Mandatory HIV testing of pregnant women would increase the detection of HIV infection, thereby allowing initiation of antiretroviral drug therapy in infected women and possibly decreasing the number of infants born with HIV infection. However, some women may avoid prenatal care because they fear HIV testing or resent the mandatory policy. These women would lose the opportunity to receive benefit from zidovudine therapy if they were infected, and, more important, they would lose the benefit of prenatal care regardless of whether they were infected with HIV. Mandatory testing might deter the women who are the most likely to be infected with HIV from receiving prenatal care [4].

Conversely, under a voluntary HIV testing policy, some HIV-infected women may refuse testing and lose the benefit of zidovudine therapy. With voluntary testing, however, no woman would avoid prenatal care because of a fear of mandatory testing. Furthermore, a voluntary testing policy respects the patient's autonomy in making this important decision.

Using a decision analysis model, we quantitatively evaluated the relative benefits and risks of mandatory compared with voluntary HIV testing of pregnant women.

Methods

Top Methods Results Discussion Author & Article Info References

The Model

We constructed a decision tree for mandatory compared with voluntary HIV testing (Figure 1) and evaluated it with Decision Maker 7.0 software (New England Medical Center, Boston, Massachusetts). The decision node represents the decision between mandatory and voluntary HIV testing. Each of these two main groups are further divided into a group that receives prenatal care and a group that does not.

View larger version (42K): [in this window] [in a new window]   Figure 1. Decision tree showing testing and treatment patterns under a mandatory and a voluntary policy of testing for HIV in pregnant women. Deterrence rate is defined as P2 – P1, the percentage of women deterred from seeking prenatal care (PNC) because of a mandatory policy. The decision tree is divided into a mandatory arm and a voluntary arm. The decision tree continues to branches A and B, which illustrate infant outcomes after maternal HIV testing (branch A) or no maternal HIV testing (branch B). Note that the last-level probability of infant death in branch B depends on whether the mother receives prenatal care. 1 –P1 = percentage of women who are currently seeking prenatal care; 1 –P2 = percentage of women who would seek prenatal care under a policy of mandatory testing.

Under a policy of mandatory HIV testing in pregnant women, women who receive prenatal care would be tested for HIV infection and all those who do not receive prenatal care would not be tested. Under the current voluntary system, however, pregnant women who receive prenatal care are given the option of HIV testing. As with mandatory testing, women who do not receive prenatal care are not tested.

Our decision tree includes two branches that relate infant outcome to whether HIV testing is performed and whether antiretroviral therapy is initiated. Women tested for HIV are either infected or not infected (Figure 1, branch A). An HIV-infected woman can either accept or refuse zidovudine therapy. The infant's outcome would be "not infected," "infected," or "dead" on the basis of the probability of HIV transmission and perinatal death. If the woman is not infected, the infant's outcome would be either "not infected" or "dead."

Women who are not tested for HIV infection would be either infected or not infected (Figure 1, branch B). Zidovudine therapy would not be offered to these women because their HIV status is unknown. The probability that the infant of an HIV-infected woman will become infected is increased if the woman does not receive zidovudine. The probability that the infant will die is higher if the woman does not receive prenatal care.

Statistical Assumptions

The probabilities for each variable in the model were either obtained from a review of the literature (based on a MEDLINE search) or estimated when reliable data were not available. We evaluated the effect of varying each probability within a clinically relevant range (Table 1).

Acceptance of Prenatal Care

In 1993, 1.6% of pregnant women received no prenatal care and 3.2% of women did not seek care until the third trimester (total, 4.8%) [5]. We chose 1.6% as the baseline value of the percentage of women receiving no prenatal care under a voluntary HIV testing policy, which translates into a 98.4% acceptance rate for prenatal care. We also tested the model at a rate of 4.8% for women receiving no prenatal care (95.2% acceptance rate). We hypothesized that acceptance of prenatal care would be lower under a mandatory HIV testing system than under a voluntary system.

Deterrence Rate

Under either a mandatory or a voluntary HIV testing policy, some women will not receive prenatal care. We defined the deterrence rate as the percentage of women deterred from seeking prenatal care under a mandatory HIV testing policy (P₂) in excess of those not currently receiving prenatal care (P₁). The deterrence rate is calculated as P₂ – P (₁) in Figure 1. For example, if 5 women in 1000 avoid prenatal care because of mandatory HIV testing, the deterrence rate is 0.5%. Because no data have been reported for this variable we varied the deterrence rate from 0.0% to 3.0% and chose 0.5% as the baseline value.

