Varghese, Beena; Peterman, Thomas A.; Holtgrave, David R.
In the almost 20 years since the first cases of AIDS were reported, many approaches to preventing infection have been designed and evaluated. As prevention resources are limited, choosing the best strategy/ approach or combination of approaches depends on the population at risk, the possible interventions, and the cost-effectiveness of those interventions. Even though such decisions are made under conditions of uncertainty, modeling available information can provide valuable insight about the best approaches and identify critical assumptions that influence the estimates.
Two major HIV prevention programs in the USA are risk-reduction counseling and HIV testing and notification of partners of HIV-infected persons. These strategies have received relatively limited attention regarding their effects on averting future HIV infections. Counseling and testing has been modeled in the past by using different outcome measures as the number of infected persons tested, number of HIV infections averted, number of life years gained and number of infected persons who return for results[1-3].
Partner notification studies have compared different referral methods used but have not evaluated partner notification for prevention of future transmission of HIV[4-6]. The main outcome measure in partner notification cost-effectiveness studies has been the number of HIV positive partners identified. Partner notification has often been dismissed as being too costly, but it has not been evaluated for cost-effectiveness in preventing future HIV infections.
This study examines the cost-effectiveness of counseling and testing and partner notification offered at HIV and sexually transmitted disease (STD) clinics in preventing future HIV infections in the USA. For the counseling and testing model we analyze the cost-effectiveness of testing and counseling infected and uninfected persons and the subsequent behavior changes that leads to prevention of future HIV infections. The partner notification model on the other hand, looks at the cost-effectiveness of finding, testing, and counseling partners of infected index patients. We use recent data on behavior change in decision models to estimate the number of future HIV infections averted, and to learn (i) if counseling and testing is a cost-effective prevention method as compared with a ‚no intervention option‚; and (ii) if partner notification increases cost-effectiveness when it is added to an existing counseling and testing program.
We developed decision tree models (Treeage DATA) that use both societal and provider perspectives. The societal perspective includes all costs and benefits incurred by providers and clients and the discounted treatment cost for HIV. The provider perspective includes only explicit economic costs to the provider (HIV and STD clinics) and excludes the cost of client time and the lifetime treatment cost of HIV/AIDS. Cost estimates for counseling and testing and partner notification were obtained from the literature and are expressed in 1997 dollars using the medical care services component of the Consumer Price Index[1,4,6,7]. All cost estimates are incremental costs required to add another unit of either counseling and testing or partner notification service to an existing clinic where these programs are currently offered.
The main components of the analysis for counseling and testing and partner notification models are: (i) estimation of the number of HIV infections averted; (ii) calculation of the total costs (societal and provider); and (iii) calculation of the cost per future HIV infection averted (incremental cost-effectiveness ratios). The prevention methods are considered cost-effective if additional benefits are obtained within an acceptable range of additional cost; if additional benefits are greater than the costs, then the programs are considered cost saving. Sensitivity analyses were performed to assess the robustness of the decision model results to changes in the assumptions in the model. The base-case parameter values and ranges used in the sensitivity analysis are listed in Tables 1, 2 and 3.
Estimation of infections averted
The number of future HIV infections averted by the intervention included transmission from HIV-infected individuals and acquisition of HIV infection by uninfected individuals reached by the intervention. A value of one was assigned to each future infection that occurred and a value of zero when no infection occurred (Figs 1 and 2). Subtracting the number of future infections that occur following an intervention from the number of infections that occur with no intervention gives the number of future HIV infections that the intervention might prevent.
Counseling and testing model
A simplified decision model used for comparing counseling and testing with no counseling and testing is shown in Fig. 1. The HIV seroprevalence at counseling and testing clinics was assumed to be 0.015 (1.5%), the average of all federally funded HIV testing sites. It is assumed that the client has arrived voluntarily at the counseling and testing site for testing and therefore there is no variable describing the acceptance of testing services in the model. The return-rate for post-test counseling varies by test site, expected serostatus, risk categories, and demographic factors. The average return rate of 75% for HIV infected individuals and 67% for uninfected individuals was used. The probability that the current partner of an HIV infected individual is infected was estimated to be 20%, based on data from partner notification studies which found that 18-40% of partners were infected [4-6, 10-12].
