As a result of the substantial investment of federal and local resources into the development and evaluation of different strategies for HIV prevention, there now is a body of evidence demonstrating that behavioral interventions can be effective in reducing sexual risk behaviors that lead to disease exposure among populations that are at increased risk for contracting and transmitting HIV infection. However, the next major challenge lies in the translation of what has been learned through behavioral research into the realities of community practice. Identifying cost-effective HIV interventions will help policy-makers select prevention strategies that optimize the use of resources to promote public health.
In this article, we report on a cost-effectiveness analysis of VOICES/VOCES, a single-session, video-based, interactive, group HIV intervention designed to be integrated relatively easily into routine practice in sexually transmitted disease (STD) clinics and similar settings that serve African American and Latino clients at high risk of HIV. This culturally appropriate intervention has been shown previously to be effective in reducing the incidence of new STD infections, as well as in increasing rates of condom acquisition and other attitudinal and self-reported behaviors associated with safer sex behaviors [1–3]. Notably, it is one of the few interventions shown to be effective for non-injecting-drug-using, heterosexual male populations . Unlike other group interventions that require participants to come to multiple sessions over time, the intervention was intended to be delivered at the time of a clinic visit and thereby inclused clinic clients who may not be sufficiently motivated to attend longer or more comprehensive HIV prevention programs.
This cost-effectiveness analysis of the VOICES/VOCES intervention has a major advantage over many previous studies that have only relied upon self-reported measures. The analysis presented herein is based on data collected on the incidence of STD infections over an average 17-month follow-up period. This modeling approach is important not only to evaluate this particular intervention, but for future cost-effectiveness studies, given the recent advances in STD diagnostic screening tools, such as urine bioassays, which are relatively inexpensive and non-invasive and thus feasible for inclusion in evaluation research [5–7].
Overview of VOICES/VOCES effectiveness study
In this analysis, data on effectiveness is drawn from the original research conducted by O'Donnell et al. at a large public STD clinic in the South Bronx from 1992–1994 . All clients attending the Morrisania STD clinic in the South Bronx during 1992 were approached at time of clinic registration and asked to participate in a randomized clinical trial of an STD prevention intervention. Following procedures approved by the institutional review boards of the Education Development Center, Inc. and the New York City Department of Health, those who provided written informed consent (99% of all approached) were assigned either to a control condition, in which they received regular clinical services, or to receive an experimental, video-based, group session in addition to routine services. (A full description of the study and randomization procedures are provided elsewhere .) Briefly, culturally relevant videos (about 20 min in length) were used to trigger a group discussion that followed a structured protocol emphasizing the importance of condom use and provided opportunities for practicing condom communication and negotiation skills. Sessions concluded with a condom education component, in which participants learned the proper ways to use condoms and how to select the right kind of condom to fit their needs, considering such issues as size, fit, lubrication and so on. Two award-winning videos were used, ‘Let's Do Something Different’ for groups composed of African Americans, and ‘Porque Si', in English and Spanish, for Latinos. Both videos were developed on the basis of extensive formative research with community groups . Groups were composed of four to eight individuals of the same gender and ethnicity and led by trained health educators and 97% of the clinic patients who were asked to enroll in the study completed it.
Several types of evaluation data were collected, including self-reports of HIV risk-related knowledge, attitudes, intentions, and behaviors as well as rates of coupon redemption for free condoms that were made available at a community location subsequent to the clinic visit. In addition, male participants (n = 2004) were tracked through the New York City STD surveillance system up to the end of 1993 to identify incident STDs. An incident STD was defined as any STD diagnosed at least 30 days after intervention. Although the clinical trial included both males and females, STD outcomes were not tracked among females because of the greater likelihood of asymptomatic gonorrhea (a relatively prevalent infection), which would lead to under-reporting and therefore greater right-censoring of observations within the follow-up period. Thus, these cost analyses, which include STD incidence as a predictor of HIV infection, model effectiveness based on STD infection among males. However, the model incorporated data on HIV/STD prevalence among females, and there were no significant differences by gender on other study outcomes.
The perspective for the cost analysis is that of STD clinic managers and funding agencies, such as federal, state, and local health departments. As the intervention was designed to be offered to clinic clients at the time of their visit (and could often be provided while they are waiting to be examined by a clinician or for test results), we excluded additional costs to the client. The costs associated with implementation were delineated into broad categories, itemized and documented over a 1-year period. Ranges of costs in each category across sites (low, average, high) were used in the model to yield ranges of outcomes.
