Standard methods of scenario and cost-effectiveness analysis were used to analyze the costs and consequences of the four cases listed above. The analyses employed a payer's perspective [16
] rather than a societal perspective so as to best estimate the resources needed to implement CDC's recommendations. All costs are expressed in 2005 US dollars, and all analyses were done in Microsoft Excel 2003 (http://office.microsoft.com
); see Table S1
under Supporting Information for the spreadsheet containing all formulae used here. A one-year time horizon was employed so as to examine intensively the initial impact of CDC's recommendations and the alternative scenarios. (In the Discussion section, the results are interpreted for the reader interested in a societal perspective and a multiyear time horizon.)
contains the input parameter values for the basic analysis, as well as the sources of the parameter values [18
]. Of course, some parameter estimates contain uncertainty, and where parameter estimation called for judgment to be made, I biased the estimates in favor of CDC's opt-out testing recommendations (as described below). The parameter values related to population size are self-explanatory.
Input Parameters, Values, and Sources
CDC recently published data that show that 73% of persons diagnosed with HIV in South Carolina had previously visited a health care facility and could have been tested for HIV, had routine testing been available [10
]. This value may be high for some populations heavily impacted by HIV with little access to health care (such as homeless youth), but is the best available estimate in the literature and likely biases the results in favor of opt-out testing.
For simplicity, it was assumed here that all incident and prevalent HIV infections in the US are among 13- to 64-year-olds; this too gives a slight bias in favor of CDC's recommendations. The analyses assume that the CDC routine HIV testing recommendations will achieve a first-year uptake of 52.2% (for a total of 109,620,000 HIV tests). This percentage is the uptake level equal to that of other recommended screening tests in the US [22
]. However, 21.0% of the adult population is already being tested for HIV [23
], so the actual uptake of testing due to the new recommendations is the difference between those two percentages (i.e., 31.2%).
The analyses assume that a two-step rapid testing strategy is to be used (an initial rapid HIV testing, followed by a confirmatory Western blot as necessary), as this would maximize client receipt of test results and is consistent with CDC's emphasis on using rapid testing where possible [1
]. The full cost of delivering a testing strategy with pre- and post-test counseling (from a payer's perspective) has been estimated at US$28.18 for HIV-seronegative people and US$103.92 for HIV-seropositive people [24
]. Prior analyses have estimated that about 61.5% of the cost of counseling and testing for HIV-seronegative clients is attributable to counseling (roughly half of that for pretest counseling, half for post-test counseling) [26
], bringing the testing-only cost down to US$13.01. Removing pretest counseling for people testing HIV seropositive reduces that expense by about US$7.59 to US$96.93 (this assumes CDC would still recommend post-test counseling to HIV-seropositive persons) [24
]. My estimates of the costs of testing services are within the ranges published by Phillips and Fernyak [39
]. (It is recognized that other testing technologies are available—such as non-rapid testing options—that may have slightly different cost levels; however, this analytic framework can be used to examine any testing technology of interest to the reader.)
The analyses separately calculate the number of persons who newly learn that they are HIV seropositive, and the number of persons who already know that they are HIV seropositive but are tested (and test seropositive) again due to new testing initiatives. Prior analyses have estimated that of persons testing HIV seropositive, about 37% already know they are HIV seropositive or do not return for results [24
]. For persons who newly learn that they are HIV seropositive, their medical care costs are estimated to be similar to those experienced by persons receiving care at a large HIV clinic in the southeastern US (where the median CD4 cell count is 367 and the range is 2 to 2,671) [30
]. However, this estimate of medical need for a newly diagnosed HIV-seropositive client can be multiplied by a constant to reflect higher or lower medical costs in any clinic population of interest.
To estimate the number of HIV infections averted, I first examined transmissions prevented from persons who newly learn that they are HIV positive due to the proposed program. It has been estimated that persons who are unaware that they are living with HIV transmit at an 8.8% annual rate; persons who are aware that they are living with HIV infection have been estimated to transmit at a 2.4% rate [24
]. Therefore, as persons learn of their HIV seropositivity, it is assumed that their transmission rate drops accordingly. It should be noted that omitting pretest counseling for persons who test HIV seropositive may diminish this change in transmission rate, but for the sake of simplicity such a “penalty” is not included here.
With these input parameters, I was able to calculate the following outputs using basic algebra: (a) number of persons tested under the recommended program; (b) number of undiagnosed HIV-seropositive persons newly reached; (c) total cost of testing program; (d) HIV transmissions averted; (e) gross cost per transmission averted; and (f) public sector medical care resources needed in one year to care for persons newly diagnosed with HIV infection. The analysis that calculates these outputs using the inputs described above is labeled the “Basic Case Analysis (Opt-Out Testing).”
