many infections. For HSP, at a clinic visit rate of 48%, the clinic strategy cost less but detected fewer infections than the eSTI strategy. However, for CTP at a clinic visit rate of 48%, the clinic strategy cost more and detected fewer infections than the eSTI strategy.
Proportion of Kits Returned in eSTI Strategy
For HSP at a kit return rate of 55%, eSTI costs more than the clinic strategy, but the cost/STI was lower. For CTP, eSTI costs less and detects more infections than the clinic strategy. At a kit return rate of 72%, for both perspectives, the eSTI strategy cost more than the clinic referral strategy, but the cost/STI was lower.
Throughout this range for both perspectives, the eSTI strategy cost less while detecting more infections than the clinic referral strategy. The higher the prevalence, the lower the cost per STD detected.
Clinic Office Visit Cost
Throughout the range of $19.50 to $39 for HSP, the clinic strategy cost less, but the cost/STI was lower for eSTI. For CTP, at an office visit cost of $25, the eSTI strategy cost more than the clinic strategy, whereas at the $85 visit cost, the clinic strategy cost more. Throughout the range, the cost/STI detected was lower for eSTI.
Clinic STI Test Cost
For HSP at a clinic STI test cost of $60.87, the eSTI strategy cost more than the clinic strategy, but detected more infections. However, the cost/STI detected was slightly lower for the clinic strategy with an incremental cost per effect ratio of $1356 for eSTI. At a test cost of $96.49, eSTI cost more than the clinic strategy but the cost/STI detected was lower for eSTI. For CTP, the clinic STI test cost would remain constant at $90, the same cost as for eSTI.
Best-Case/Worst-Case Sensitivity Analysis
The 5 parameters that have the most influence in 1-way sensitivity analyses (STI prevalence, kit return rate, clinic visit rate, office visit cost, and clinic STI test cost) were varied simultaneously along the ranges found in Tables 1 and 2. The eSTI best-case and worst-case cost-effectiveness results for both perspectives are shown in Table 3.
Our model suggests that an eSTI strategy of self-sampling with a home collection kit and e-prescription for treatment of uncomplicated infections would be likely to detect more STIs and cost less per STI detected than a standard clinic referral screening and treatment strategy.
We are aware of only 2 other cost analyses of home collection screening versus clinic screening.10,20 In one, clinician time and STI test costs for clinic screening were compared with test kit, packaging, and postage costs for home screening.20 Screening rates in the 2 strategies were assumed to be equal. Time required to notify and assist with treatment was not incorporated, nor was the cost of a clinic visit beyond the clinician’s time. The authors concluded that home screening could be cost-effective, but only if it resulted in less utilization of clinic services. Another study included not only the costs of testing and treatment but also the theoretical cost savings from prevention of PID.10 Rates of screening in the Internet sampling group were estimated to be higher than in the clinic group based on the authors’ primary data and the literature. The authors concluded that an Internet-based self-swab screening strategy led to more PID prevented at a lower cost than clinic screening. Although our study did not collect data on PID prevention, we were able to collect prospective data on the proportion of participants enrolling with staff assistance (and associated labor cost), proportion receiving positive and negative results online (and associated labor costs), and the proportion treated using eRx (and associated reduction in labor costs).
There are a few published studies that we can look to for estimates of clinic-based treatment rates for STIs that are detected via home screening and/or Internet methods.5,10,21 However, the availability of eRx in a Web-based system has not been evaluated, except in our demonstration study,6 and requires validation in a larger comparative effectiveness trial. If the overall study prevalence of STIs (CT + GC + TV) is greater than our estimate of 8%, cost per STI detected in a future trial may be lower than results of this cost analysis suggest.
There are limitations to our analysis. Results of cost-effectiveness analyses must be interpreted with caution, especially when parameter estimates that heavily influence the results are uncertain. The results of this analysis are strongly influenced by the home collection kit return rate estimate and the clinic visit rate estimate. We do not have data on the proportion of symptomatic versus asymptomatic participants who return kits and visit the clinic, and our estimates are limited because prospective randomized trials that directly compare these strategies have not been conducted. Furthermore, our estimates for trial staff labor costs are based primarily on investigator assumptions. Nevertheless, we have provided results of sensitivity analyses that were conducted to explore the influence of a range of estimates for these parameters. Lastly, this analysis is a static analysis in that it only considered the potential impact on the individual participants, but did not consider transmission. As part of a future scaled up comparative effectiveness study, we intend to include dynamic modeling to evaluate the cost-effectiveness of primary prevention for women who will not become infected due to decreased incidence in the population in addition to the secondary prevention of index women who are screened and treated.