Acceptance of Voluntary HIV Testing

Depending on the setting, 77% to 97% of pregnant women agree to HIV testing [6-8]. Barbacci and colleagues [7] found that acceptance of HIV testing among pregnant women at high risk for HIV increased from 77% in 1987-1988 to 90% in 1989-1990 when HIV testing was offered with counseling. Even when HIV testing was not offered to pregnant women reporting a low risk for HIV infection, 48% asked for the test. However, when the same group was offered testing after being counseled, 90% accepted the test. Lindsay and colleagues [8] found a 97% acceptance rate among 4731 inner-city women who had been counseled about HIV testing during pregnancy. Using these data, we varied the acceptance of voluntary HIV testing from 72% to 97% in the model and used 85% as the baseline estimate. Acceptance of HIV testing in the mandatory branch was, by definition, maintained at 100%.

Prevalence of HIV in Women of Child-Bearing Age

The nationwide prevalence of HIV infection among women of child-bearing age was 1.7 per 1000 (0.17%) in 1993 [9]. At 5.8 cases per 1000, New York State has the highest prevalence of HIV infection among women of child-bearing age in the United States [10]. We used 0.17% as the baseline value for the prevalence of HIV infection among women of child-bearing age and varied the overall HIV prevalence in this segment of the population to 0.58% in the model. Throughout this article, the term national HIV prevalence refers to the prevalence of HIV infection among all women of child-bearing age in the United States.

Prevalence of HIV in Pregnant Women Not Receiving Prenatal Care

In the model, we used a range of 0.17% to 5.0% for HIV prevalence in women not receiving prenatal care. We chose this range because some demographic subgroups at higher risk for HIV (such as injection drug users) may also have decreased access to and use of prenatal care. Broekhuizen and colleagues [16] found that 45.6% of illicit drug users in their study compared with 17.2% of those not using drugs had five or fewer prenatal visits.

Acceptance of Zidovudine by HIV-Infected Pregnant Women

In ACTG-076, acceptance of zidovudine therapy was 95% [1]. Because the ACTG-076 cohort was highly motivated, the Figure isprobably an overestimate of the actual acceptance rate. We estimated that the acceptance rate ranges from 72% to 95% and used 85% as the baseline value for acceptance of zidovudine therapy in both the mandatory and voluntary branches of the model [11].

Benefit of Zidovudine in Reducing Vertical Transmission of HIV

In ACTG-076, maternal-fetal transmission of HIV was 7.9% in the zidovudine group and 27.7% in the placebo group [1]. In our model, the transmission rate without antiretroviral therapy was held constant. However, we used a transmission rate among the recipients of antiretroviral therapy that ranged from a high of 7.9% to a low of 0.0% to account for the possibility that some of the newer antiretroviral agents may eliminate maternal-fetal transmission of HIV. Of note, ACTG-076 found that gestational age at study entry had no statistically significant influence on maternal-fetal transmission [1].

Benefit of Prenatal Care

Quantifying the effect of prenatal care on perinatal mortality is difficult, and various studies have reached different conclusions [17]. Broekhuizen and colleagues [16] found that prenatal care had a greater effect on perinatal mortality in illicit drug users than in those not using illicit drugs. Among illicit drug users, perinatal mortality was 2.0 to 3.0 times higher in those who had fewer than five prenatal visits than in those who had more than five prenatal visits. Foster and colleagues [18] found a significant decrease in fetal deaths, from 35.4 to 7.0 per 1000 live births, in uninsured patients receiving prenatal care. In contrast, Piper and colleagues [19] found that expansion of Medicaid eligibility for prenatal care had no effect on perinatal mortality. The differences in the results of these studies are probably caused by differences in the study populations, the nature of prenatal care, and the size of the study samples.

In 1988, the U.S. Office of Technology Assessment conducted an extensive analysis of the effectiveness of prenatal care and found that prenatal care has a beneficial effect on two critical aspects of infant health: low birthweight (as a predictor of infant mortality) and neonatal mortality (independent of birthweight) [13]. Eisner and colleagues [14] examined U.S. birth registration data for single live births and found that lack of prenatal care was associated with a two- to sixfold increase in low birthweight.

