Objective
We sought to determine the prenatal human immunodeficiency virus (HIV) screening rate when using an opt-in policy and to find variables predictive of screening.
Study Design
This was a case-control study examining gravid women with a prenatal visit and a delivery at our hospital in 2005. Cases were defined as women who did not undergo HIV screening during the first or second prenatal visit. Our institution used an opt-in approach to HIV screening.
Results
Overall, 71% (291/412) of women underwent HIV screening at the first or second prenatal visit. Patient refusal was the most common reason for not being screened (15%; 62/412). Women who were ≤25 years old, were unmarried, and received care from maternal-fetal medicine attendings or family practitioners were more likely to undergo HIV screening.
Conclusion
With an opt-in approach, 29% of women were not screened for HIV during their early prenatal care. An opt-in policy also leads to screening rates that are provider dependent.
Identifying gravid women infected with human immunodeficiency virus (HIV) is important, as antepartum highly active antiretroviral therapy (HAART) reduces rates of vertical transmission from a baseline of 26% down to 1-2%. Yet, an estimated 882 cases of perinatal HIV transmission occurred between 2001-2004, and 92% of cases of acquired immunodeficiency syndrome among children <13 years of age in the United States were attributed to perinatal HIV transmission. Given the effectiveness of HAART, these cases were likely due to unrecognized HIV infection in gravid women.
The Centers for Disease Control and Prevention (CDC) recommends that HIV screening be part of the routine panel of prenatal tests for all pregnant women and that screening should follow an opt-out approach. The American College of Obstetricians and Gynecologists (ACOG) also recommends an opt-out approach and urges that all gravid women be screened for HIV as early as possible during each pregnancy. Despite these recommendations from the CDC and the ACOG, policies for prenatal HIV screening vary from state to state. A total of 22 states use an opt-out policy where HIV screening is a routine component of prenatal testing with women having the option to opt out. In the remaining 28 states, which includes Pennsylvania, a mandatory opt-in approach requires both detailed pretest HIV counseling and written informed consent. We sought to examine the impact of an opt-in policy on the rate of prenatal HIV screening among gravid women and to determine factors predictive of such screening.
Materials and Methods
Thomas Jefferson University Hospital, Philadelphia, PA, is an urban, tertiary referral, teaching facility with 4 prenatal clinics. In 2005, the Jefferson Center for Women’s Medical Specialties and the Jefferson Center for Maternal-Fetal Medicine (MFM) were staffed by 9 obstetrics-gynecology (OB-GYN) and 4 MFM attending physicians, respectively. In all, 32 OB-GYN residents (supervised by OB-GYN attendings) managed the Jefferson Obstetrics and Gynecology Associates clinic. The Jefferson Family Medicine Associates prenatal clinic was staffed by 3 family medicine attendings and 27 family medicine residents.
Our institution’s 4 prenatal clinics followed Pennsylvania state law’s opt-in policy to prenatal HIV screening. All obstetric care providers in 2005 were required to provide women with detailed pretest HIV counseling and education and to obtain written informed consent before performing HIV screening. Pretest counseling and education was provided by individual physicians, as our clinics did not have staff dedicated to offering such counseling. Physicians were not provided with a scripted recommendation for counseling. If a woman did not desire HIV screening, she only needed to verbally decline testing and she was not required to provide an explanation. Yet, if a woman gave a reason for her refusal, physicians were allowed to document the reason in the chart, but they were not required to do so.
Using our university hospital’s perinatal database, we identified all women with a documented delivery from Jan. 1, 2005, through Dec. 31, 2005. This search identified all women who had a live birth at any gestational age. Our search also identified women with stillbirths, but it was limited to identifying those who had one >28 weeks’ gestation. Given that our search for women with stillbirths was restricted, we excluded them from the study. From the population of women with a documented delivery, we then identified those with at least 1 prenatal visit in our outpatient clinics prior to delivery. This was done by performing a query of the billing claims data.
We determined that a sample size of 384 gravid women would provide an acceptably precise interval estimate, ie, a 95% confidence interval equal to the sample proportion ± .05. We estimated a 10% nonresponse rate (missing chart or incomplete records), and thus required a sample of 423 women from the population of women with a documented delivery and at least 1 prenatal visit. To identify a random group, we used the simple random sampling technique. We calculated the required sampling fraction, ie, required sample size of 423 divided by the total number of cases in the population, and then used a software random number generator (Excel; Microsoft Corp, Redmond, WA) to identify the random sample.