A scaled CER trial will permit us to collect the following data for a more robust cost-effectiveness analysis: (1) rates of symptomatic participants in the clinic referral group that visit the clinic and in the eSTI group that return an STI kit; (2) rates of partner notification, partner treatment, and associated staff labor costs for each strategy; and (3) proportion of infections requiring DISs and associated labor costs for each strategy. Furthermore, we will be able to incorporate cases of PID and costs of PID sequelae prevented by each screening strategy as additional and important outcomes of a comprehensive cost-effectiveness analysis. If in the larger analysis the clinic referral strategy has a higher infection detection rate and/or treatment rate, then even if it costs more per STI detected, it may actually prevent more sequelae and lead to cost savings.
Our analysis indicates that eSTI is likely to be more effective and cost less per infection detected than clinic referral for STI screening in the context of a clinical trial as well as for clinical care. If confirmed, our findings would support the routine use of eSTI in clinical trials where longitudinal STI testing is required as well as the development of national scale-up and financing strategies for online STI testing and treatment programs in the context of routine clinical care.
1. Centers for Disease Control and Prevention. CDC Fact Sheet: Incidence, Prevalence, and Cost of Sexually Transmitted Infections in the United States. Atlanta, GA: Department of Health and Human Services, 2013.
2. Centers for Disease Control and Prevention. Sexually Transmitted Disease Surveillance 2011. Atlanta, GA: Department of Health and Human Services, 2012.
3. Kissinger P, Adamski A. Trichomoniasis and HIV interactions: A review. Sex Transm Infect 2013; 89: 426–433.
4. Gaydos CA, Barnes M, Aumakhan B, et al. Can E-technology through the Internet be used as a new tool to address the Chlamydia trachomatis
epidemic by home sampling and vaginal swabs?. Sex Transm Dis 2009; 36: 577–580.
5. Rotblatt H, Montoya JA, Plant A, et al. There’s no place like home: First-year use of the “I Know” home testing program for chlamydia and gonorrhea. Am J Public Health 2013; 103: 1376–1380.
6. Spielberg F, Levy V, Kapur I, et al. Online access to home STI specimen collection and e-prescriptions linked to public health—Is a comparative effectiveness trial feasible? International Society for STD Research; 2013; Vienna, Austria.
7. Graseck AS, Secura GM, Allsworth JE, et al. Home screening compared with clinic-based screening for sexually transmitted infections. Obstet Gynecol 2010; 115: 745–752.
8. Reagan MM, Xu H, Shih SL, et al. A randomized trial of home versus clinic-based sexually transmitted disease screening among men. Sex Transm Dis 2012; 39: 842–847.
9. Centers for Disease Control and Prevention. Chlamydia positivity among women aged 15-24 years tested in family planning clinics, by state, infertility prevention project, United States and outlying areas, 2011. 2013.
10. Huang W, Gaydos CA, Barnes MR, et al. Cost-effectiveness analysis of Chlamydia trachomatis
screening via Internet-based self-collected swabs compared with clinic-based sample collection. Sex Transm Dis 2011; 38: 815–820.
11. Centers for Disease Control and Prevention. Update to CDC’s Sexually Transmitted Diseases Treatment Guidelines, 2010: Oral cephalosporins no longer a recommended treatment for gonococcal infections. MMWR 2012; 61: 590–594.
12. Centers for Disease Control and Prevention. Recommendations for the laboratory-based detection of Chlamydia trachomatis
and Neisseria gonorrhoeae
—2014. MMWR 2014; 63: 1–19.
14. Bureau of Labor Statistics, Labor USDo. Employer Costs for Employee Compensation for the Regions—March 2013. Available at: http://www.bls.gov/ro7/ro7ecec.htm
. Accessed July 25, 2013.
18. Abramowicz M. Drugs for bacterial infections. Med Lett 2013; 11: 65–74.
19. Partnership Healthplan of California. Formulary Guide. Medi-Cal; 2014.
20. Smith KJ, Cook RL, Ness RB. Cost comparisons between home- and clinic-based testing for sexually transmitted diseases in high-risk young women. Infect Dis Obstet Gynecol 2007; 2007: .
21. Cook RL, Ostergood L, Hillier SL, et al. Home screening for sexually transmitted diseases in high-risk young women: randomised controlled trial. Sex Transm Infect 2007; 83: 286–291.