On the basis of findings from Eisner and coworkers' study, we estimated the benefit of prenatal care indirectly by using birthweight, neonatal mortality, and fetal mortality (Appendix). We used a baseline perinatal mortality rate of 12.7 per 1000 live births [12] and then calculated the perinatal mortality rate without prenatal care. We derived a perinatal mortality rate without prenatal care of 21.8 per 1000 live births. Some studies have shown that the effectiveness of prenatal care depends on many variables, including the nature of services provided [17, 18]. We therefore varied the perinatal mortality rate without prenatal care from 17.25 per 1000 (half our calculated benefit) to 30.9 per 1000 (double our calculated benefit).

Estimated Utility

The estimated utility of a healthy infant was 1. The estimated utility of a dead infant was 0. The estimated utility of an HIV-infected child was conservatively estimated to be 0.1; this value was calculated by dividing the median life expectancy of an HIV-infected newborn (approximately 8 years [2]) by the average life expectancy in the United States. The estimated utility of infected infants was varied from 0 to 1 in the sensitivity analysis to demonstrate the effect of this utility on the deterrence rate.

Sensitivity Analysis

Deterrence Rate

While varying each of the model components through their specified ranges (Table 1), we determined the threshold deterrence rate for prenatal care above which a voluntary HIV testing policy would be preferred to mandatory testing. In this analysis, if the actual deterrence rate is higher than the calculated threshold deterrence rate, the model favors voluntary HIV testing. Conversely, a lower actual deterrence rate favors mandatory testing.

Absolute Numbers of Infants Spared HIV Infection

It has been estimated that HIV infection would be prevented each year in about 1400 infants in the United States if 100% (compared with 0%) of HIV-positive women received zidovudine during pregnancy [2]. The number of infants spared infection under the current voluntary HIV testing policy would approximate this Figure if the rates of voluntary testing and acceptance of zidovudine were to approach 100%. We used the model to calculate the absolute number of infants who would be spared HIV infection under a mandatory compared with a voluntary HIV testing policy. We compared this difference to the difference in perinatal mortality associated with mandatory compared with voluntary testing. The advantage of this analysis is that the model does not require assignment of estimated utilities for infants.

Results

Top Methods Results Discussion Author & Article Info References

Sensitivity Analysis: Deterrence Rate for Prenatal Care

Deterrence Rate at Baseline Values

At baseline values for all of the other model components, the threshold deterrence rate is 0.40%. Therefore, if more than 4 women per 1000 women were deterred from seeking prenatal care because of a mandatory HIV testing policy, a voluntary policy is preferred in the model (Figure 2).

View larger version (26K): [in this window] [in a new window]   Figure 2. Sensitivity analysis of the decision model for mandatory compared with voluntary HIV testing in pregnant women. Threshold deterrence rate is the deterrence rate above which a voluntary HIV testing policy is preferred. The horizontal bars illustrate the effects on the deterrence rate of varying each component in the model through its specified range of values (shown at the ends of each bar) while holding the values of the other components at baseline.

Effects on Deterrence Rate

The deterrence rate did not deviate widely from baseline when acceptance of zidovudine therapy or efficacy of antiretroviral therapy was varied through the specified ranges. When the estimated benefit of prenatal care was doubled, the threshold deterrence rate decreased to 0.20%; at half the benefit, it increased to 0.78%. Varying the estimated utility of HIV-infected infants changed the threshold deterrence rate from 0% to 0.44%.

When the overall prevalence of HIV infection was increased to 0.58%, the deterrence rate increased to 1.29%. However, even at this high overall prevalence, when voluntary HIV testing was increased from baseline to 97%, the deterrence rate decreased to 0.26%. When the overall HIV prevalence was kept at the baseline value (0.17%) and the prevalence in the group not receiving care was increased to 5%, the deterrence rate decreased to 0.18%.

Of all of the model components, acceptance of HIV testing had the most dramatic negative effect on the deterrence rate. At the baseline values, a 10-fold decrease in the deterrence rate, from 0.78% to 0.08%, occurred when acceptance of voluntary HIV testing was increased from 72% to 97%.

Stacking against Voluntary Testing

We evaluated the model using values that favor mandatory testing while varying acceptance of voluntary HIV testing. When the prevalence of HIV infection was set at 0.58%, acceptance of zidovudine therapy was set at 95%, acceptance of HIV testing was set at 72%, and all other model components were set at baseline, the deterrence rate increased to 2.66%. However, in this same scenario, when acceptance of HIV testing was increased to 97%, the deterrence rate decreased 10-fold, to 0.28%.

