Background: The benefits of combination antiretroviral (ARV) prophylaxis for infants whose HIV exposure is recognized near birth have been established, and the benefits of early ARV therapy are well known. Decisions about ARVs can be supported by the probability that the child has acquired HIV.
Methods: Using 2005-2010 data from Enhanced Perinatal Surveillance of the Centers for Disease Control and Prevention, we developed a tool for use at birth to help predict HIV acquisition of HIV-exposed infants to support ARV management. A logistic regression model, fit using a fully Bayesian approach, was used to determine maternal variables predictive of infant HIV acquisition. We created a score index from these variables, established the sensitivity and specificity of each possible score, and determined the distribution of scores among infants, with and without HIV, in our study population.
Results: Multivariable analysis of data from 8740 HIV-exposed infants (176 infected and 8564 uninfected) yielded 4 maternal variables in the perinatal HIV acquisition prediction model: sexually transmitted infection, substance use, last HIV viral load before delivery and ARV use. Using the regression coefficient estimates, we rescaled each possible score to make the maximum score equal to 100. For each score, sensitivity and specificity were determined; the area under the receiver operating characteristic curve was 0.79. Median index scores for infants with HIV and without HIV were 43 (first quartile 27 and third quartile 60), and 12 (first quartile, 0 and thirs quartile, 29), respectively.
Conclusions: Decisions to begin infants on 3 ARVs-whether considered therapeutic or prophylactic-can be supported by data available on the day of birth.