The risk of HIV transmission from an infected individual to an uninfected partner was taken from a study of discordant couples in which 7% of sexually active asymptomatic persons transmitted HIV infection to their partners within 2 years. Another study has reported a per-partner HIV transmission probability of 10-30%. Based on this, a transmission rate of 7% was used with no post-test counseling and a range of 5-30% was used in the sensitivity analysis. Studies have found that 20-80% of those who learn of their seropositive status reduce their risk behavior[15,16]. Another study has used an estimate of 20% and 50% (lower and upper limit used) as the proportion of seropositive people who change their risk behavior after counseling. Post-test counseling of an infected person was therefore assumed to reduce the risk of transmission for half of the seropositive clients who receive counseling and reduce risk of transmission by 50% (to 0.035); a range of 10-70% was used in the sensitivity analysis.
The risk of acquiring HIV infection for an uninfected individual attending a STD clinic was 0.35% in the year following enrollment in a prevention study. The study found a 20% reduction in the risk of acquiring a STD at 12-month follow-up for client-centered counseling compared to didactic messages (the reduction compared to no counseling is unknown). Based on this, we estimated that post-test counseling reduced the risk of acquiring an infection by 20%, within a 1-year time period, with a range of 10-50% in the sensitivity analysis.
Partner notification model
Introducing partner notification to an existing counseling and testing program brings additional cost and benefits (Fig. 2, Table 2). For partner notification, every infected index patient provides a list of partners who might be infected (usually partners from the past year). The partners are notified either by the index patient (patient referral) or by the provider (provider referral); sometimes a combination of both may also be used. However, not all partners are found. In four partner notification studies, between 0.57 and 0.67 partners were found and tested for every infected person who was offered partner notification services through provider referral[4,6,11,19]. Provider referral was found to be almost eight times more successful in identifying partners. Therefore, we assumed that for every index patient identified by counseling and testing, 0.6 partners were found and offered testing via provider-referral and 0.08 partners via patient-referral[6,20]. Studies have reported an HIV seroprevalence of 18-46% among partners; in this study an estimate of 20% was used[4,10,12]. The probability that a partner who is found and offered HIV counseling and testing will receive post-test counseling has varied from 54% to 100% among different partner notification studies, with most studies showing high return rates[4,6,10,12]; a return rate if 90% was used in this model.
In the partner notification model, we included the risk of HIV transmission from the infected index patient to their uninfected partners and from infected partners to their uninfected partners. We assumed that for 25% of the infected partners identified by partner notification, the infected index patient is their only partner, and no HIV transmission could occur. We assumed the remaining infected partner notification partners would have additional partners besides the index patient and that 20% of those partners were already HIV infected and not at risk for acquiring HIV[4,6,10-12].
For the infected partners never reached by the partner notification program, the risk of transmitting HIV was similar to the individuals in the counseling and testing models who did not learn that they were infected (0.07)[13,14]. We assumed that infected partners of the index patient who were notified about their exposure, but refused additional counseling or HIV testing, still had a 25% reduction in risk of transmitting HIV (from 0.07 to 0.052)[15-17]. In the sensitivity analysis, this decrease in risk was varied between 10% and 50%. Post-test counseling of an infected partner was assumed to reduce the risk of transmission to their partners (second generation infection) by 50% (to 0.035) as in the counseling and testing model[15-17].
Among the uninfected partners who are notified but do not get tested/counseled, this risk is assumed to decrease by 25% (to 0.0263), with a range of 10% to 50% (0.017-0.0236) in the sensitivity analysis. If the uninfected partner is also notified and receives post-test counseling, we assumed an additional 25% reduction in risk of transmission in this couple to 0.017.