Start-up costs included the costs for purchasing VOICES/VOCES materials, or ‘tool kits’ (two culturally relevant videos; administrator guide and facilitator manual; condom display board, condom samples, and carrying case), staff training (2-day course for site supervisor and facilitators), and equipment (TV/VCR). On average, we included the purchase of three VOICES/VOCES tool kits per site (range 2–4). Recurrent costs included staff salaries; space and facilities costs; refresher training (1 day), and ongoing technical assistance provided by the developer (averaged at 4 h/month). Staff salaries were the highest cost item, and estimated on the basis of a range from 5 to 15% of a supervisor's time, along with 1.5 to 2 full-time facilitators. Client-specific costs included a condom kit (envelope package containing three condoms) given at the conclusion of the intervention to each participant. Following conventions used in similar cost analyses, a discount rate of 5% was utilized for both cost and outcome estimations. Capital expenditures (VCR/TV purchase only) were annuitized and distributed across the life of the project with a 5% discount rate used in the annuity function. No research-related costs were included.
A summary of the ranges for calculating the above costs is presented in Table 1. Recurrent costs associated with the intervention averaged US$38.33 per client; start-up costs averaged only US$0.71 per client and client-specific costs were US$4.25. Thus, the total cost per client for the intervention averaged US$43.30.
Estimates of the number of incident HIV infections were derived from the following widely used probability-based formula: EQUATION where p is the average HIV prevalence among sexual partners of the target population, r is the risk of HIV transmission per act of unprotected sex (also known as 'infectivity'), f is the fraction of sex acts when a condom is used, e is the effectiveness of condoms, n is the average number of sex acts per partner of the target population, and m is the average number of sex partners for the target population . Model estimates are calculated separately for index cases who are HIV infected and those who are HIV uninfected. In the analysis for HIV-infected individuals, the result reflects the number of HIV infections that result from contact with HIV-uninfected sexual partners. The model estimates for HIV-seronegative individuals reflects clients becoming infected with HIV from their HIV-infected sexual partners. The time interval for the analysis was 1 year, and each parameter was estimated to an annualized value. Estimates were calculated by applying average parameter values across a population size of 10 000 and multiplying the probability of HIV by the population size. The number of HIV infections averted was calculated by taking the difference in estimates across the intervention and control groups. Base-case values for all parameters are shown in Table 1. All results are expressed in terms of the impact per 10 000 persons exposed, which is a both a common metric in epidemiologic analysis and a reasonable estimate of the number of clients who might be exposed to the intervention in a typical year at multiple clinics in a large urban setting.
As described earlier, the primary outcome for the original study was STD incidence. Baseline rates of condom use and sexual behavior were collected, but follow-up rates for the self-reported behavioral outcomes were not available as participants were tracked through the New York City STD surveillance system and not required to make a repeat visit. Rates of condom use were therefore back-calculated from STD incidence rates collected from the study using the following formula, which is derived from Equation 1 for STD rather than HIV incidence: EQUATION where s is STD incidence, z is the average STD prevalence among sexual partners of the target population, q is the risk of STD transmission per act of unprotected sex, e is the effectiveness of condoms, n is the average number of sex acts per partner of the target population, and m is the average number of sex partners for the target population. Values for the e, n, and m parameters are common across equations (1) and (2). The results from equation (2) are used for the f parameter in equation (1). As before, the time interval for the analysis was 1 year, and thus each parameter is annualized. Note that the model is constrained by the STD incidence values, which limits the range of values for other parameters as only certain combinations of STD prevalence and infectivity, condom use and efficacy, and frequency of contact with sexual partners generate valid results.
