APPENDIX B
COST AND PROBABILITY ASSUMPTIONS USED IN COST-EFFECTIVENESS ANALYSIS OF PRENATAL HIV SCREENING
Category |
HIV Negative on ELISA Test |
HIV Positive or Indeterminate on ELISA Test |
Laboratory |
||
First ELISA |
$5 |
|
Three ELISA tests |
|
$10 |
Western Blot |
|
$35 |
Risk assessmenta |
$2 |
$ 2 |
Counseling |
||
Pretest |
$7 |
$ 7 |
Posttest |
$1 |
$45 |
a For selective screening only. |
Cost Assumptions
-
All cost figures are estimates of average costs, in 1989 dollars, per person screened.
-
The column ''HIV Negative on ELISA Test'' shows average costs for individuals screened who are negative on the ELISA and require no confirmatory Western Blot testing. This includes HIV-negative individuals who test negative on the first ELISA (the majority) and require no further testing. It also includes HIV-negative persons whose first ELISA is borderline reactive but whose two repeat ELISA tests run in duplicate on the same blood specimen are clearly negative.
-
According to blood donor screening data, about two-thirds of all initially positive ELISA results are negative on repeat.1
It is assumed that the blood sample for the HIV test is drawn with the standard prenatal panel after informed consent has been obtained. The average cost of $5 per person screened includes the estimated cost of the laboratory test and the cost of providing information that is sufficient for informed consent concerning the purpose, benefits, and risks of the HIV test.
-
The column "HIV Positive or Indeterminate on ELISA Test" shows the average cost for individuals screened who require Western Blot testing. This includes those individuals who test positive on the first and two repeat ELISA tests. It also includes those who test positive on the first and one of the two repeat ELISA tests; some of these will yield a positive Western Blot result. Some will have a negative or indeterminate Western Blot and require further testing with additional blood specimens. The blood donor screening studies cited earlier have shown that about one-third of those testing positive on the first ELISA were persistently positive on the repeat ELISA, and 10 percent were also positive on the Western Blot. Of those repeatedly but weakly positive on the ELISA, only about 1 percent were positive on Western Blot testing.
-
To estimate the average laboratory cost a telephone survey was conducted of manufacturers of the reagents and of several commercial and public laboratories in the San Francisco Bay area. Average cost estimates are based on information provided by large public laboratories where economies of scale occur both in terms of manufacturer discounts and in processing the specimens with adequate quality control.
-
In selective screening, a risk assessment cost will be incurred for all women who present for prenatal care. It is assumed that risk assessment consists of a written instrument that will be part of an intake questionnaire. For most individuals this will be sufficient, although some women with risk factors may need further explanation in the course of the intake visit; the estimate is for average cost.
-
Under all screening strategies, universal as well as selective, all women who test HIV positive will receive specific HIV-related counseling. In selective screening, all women will receive specific HIV-related counseling, even if they test HIV negative, because all women screened have risk factors.
In universal screening, the cost implications of providing specific HIV-related counseling to HIV-positive women only are compared with the costs of providing such counseling to all women screened.
-
The lowest basic cost estimates for counseling have been used in the analysis because counseling is by far the most expensive item in the screening program cost. It was assumed that specific HIV-related pretest counseling would be added on to general pregnancy-related health counseling and would require a minimal marginal cost for most women who do not have risk factors. On the other hand, for some women with risk factors, the HIV-specific additional counseling would be more intensive. The analysis includes an estimated average pretest counseling cost of $7 per woman screened. It was also assumed that posttest counseling for HIV-negative women would consist of handing out an informative pamphlet about HIV prevention, which would be sufficient for most women. Some women who have known risk factors, however, may require specific information concerning risk reduction in their particular circumstances.
Counseling for HIV-positive women is assumed to involve intensive posttest counseling, including information on the risk of transmission, prevention, and health care and reproductive options. The average cost of counseling has been estimated at $53; in the case of screening programs with counseling for all, this would include an average cost of $7 for pretest counseling.
Probability Assumptions
-
p(D/NoRF) = 0
i.e., the probability of disease, given that there are no risk factors, is equal to 0; all HIV infection occurs in women with risk factors only.
If, on the other hand, there is a nonzero background rate in women without risk factors, then selective screening will miss more than the assumed 50 percent of HIV-positive women.
For example, if the background rate in women without risk factors is 0.0001 (the estimated probability of HIV infection in a U.S. adult without risk factors), then selective screening will miss about 6 HIV-
-
positive women per 100,000 women. At low overall prevalences, this means that selective screening can at best pick up only about a quarter of HIV-positive women rather than one-half.
-
To calculate the prevalence of HIV infection among women with risk factors, the following identity has been used:
p(D) = p(D/RF) × p(RF) + p(D/NoRF) × p(NoRF),
where
p(D) |
= |
the overall regional prevalence among childbearing women (from newborn screening data) |
p(D/RF) |
= |
the conditional probability of disease given risk factor(s) |
p(RF) |
= |
the estimated regional prevalence of risk factors among childbearing women (estimated to be 0.10 at baseline) |
p(D/NoRF) |
= |
the conditional probability of disease given no risk factor (estimated to be 0 in the present analysis, or 0.0001 in sensitivity analysis) |
p(NoRF) |
= |
1 minus the estimated regional prevalence of risk factors among childbearing women (estimated to be 0.90 at baseline) |
-
p(TEST) = p(TEST/HIV+)
i.e., the probability of testing is equal to the probability of testing HIV-positive women.
This is a severe restriction that tends to underestimate the cost of selective screening and overestimate the cost of universal screening:
-
For selective screening, this means that 50 percent of HIV-positive women are identified when only 50 percent of the women with risk factors are tested.
In reality, considerably more than 50 percent of the women may need to be tested in order to detect 50 percent of HIV-positive women. Thus, the cost per HIV-positive woman identified is considerably higher, as shown by sensitivity analysis.
-
For universal screening, this means that 95 percent of pregnant women would need to be tested in order to detect 95 percent of HIV-positive women.
In reality, a much smaller proportion of the pregnant population may need to be tested. In some low-prevalence regions, many women may correctly assess themselves as having no risk, so that little is lost if they refuse testing.
N.B.: If the goal is not only to detect the maximum number of HIV-positive women but also to counsel the maximum number of women at risk who are not yet infected, then the assumption concerning the prevalence of risk factors becomes crucial. Depending on that prevalence, universal screening with counseling for all may be advantageous even at a relatively low rate of infection in the region.