The incremental cost of adding counseling/testing and partner notification services to an existing clinic is expressed in 1997 dollars (Table 3). The cost estimates of Farnham et al.  and Toomey et al.  were adjusted for inflation using the medical care services component of the Consumer Price Index. The total costs of counseling and testing were estimated for HIV infected and uninfected individuals who did or did not receive post-test counseling. These costs reflect the incremental costs to a clinic visit, and so do not include patient travel and waiting time costs during the initial visit to the clinic or the cost of collecting blood specimens.
For an infected individual who received post-test counseling, the societal cost estimate included pretest and post-test counseling, tests [three enzyme linked immunosorbent assays (ELISA) and Western blots], administrative costs, and patient time costs for counseling. This total was estimated to be $129 in 1997 dollars. The cost for an infected individual who did not receive post-test counseling was estimated to be $90.
The societal costs for counseling and testing were much lower for uninfected individuals. For an uninfected individual who received post-test-counseling, the cost included pretest counseling, brief post-test counseling, one ELISA, and patient time cost, and administrative costs, adding to a total $41. The total cost for the uninfected individual who does not receive post-test counseling is $25.
Partner notification studies have reported the cost of identifying a partner in the range $300-700 (1997 dollars)[4,6,11,19]. We used $439 as the cost of finding and offering an HIV test to a partner and a range of $300-700 in the sensitivity analysis. We assumed that there would be no provider cost for partner notification via patient referral, although some programs may spend time teaching strategies for informing partners. The cost of counseling and testing were added to obtain the total cost of partner notification. The estimated lifetime treatment cost of HIV, discounted at 3% ranges from $154000 to $195000[21-23]. We used $175000 in our model and a range of $100000 and $250000 in the sensitivity analysis.
The provider perspective is designed to identify costs for the provider, and thus always excludes patient costs and disease treatment costs. The provider cost for an infected individual who receives post-test counseling was estimated to be $104; no post-test counseling was $84. Similarly, for uninfected individuals with post-test counseling, the cost was $26; no post-test counseling was $19.
Our model predicts that counseling and testing 10000 individuals would prevent almost eight future HIV cases and result in a net societal cost saving of almost $1million (Table 4). Provider costs for adding this intervention to existing facilities would be almost $248000 or approximately $32000 per case averted.
Testing 10000 individuals when the HIV prevalence is 1.5% would identify 150 infected individuals. However, because only 75% of them return for results, we assumed that 113 index patients would be available for partner notification. Provider referral for these 113 index patients would prevent another 1.24 future cases and save an additional $182000 (net savings). To the provider, this translates to an added cost of $28000 per additional case averted. Partner notification via patient referral prevents only 0.17 additional cases (over counseling and testing alone) but at what we assume to be essentially no additional cost to the provider.
The one-way sensitivity analyses for the models show that the cost savings ratios for counseling and testing and partner notification appear to be most sensitive to risk of HIV transmission, risk of acquiring HIV infection, effectiveness of counseling on behavior change, and treatment cost of HIV. Threshold analysis indicates that counseling and testing will cease to be cost saving to society if: (i) lifetime treatment costs of HIV/AIDS are less than $40000; or (ii) the risk of acquiring HIV infection (for the uninfected) decreases from 3.5/1000 to 1/1000 and at the same time, the risk of HIV transmission (from the infected) decreases from 7/100 to 1/100. Partner notification would cease to be cost saving if: (i) lifetime treatment cost of HIV/AIDS is less than $20000; or (ii) the risk of HIV infection for the unnotified uninfected partner decreases from 3.5/100 to 1.6/100.
Changing the effectiveness of counseling also had an impact on the results of the counseling and testing model. If counseling infected persons leads to a 70% reduction in risk behavior, 10 additional future infections would be prevented. This would result in a 6% increase in societal savings and a 10% decline in provider cost per case averted. If counseling uninfected persons leads to a 50% reduction in risk behavior, 15 future infections would be prevented. This would result in 11% increase in societal savings and a 40% decrease in the provider cost per case averted.