Sexual and epidemiologic parameters for models
Sexual behavior parameters (shown in Table 1) were based primarily on results from the baseline survey that was conducted at enrollment into the study. Values for the average number and frequency of sexual contacts were annualized and set to a constant as the analysis is designed to identify the number of HIV infections averted from changes in condom use alone. The prevalence of HIV among the study participants and their sexual partners were based on more recent seroprevalence data collected on clinic clients during 1997–1998. We assumed that the prevalence of HIV infection among sexual partners of HIV-infected participants would be higher than the rate among the sexual partners of HIV-uninfected participants, with average values of 7 and 5% assigned to each group, respectively. For the STD prevalence rate of sexual partners, we began with an estimate of the prevalence rate among study participants at the time of the clinic visit, as obtained through a random sample of patient charts conducted at the time of the original study (STD prevalence = 54%). This prevalence rate was used as an input into the model for calculating cases of HIV, along with other pertinent variables. Model output revealed that the range of STD prevalence was constrained by the measured incidence rate (that is, to achieve the given incidence of infection subsequent to the intervention, the prevalence rate was constrained to a given range of values (0.172 to 0.30). The STD incidence data were obtained from the New York City STD surveillance database. Estimates of condom effectiveness, HIV and STI infectivity were drawn from recently published studies. A wide range of studies have identified a condom effectiveness rate of approximately 90% . Values for HIV infectivity were reported to be approximately 0.0056 [95% confidence interval (CI), 0.0041–0.0075] . Values for STD infectivity were based on the likelihood of transmission in a single sex act for syphilis, gonorrhea, and chlamydia and were approximately 0.150 (95% CI, 0.10–0.253) .
Threshold analysis was conducted to identify when intervention costs would equal treatment costs, or how much can be spent per client on the intervention before there is a net loss in terms of treatment expenditure. This is also useful because variations in the implementation of the intervention may influence cost (such as, for example, if clients had to return to the clinic for intervention at a different time and thus incur transportation and opportunity costs, or if additional training was required for inexperienced facilitators). We calculated the amount of money that would have been spent on the HIV infection averted in lieu of the intervention, and divided this figure by the population size (10 000). This threshold is a useful gauge of how cost-effective the intervention is relative to not having the intervention.
Estimation of quality-adjusted life years saved and cost–utility ratio
Quality-adjusted life years (QALY) were estimated using a method developed by Holtgrave et al.  that was based on a survival and disease progression framework in the US setting. This involves dividing HIV infection into four stages: (1) full health and unaware of HIV infection; (2) 0.90 of full health for people aware of HIV infection and with a CD4 cell count of between 200 and 499 × 106 cells/l (based on a prior expert judgment study); (3) 0.65 of full health for persons with a CD4 cell count of less than 200 × 106 cells/l (based on prior empirical studies of AIDS patients); and (4) 0.40 of full health for persons with AIDS as defined by clinical conditions (based on four prior empirical studies with AIDS patients). These values can then be translated into an average QALY saved per infection averted by fitting the structure of these stages of HIV infections to the age distribution of the target population. For example, Holtgrave et al. reported that when the average age of infection was 26 years, the average QALY saved per infection averted was 11.23 when a 3% discount rate wais applied. Discount rates of zero and 5% generate an average of 23.87 and 7.10 QALY per infection averted, respectively. As the average age of HIV infection in our study population was approximately the same as in the Holtgrave et al. analysis, we used a triangular probability distribution function in our analysis with values of 11.23 for most likely, with 7.10 and 23.87 for the low and high extremes, respectively.
The results of cost–utility analyses are typically presented in terms of a cost–utility ratio comparing two alternative intervention approaches. The cost–utility ratio is a ratio of the net cost of the intervention program and the total QALY saved from the intervention. The formula for this is: EQUATION where C is the cost of the intervention compared to no intervention, A is the number of HIV infections averted from the program, T is the lifetime treatment costs saved by averting the infection, and Q is the number of QALY saved by averting an infection. Cost–utility ratios less than zero denote cost-savings from the intervention. Again, we adopted the approach of Holtgrave et al.  who suggest that health service programs with a cost–utility ratio of less than US$30 000 are frequently considered cost-effective, whereas those with a cost–utility ratio of greater than US$140 000 are difficult to justify as cost-effective. We calculated the cost–utility ratio using US-based, HIV-related treatment costs identified by Pinkerton and Holtgrave with conversion to 1999 dollars . On the basis of a synthesis of the published literature, they estimated an intermediate lifetime cost to treat HIV and AIDS of US$195 188 using 3% discounting, and US$157 348 to US$274 766 with zero and 5% discounting. We used a triangular probability distribution function for estimating treatment costs, adjusting the figures of Pinkerton and Holtgrave  for cost-of-living increases for New York City to 1999 dollars, with values of US$199 960 for the most likely, with US$161 219 and US$281 525, respectively, for the low and high extremes.