Each chart was reviewed by a team of trained chart reviewers. Demographic information obtained from the chart included maternal age, race/ethnicity, marital status, and insurance. The clinic where a woman received her prenatal care determined the type of obstetric care provider. Our institution’s 4 prenatal clinics were each primarily managed by a different type of provider in terms of specialty and level of training. The 3 OB-GYN clinics were staffed by OB-GYN attendings, OB-GYN residents, or MFM attendings. The family medicine prenatal clinic, on the other hand, was managed by both family medicine attendings and residents. Because we did not record the screening rates for individual providers, we combined the family medicine attendings and residents into a single “family practitioner” group.
From the chart, we also determined whether HIV screening was performed. This was accomplished by searching for documentation of an HIV antibody test result in the progress notes or by locating actual HIV antibody test results in the laboratory section of the chart. Only results from HIV screens ordered at the first or second prenatal visit were accepted as controls for our case-control study. We defined controls as such because both the CDC and the ACOG recommend HIV testing be performed as early as possible during prenatal care. For women who did not undergo HIV screening during the first 2 visits (cases), we attempted to find explanations from the chart.
Due to our study’s primary focus on the rate of HIV screening and our desire to protect patient privacy in regard to HIV status, we did not record the results of the HIV screening tests. In Philadelphia, PA, the estimated incidence of HIV in 2006 among female individuals aged 13 years and older was 62 per 100,000 population. There were an estimated 4835 women living with HIV or acquired immunodeficiency syndrome in 2006. HIV data in Philadelphia, PA, were considered incomplete before 2006, as HIV name-based reporting started in October 2005.
All statistical tests were performed using software (SPSS 17.0; SPSS Inc, Chicago, IL). To analyze continuous variables, we used either Student t test or the analysis of variance test. The χ 2 test was used to analyze categorical variables. P < .05 was considered statistically significant. Both univariate and multivariate analyses were performed to identify categorical variables predictive of women undergoing HIV screening. For the multivariate analysis, we used a binary logistic regression model. Covariates in the binary logistic regression model included any variable from univariate analysis that was associated with HIV screening with P < .20. This cutoff value was chosen so that we could liberally include variables in the model and therefore assess confounding by more variables. Results from the binary logistic regression were reported as adjusted odds ratios with 95% confidence intervals. This study was approved by our institutional review board.
Results
In 2005, 2073 women had a delivery and at least 1 prenatal visit prior to their delivery at our university hospital. From this population, we identified a random sample of 423 women. Two charts had insufficient prenatal records, 4 women had a stillbirth, and 5 charts could not be located. Therefore, our study included 412 gravid women. Demographic characteristics are presented in Table 1 . No woman in our random sample had a history of HIV infection.
| Variable | Controls: screened for HIV (n = 291) a | Cases: not screened for HIV (n = 121) a | P |
|---|---|---|---|
| Age, y | 26.0 ± 6.0 | 28.9 ± 6.3 | < .001 b |
| Marital status | |||
| Unmarried c | 208 (71) | 66 (55) | .001 d |
| Married | 83 (29) | 55 (45) | |
| Race/ethnicity | |||
| African American | 165 (57) | 52 (43) | .05 d |
| Asian | 33 (11) | 12 (10) | |
| Caucasian | 68 (23) | 45 (37) | |
| Latina | 11 (4) | 6 (5) | |
| Other or unknown | 14 (5) | 6 (5) | |
| Insurance | |||
| Medicaid | 156 (54) | 62 (51) | .66 d |
| Commercial | 135 (46) | 59 (49) | |
| Obstetric care provider | |||
| Family practitioner e | 39 (13) | 1 (1) | < .001 d |
| MFM attending | 24 (8) | 6 (5) | |
| OB-GYN resident | 139 (48) | 63 (52) | |
| OB-GYN attending | 89 (31) | 51 (42) |
a During first or second prenatal visit;
c Unmarried women included those who were single, widowed, or divorced;
OB-GYN attendings and OB-GYN residents provided care for 140 of 412 (34%) and 202 of 412 (49%) women, respectively. MFM attendings and family practitioners treated 30 of 412 (7%) and 40 of 412 (10%) women, respectively. Table 2 shows the characteristics of the population seen by each obstetric care provider group. Women seen by OB-GYN residents and family practitioners tended to be younger, unmarried, and on Medicaid when compared to those treated by OB-GYN and MFM attendings. OB-GYN residents and family practitioners also provided care for a population with a higher proportion of African Americans.