The sensitivity analyses on all other variables do not show a substantial impact on the cost savings; counseling and testing and partner notification remain cost saving HIV prevention methods under a variety of other probability estimates.
Our estimates of the costs and benefits of counseling and testing suggest that counseling and testing saves societal dollars. From the provider (clinic‚s) perspective, the cost of preventing an HIV infection seems reasonable at almost $32000 per infection averted. However, if the term provider includes the entire health care system or ministry of health, the lifetime treatment cost of HIV would add to the benefits obtained from counseling and testing and would make counseling and testing a cost saving HIV prevention method from the provider perspective as well. With such an approach, the difference between the societal and provider perspective becomes minimal.
In the USA, partner notification for HIV has generally been considered to be very expensive. However, our analysis suggests that partner notification results in societal cost saving under a wide range of probability estimates. This model incorporates the benefits of counseling and testing into partner notification services which reach a group of people most likely to benefit from counseling and testing. Therefore this study suggests that policies should encourage provider referral as an added prevention tool along with counseling and testing, as these two methods provide large societal savings and prevent HIV infection. Even from the provider perspective partner notification appears to be a worthwhile effort at HIV prevention at $28000 per additional infection averted.
Our estimates of the effectiveness of the interventions used heterosexual transmission and infection rates and used a higher incidence of HIV (0.35%) among STD clinic patients. For the general population in the USA with much lower incidence, our model might be overestimating the benefits of counseling and testing. However, among men who have sex with men and injecting drug-users, the incidence and transmission rates may be higher, and if other parameters remain unchanged an increase in risk of transmission would increase the cost savings obtained from the counseling and testing and partner notification prevention methods among these populations. As information on the risk of transmission or infection among other groups becomes available, required changes can be made in the model to enhance the accuracy of our estimates.
Our cost estimates used a 3% discount rate and a low estimate of the lifetime treatment cost of HIV($175000). With higher estimates, these prevention tools will result in increased savings to society. Also, the model did not include the cost of lost productivity from HIV morbidity and mortality and so it underestimated the societal savings. However, we used only the incremental cost of adding counseling and testing and partner notification to a clinic visit, thus underestimating the total costs of these services. Fixed and other related costs can be added to the model in settings where counseling and testing and partner notification are not currently offered.
This model (like all other models) has certain limitations and therefore must be considered within the context of the probabilities and information used here. One important limitation of the counseling and testing and partner notification model is that it does not address the possibility of a person having more than one partner. Neither does the counseling and testing model take in to account the number of secondary and tertiary infections averted. Adding these to the decision tree would surely increase the cost savings obtained from counseling and testing and partner notification, but would also make this model very complex. Another important limitation of our model is the small amount of data available on the effectiveness of counseling. Better information would enhance the model‚s reliability.
The third and perhaps the most important limitation is the use of 1-2 year risk as compared with a lifetime risk of HIV infection to measure the final outcome of future HIV cases prevented. Persons who do not acquire an infection during this first year are at risk in subsequent years, so for some, infection is only postponed. On the other hand, our model did not take into account second and third generation transmission that would have occurred if infections were not averted. Estimating the lifetime probability of HIV infection is beyond the scope of this paper and could be a topic for future research. Finally our model addressed the costs and benefits of prevention counseling and does not address the costs and benefits associated with treating HIV infected persons who are identified by the intervention.
The main role of HIV counseling is prevention of future HIV cases. Studies have shown that quality counseling does prevent infection. Our analysis suggests that quality counseling and testing and partner notification services save money for society in this setting. This model can be adapted to other settings.
1. Farnham PG, Gorsky RD, Holtgrave DR, et al. Counseling and testing for HIV Prevention: Costs, Effects, and Cost-Effectiveness of More Rapid Screening Tests. Public Health Rep 1996, 111:44-53.
2. Kassler WJ, Dillon BA, Haley C, Jones WK, Goldman A. On-site rapid testing with same day results and counseling. AIDS 1997, 11:1045-1051.