In line with CDC's recommendations, the Basic Case Analysis (Opt-Out Testing) makes a simplifying assumption that the removal of HIV counseling for seronegative persons at high behavioral risk of infection does no harm. This assumption is counter to the literature, which notes that client-centered counseling accompanying testing can reduce incident sexually transmitted infections by 20% (or even more among adolescents) [5
]. Therefore, I created and assessed alternative scenarios.
The “Behavioral Offset Case Analysis” is exactly the same as the Basic Case Analysis (Opt-Out Testing) with one exception. It has been estimated by CDC's National Center for Health Statistics that roughly 11.7% (11.9% in one publication) of the US population 15–44 years old is at high behavioral risk of HIV infection [35
]. I make a simplifying assumption that this percentage holds for 13- to 64-year-olds but recognize that the actual percentage is not known and may vary by age. (Note that the Basic Case Analysis is actually the same as a Behavioral Offset Case sensitivity analysis but assumes that 0.0% rather than 11.7% of persons are at high behavioral risk.) It is possible that persons at high risk of HIV infection who are tested via CDC's recommended program (which omits risk assessment and counseling for HIV seronegative persons) could actually increase their risk behavior. For instance, if an injecting drug user who is given an opt-out HIV test without being questioned about substance use or counseled about risk gets an HIV-negative result, the individual could easily take that testing experience as a confirmation that injecting drugs is not posing an HIV-related risk. Indeed, some persons repeatedly seeking HIV testing use the experience as a risk confirmation strategy [40
]. Further, CDC and Kaiser Family Foundation have estimated that roughly four in ten persons in the US have some basic misconceptions about HIV [42
]. Hence, perfect HIV-related knowledge cannot be assumed among patient populations. In the Behavioral Offset Case, the rate of HIV incidence is calculated for persons at high risk of infection tested under CDC's recommended program, and it is assumed that the rate increases by 5% due to behavioral offset. This behavioral offset parameter is not known with much precision and suggests an important area for additional research.
Next, I created a Routine Counseling and Testing Case. This case assumed that clients received counseling and testing (but if someone reported no risk behaviors and tested HIV negative, no post-test counseling would be needed). While the additional counseling in this scenario would of necessity increase the cost relative to the basic program, this counseling would prevent infections among high-risk seronegative persons. The best available estimate in the literature for the effect size of counseling among high-risk seronegative persons is 20% [5
]. So as not to overstate the case for counseling (and so as to ensure that any potential bias is in favor of opt-out testing rather than against it), that effect size is reduced here to 15%. In other words, it was assumed that the number of incident HIV infections among HIV-seronegative persons at high risk of infection is reduced by 15% due to the provision of client-centered counseling and testing. Given that a small number of HIV-seronegative persons would fail to get their results even in a rapid testing situation, this benefit is decreased by 1% (consistent with prior peer-reviewed analyses) [24
Once the estimated cost of the Basic Case program is known, another type of analysis is possible. Recently, the US president proposed US$93 million for targeted HIV counseling and testing efforts in the US (focusing particularly on incarcerated populations, persons in drug treatment, and other persons in clinical and community-based service delivery settings at high risk of undiagnosed or imminent HIV infection). I have previously estimated that the president's proposed US$93 million counseling and testing initiative might (in one year) identify as many as 26,984 undiagnosed persons living with HIV, and prevent up to 2,537 transmissions and infections at an average cost of US$36,663 each [24
]. I used this framework of a targeted counseling and testing program analysis here to ask, “At the level of resources needed to fund CDC's recommended routine testing program, what would be the impact of a targeted counseling and testing program?” Hence, a scenario labeled “Targeted Counseling and Testing” was created.
In the Targeted scenario, it was assumed that the level of available resources for service delivery was the same as that estimated in the Basic Case Analysis described above (in other words, the resource level output of the Basic Case Analysis was used as the resource level input in the Targeted Counseling and Testing analysis). Besides the other input parameters in , additional assumptions were needed for this scenario. Given that this case assumes a highly targeted program, it would be possible to essentially ensure that counseling and testing was offered only to persons at high risk of infection. However, this percentage was instead assumed to be 50%, so as to allow for some offering of counseling and testing to lower-risk persons in areas of higher HIV seroprevalence. This scenario analysis assumes that 1% of persons being tested are HIV seropositive (which is less than the 1.5% seropositivity typically seen in publicly funded HIV counseling and testing sites in the US and used in previously published analyses) [24
]. Since this targeting might be achieved via ongoing surveillance and evaluation activities, no additional costs were allocated for additional targeting efforts or outreach; however, the assumptions about the ability of this program to reach populations with high HIV seropositivity was examined in sensitivity analyses.