Sensitivity Analysis: Absolute Numbers of Infants

Varying Acceptance of Zidovudine Therapy

Table 2 shows the effect of varied acceptance of zidovudine therapy across a wide range while holding the other components of the model at their baseline values. According to our model, the decrease in the estimated number of infants with HIV infection would be less than the increase in the number of perinatal deaths resulting from a change to a mandatory policy, although the benefit of a voluntary system decreases moderately as acceptance of zidovudine therapy increases.

Varying the Acceptance of Voluntary HIV Testing

Table 2 also shows the effect of varying degrees of acceptance of HIV testing under a voluntary system. If voluntary HIV testing is at a low level, the decrease in the number of infants with HIV infection as a result of mandatory testing is greater than the increase in perinatal deaths caused by such a policy. However, as voluntary HIV testing increases to 97%, the benefit of a mandatory system diminishes significantly while the adverse effect of deterrence remains.

Varying Both Deterrence Rate and Prevalence of HIV

We calculated the number of infants born with HIV infection and the number of perinatal deaths under a mandatory screening policy at baseline and high HIV prevalence rates using two values for the deterrence rate (Table 3). The difference in the number of perinatal deaths depends solely on the deterrence rate, whereas the difference in the number of infants born with HIV infection depends on HIV prevalence and deterrence rate.

At the current national HIV prevalence of 0.17%, the decrease in the number of infants born with HIV infection under a mandatory policy would be less than the increase in the number of perinatal deaths under such a policy, assuming that mandatory testing would deter at least 0.5% of women from seeking prenatal care. With greater than a threefold increase (from 1.7 to 5.8 per 1000) in the national HIV prevalence among women of child-bearing age and at a deterrence rate of 0.5%, the decrease in the number of HIV-infected infants under a mandatory policy would be more than the number of excess perinatal deaths resulting from lack of prenatal care. However, when the more likely estimate of a deterrence rate of 1.5% was used, given the higher HIV prevalence, the analysis again supports a voluntary testing policy (Table 3).

Effect of a Deterrence Rate of 0.0%

Even if we assume that no woman would be deterred from seeking prenatal care by a mandatory screening policy, the maximum number of infants spared HIV infection annually in the United States would be only 34 if acceptance of voluntary HIV testing is 97% and 173 if acceptance of voluntary HIV testing is 85% (Table 4). These results are the same as those from an analysis that assumes no benefit of prenatal care.

Discussion

Top Methods Results Discussion Author & Article Info References

The purpose of our decision analysis was to quantify the tradeoffs of a mandatory compared with a voluntary HIV testing policy for pregnant women. Our model suggests that a voluntary policy is preferable to a mandatory policy over a broad range of variables used in the analysis.

One of the most important variables in our model is the acceptance of HIV testing under a voluntary policy. Increasing voluntary HIV testing to 97% significantly reduced the benefit of mandatory testing, even when the model was stacked against voluntary testing and when the deterrence rate was assumed to be 0.0%. This level of acceptance is already a reality in many settings [7, 8]. Outreach and educational efforts substantially improve acceptance of HIV testing and would have the greatest impact on reducing the number of infants born with HIV infection. Such an approach would also circumvent the potential adverse effect of deterrence from prenatal care as a result of a mandatory policy.

The deterrence rate emerged as another important variable in the model. If mandatory testing were to result in deterrence rates in the 0.5% range, implementing such a policy would mean that the number of infant deaths resulting from lack of prenatal care would be greater than the number of infants spared HIV infection. At a higher prevalence of HIV and lower rates of HIV testing under a voluntary policy, the threshold deterrence rate increases. If any deterrence from prenatal care occurs, however, the benefit of a mandatory system must be weighed against the resulting increase in perinatal mortality. The actual deterrence rate under a mandatory policy is not known. On the basis of our findings, we conclude that research analyzing the effect of mandatory HIV testing on the deterrence rate for prenatal care must be done before a mandatory policy is introduced.

The prevalence of HIV infection among women of child-bearing age was also important in the model. At the current prevalence, a voluntary policy is warranted. However, if the overall prevalence increases to 0.58% or more, the question of a mandatory HIV testing policy should be revisited. At that time, particular attention should be paid to the level of voluntary acceptance of HIV testing and deterrence rate from prenatal care, both of which are important variables.

Variables that were less important in the model included the acceptance of zidovudine therapy, the efficacy of antiretroviral therapy, and the prevalence of HIV in the women not receiving prenatal care.