The results of the analysis are summarized in Table 2. Our model estimated that the average annualized condom use rate was 0.862 for the control group, and 0.948 for the intervention group. Even with relatively high rates of condom use, there was still a relatively high rate of incidence of STD among the study population.
The average annual cost to provide the intervention to 10 000 clients was estimated to be US$447 005 (range, US$340 163–557 174), with an average cost per client of US$43.30. This expenditure would result in an average of 27.69 HIV infections averted: 14.0 among the 9500 HIV-uninfected participants, and 13.4 among the HIV-negative sexual partners of the 500 HIV-positive clients. Almost 50% of the incident HIV infections were attributable to the 5% of HIV-infected clients who received the intervention. Note also that although the mean value for HIV infections averted is 27.69, the range is 4.79 to 202.78. However, the distribution is heavily skewed to the left. Thus, the mean value is a reasonable measure of central tendency.
The number of QALY saved from the intervention in this analysis averaged 387.61 (range 65.44–3488.30). Again, Table 2 shows that the distribution of results over the simulation analysis is strongly left skewed. The cost per HIV infection averted from the intervention averaged US$21 486 (range, US$2399–87 670). The QALY saved from the intervention averaged US$2217 (range, US$200–7664). The cost–utility ratio averaged −US$14 628 (range, −US$32 485 to −US$6022), with negative values indicating a cost savings at a level that is considered highly cost-effective . This cost-effectiveness is further evidenced by the cost savings from the intervention which averaged US$5 544 408 (range, US$428 950–45 359 520). The threshold for intervention to treatment costs averaged US$599 (range, US$83.86–4585) with the distribution summarized in Table 2. This indicates that, on average, the intervention could cost up to US$599 per client before these costs eclipse the treatment costs (compared with the actual cost of about US$43.00).
This study shows that a brief HIV behavioral intervention targeted to a high-risk population, namely male clients attending an urban STD clinic, can be highly cost-effective in reducing the economic burden of disease to the community. Our findings provide a counterpoint to the widely held belief that brief interventions have limited benefits, although increasingly there are questions about the level of intervention ‘dosage’ that is necessary to show evidence of behavioral impact or disease reduction [17,18]. Notably, reductions in STD outcomes persisted over an average of 17 months of follow-up and were greatest among those men who, at the time of study enrollment, reported having had recent sex with multiple partners. By reducing unprotected sexual encounters among this particular population over even this relatively short period of time, our analyses indicate it is possible to reduce the spread of disease in a cost-effective manner.
One reason that the intervention is so cost-effective is that it appears to work particularly well with men who report high numbers of sexual partners; our modeling illustrates how even a small reduction in the relatively small proportion of men with high numbers of sexual partners and encounters can have a significant impact on disease transmission. This is particularly true in locations, such as New York City, and among populations, including African American and Latino men, where rates of STDs and HIV seroprevalence are especially high. By providing brief behavioral interventions at the time of an STD clinic visit, it is possible to reach a high-risk population of men who may not have the time, resources, or motivation to seek out prevention services or attend longer or multiple session interventions, but who are critical vectors in the spread of HIV and other STDs in the community.
Given the importance of cost information for HIV prevention planning, another question that arises is, how does the cost-effectiveness of this intervention compare with other approaches? In 1992 Holtgrave et al. presented a review of 47 published abstracts or papers that provided data on costs and benefits ; this was updated in 1995, with an additional 46 studies . The majority of studies related to HIV counseling and testing and AIDS care and treatment. Although there was a doubling of total studies between the two reviews, only a minority pertain to primary prevention and none concerning a heterosexual male population. In interventions aimed at gay men, for example, costs per HIV infection averted have been estimated at US$65 000 for a community-level HIV risk reduction intervention , to US$4150.14 for a 1.5 h skills training workshop . Our estimated cost per infection averted (US$21 486) lies between the two; although it is substantially less cost-intensive than a community intervention, the lower rates of HIV infection in the target population of male heterosexuals results in a higher cost per case averted. On the basis of these finding we recommend that this effective HIV/STD intervention be considered for replication on a much larger scale.
1. O'Donnell L, San Doval A, Duran R, O'Donnell C. Video based sexually transmitted disease patient education: its impact on condom acquisition. Am J Public Health 1995, 85: 817 –822.