3. Holtgrave DR, Qualls NL, Graham JD. Economic evaluation of HIV prevention programs. Ann Rev Public Health 1996, 17:467-488.
4. Rutherford GW, Woo JM, Neal DP, et al. Partner notification and the control of human immunodeficiency virus infection. Sex Trans Dis 1991, 18:107-110.
5. Hoffman RE, Spencer NE, Miller LA. Comparison of partner notification at anonymous and confidential HIV test sites in Colorado. J Acquir Immune Defic Syndr Hum Retrovirol 1995, 8(4):406-410.
6. Toomey KE, Peterman TA, Dicker LW, et al. HIV partner notification: cost and effectiveness data from an attempted randomized controlled trial. Sex Trans Dis 1998, 25:310-316.
7. Consumer price indexes of medical care services U.S. Bureau of Labor Statistics, CPI Detailed Report. Http://126.96.36.199/cgi-bin/surveymost
8. HIV Counseling and Testing in Publicly Funded Sites: 1996 Summary Report. Atlanta, U.S. Department of Health and Human Services. Centers for Disease Control and Prevention, 1997.
9. Valdiserri RO, Moore M, Gerber AR, Campbell CH, Dillon BA, West GR. A study of clients returning after HIV testing: implications for improving rates of return. Public Health Rep 1993, 108:12-18.
10. Wells KD, Hoff GL. Human immunodeficiency virus partner notification in a low incidence urban community. Sex Trans Dis 1995, 22:377-379.
11. Pavia AT, Benyo M, Niler L, Risk I. Partner notification for control of HIV: results after 2 years of a statewide program in Utah. Am J Public Health 1993, 83(10):1418-1424.
12. Wykoff RF, Jones JL, Longshore ST, et al. Notification of the sex and needle-sharing partners of individuals with human immunodeficiency virus in rural South Carolina: 30-month experience. Sex Trans Dis 1991; 18:217-222.
13. De Vincenzi I. A longitudinal study of human immunodeficiency virus transmission by heterosexual partners. New Engl J Med 1994, 331:341-346.
14. Mastro TD, De Vincenzi I. Probabilities of sexual HIV-1 transmission. AIDS 1996, 10(Suppl A):S75-S82.
15. McKay NL, Phillips KM. An economic evaluation of mandatory premarital testing for HIV. Inquiry 1991, 28:236-248.
16. Higgins DL, Galavotti C, O‚Reilly KR, et al. Evidence for the effects of HIV antibody counseling and testing on risk behaviors. JAMA 1991, 266:2419-2429.
17. Holtgrave DR, Valdiserri RO, Gerber RA, Hinman AR. Human immunodeficiency virus counseling, testing referral and partner notification services. A cost-benefit analysis. Arch Intern Med 1993, 153:1225-1230.
18. Kamb ML, Fishbein M, Douglas JM, et al. Efficacy of risk-reduction counseling to prevent human immunodeficiency virus and sexually transmitted diseases: a randomized controlled trial. JAMA 1998, 280:1161-1167.
19. Spencer NE, Hoffman RE, Raevsky CA, Wolf FC, Vernon TM. Partner notification for human immunodeficiency virus infection in Colorado: results across index case groups and costs. Int J STD AIDS 1993, 4:26-32.
20. Landis SE, Schoenbach VJ, Weber DJ, et al. Results of a randomized trial of partner notification in cases of HIV infection in North Carolina. New Engl J of Med 1992m 326:101-106.
21. Hellinger FJ. The lifetime cost of treating a person with HIV. JAMA 1993, 270:474-478.
22. Holtgrave DR, Pinkerton SD. Updates of cost of illness and quality of life estimates for use in economic evaluations of HIV prevention programs. J Acquir Immune Defic Syndr Hum Retrovirol 1997, 16:54-62.
23. Gable CB, Tierce JC, Simison D, et al. Costs of HIV/AIDS at CD4+ counts disease stages based on treatment protocols. J Acquir Immune Defic Syndr 1996, 12:413-420.
© 1999 Lippincott Williams & Wilkins, Inc.