From an ethical perspective, a mandatory testing policy sacrifices the autonomy of an entire segment of the population in deciding whether to be tested for HIV. Therefore, the benefits of mandatory HIV testing must be substantial and must be clearly demonstrated before such a policy is implemented. No data show that mandatory HIV testing would be beneficial. Our model suggests not only that mandatory testing would be of minimal benefit but also that it may have substantial adverse effects.

Focusing on mandatory testing as a quick solution to maternal-fetal transmission of HIV without addressing these important issues is probably futile and may do more harm than good, as shown in our model. Research should address strategies for improving voluntary acceptance of HIV testing through counseling and education. It also should address intermediate steps, such as acceptance of antiretroviral therapy and access to prenatal care, and consider the potential adverse effects of mandatory testing, such as deterrence from prenatal care.

Limitations

Our model assumes that zidovudine therapy causes no negative effects in the 92.1% of HIV-negative infants born to HIV-positive mothers who receive the drug. A recent National Institutes of Health panel [20] reviewed two studies in mice and found that exposure to high doses of zidovudine in utero increased the risk for cancer in offspring. Although these findings cannot be directly applied to humans, they emphasize our lack of knowledge about possible long-term adverse effects of exposure to zidovudine in utero. A voluntary testing policy would be even more preferable than a mandatory policy if such effects indeed occur in humans.

We did not include in our model the potential benefit derived from a woman's knowledge of her HIV status and receiving medical treatment if she is infected with the virus. We also did not include the benefit of a reduction in HIV transmission to sexual partners of infected women who undergo screening [21]. Finally, our model did not evaluate the effect of a woman's knowledge of her HIV status on her decision to avoid breast feeding, which would decrease the risk for transmission of HIV to the infant.

The lack of data on the actual deterrence rate for prenatal care with mandatory HIV testing makes it difficult to quantify the effect of such a policy in pregnant women. If the actual deterrence rate is significantly lower than we assumed, mandatory HIV testing might be preferred (depending on the values of the other variables). If the actual deterrence rate is close to or greater than our assumption, however, then HIV testing should remain voluntary.

Conclusions

A policy of voluntary HIV testing in pregnant women is preferable to mandatory testing over a broad range of variables in our model if the deterrence rate for prenatal care is 0.5% or more. At high levels of acceptance of voluntary HIV testing (a goal already realized in many settings), the benefits of a mandatory policy are significantly reduced even if the deterrence rate is 0%. Therefore, educational and outreach efforts to increase acceptance of voluntary HIV testing during pregnancy would be the most effective strategy for identifying women with HIV infection and reducing maternal-fetal transmission of HIV.

Our model supports the current policy of voluntary HIV testing in pregnant women, as recommended by the Centers for Disease Control and Prevention, and argues against the implementation of a mandatory testing policy. Additional data on the results of current policy recommendations are urgently needed before any major change in policy is implemented.

Appendix: Quantifying the Effect of Prenatal Care on Perinatal Mortality

In 1988, the U.S. Office of Technology Assessment found a relation between low birthweight (a reliable predictor of perinatal mortality) and a lack of prenatal care [13, 15].

The Equation forneonatal mortality with prenatal care is as follows: (NL x IL) + (NN x IN), where NL is the neonatal mortality rate for infants of low birthweight, NN is the neonatal mortality for infants of normal birthweights, IL is the incidence of low birthweight, and IN is the incidence of normal birthweight. We used a conservative estimate based on a national study by Eisner and colleagues [14], who found that prenatal care decreases the incidence of low birthweight (IL) by at least 50%. By using this Figure andIN = 1 –IL, we derived a neonatal mortality rate without prenatal care, as follows: (NL x IL/50%) + (NN x [1 –IL/50%]).

We used the same Equation for the rate of fetal deaths without prenatal care. By adding the neonatal and fetal mortality rates, we calculated the perinatal mortality rate without prenatal care at 21.8 per 1000 using 1991 vital statistics ([12]; Unpublished data from the linked birth/infant death data set, 1991 birth cohort, National Center for Health Statistics).

Dr. Longenecker: Johns Hopkins University, 2024 East Monument Street, Suite 2500, Baltimore MD 21205.

Ms. Ricksecker: Department of Community and Preventive Medicine, Allegheny University of the Health Sciences, Mailstop 921, Broad and Vine Streets, Philadelphia, PA 19102.

Dr. Healton: Columbia University School of Public Health, 600 West 168th Street, New York, NY 10032.

Dr. Smith: Department of Medicine, 2B Elkins, Abington Memorial Hospital, 1200 Old York Road, Abington, PA 19001.