2. O'Donnell L, San Doval A, Duran R, O'Donnell C. The effectiveness of video-based interventions in promoting condom acquisition among STD clinic patients. Sex Transm Dis 1995, 22: 97 –103.
3. O'Donnell C, O'Donnell L, San Doval A, Duran R, Labes K. Reductions in STD infections subsequent to an STD clinic visit: using video-based patient education to supplement provider interactions. Sex Transm Dis 1998, 25: 161 –165.
4. Cohen DA, MacKinnon DP, Dent C, Mason HR, Sullivan E. Group counseling to promote use of condoms. Public Health Rep 1992, 107: 727 –731.
5. Oh MK, Richey CM, Pate MS, Brown PR, Hook EW. High prevalence of Chlamydia trachomatis
infections in adolescent females not having pelvic examinations: utility of PCR-based urine screening in urban adolescent clinic setting. J Adolesc Health 1997, 21: 80 –86.
6. Carroll KC, Aldeen WE, Morrison M, Anderson R, Lee D, Mottice S. Evaluation of the Abbott LCX ligase chain reaction assay for detection of chlamydia and gonorrhoeae in urine and genital swab specimens for an STD clinic population. J Clin Microbiol 1998, 36: 1630 –1633.
7. Gaydos CA, Howell MR, Quinn TC, Gaydos JC, McKee KT. Use of ligase chain reaction with urine versus cervical cultures for detection of chlamydia in an asymptomatic military population of pregnant and nonpregnant females. J. Clin Microbiol 1998, 36: 1300 –1304.
8. O'Donnell L, San Doval A, Vornfett R, DeJong W. Reducing AIDS and other STDs among inner-city hispanics: the use of qualitative research in the development of video-based patient education. AIDS Educ Prev 1994, 6: 140 –153.
9. Mastro TD, Satten GA, Nopkesomnn T, Sagnkharomya S, Longini IM. Probability of female-to-male transmission of HIV-1 in Thailand. Lancet 1994, 343: 204 –207.
10. Stigum HE, Mangus PE, Bakketeig LE. Effect of changing partnership formation rates on spread of sexually tranmitted diseases and human immunodefeficency virus. Am J Epidemiol 1997, 145: 644 –645.
11. Weinstein M, Graham J, Siegel J, Fineberg H. Cost-effectiveness Analysis of AIDS Prevention Programs: Concepts, Complications, and Illustrations.
Washington, DC: National Academy Press; 1989.
12. Pinkerton SD, Abramson PR, Turk ME. Updated estimates of condom effectiveness. J Assoc Nurses AIDS Care 1998, 9: 88 –89.
13. Holtgrave D, Qualls N, Graham J. Economic evaluation of HIV prevention. Ann Rev Public Health 1995, 17: 467 –488.
14. Pinkerton SD, Holtgrave DR. Lifetime costs of HIV/AIDS medical care. J Acquir Immune Defic Syndr Hum Retrovirol 1997, 14: 380 –382.
15. Holtgrave DR, Qualls NL. Threshold analysis and programs for prevention of HIV infection. Med Decis Making 1995, 15: 311 –317.
16. Pinkerton SD, Holtgrave DR, Bloom FR. Cost-effectiveness of post-exposure prophylaxis following sexual exposure to HIV. AIDS 1998, 12: 1067 –1078.
17. Branson BM, Peterman TA, Cannon RO, Ransom R, Zaidi AA. Group counseling to prevent sexually transmitted disease and HIV: a randomized controlled trial. Sex Transm Dis 1998, 25: 553 -560.
18. Kamb ML, Fishbein M, Douglas JM Jr. et al
. Efficacy of risk-reduction counseling to prevent human immunodeficiency virus and sexually transmitted diseases: a randomized controlled trial.Project RESPECT Study Group.
JAMA 1998, 280: 1161 –1167.
19. Holtgrave D, Valdiserri R., West G. Economic evaluation of HIV-related prevention and treatment services: a review.IX International Conference on AIDS
. Berlin, June 1993 [abstract PO-D36-4378].
20. Pinkerton S, Holtgrave D, DiFranceisco M, Stevenson L, Kelly J. Cost-effectiveness of a community-level HIV risk reduction intervention. Am J Public Health 1998, 88: 1239 –1242.
21. Pinkerton S, Holtgrave D, Valdiserri R. Cost-effectiveness of HIV prevention skills training for men who have sex with men. AIDS 1997, 11: 347 –357.