A coarsened multinomial regression model for perinatal mother to child transmission of HIV
2008

New Model for Estimating HIV Transmission from Mother to Child

Sample size: 2052 publication Evidence: high

Author Information

Author(s): Gard Charlotte C, Brown Elizabeth R

Primary Institution: University of Washington

Hypothesis

Can coarsened multinomial regression models provide better estimates of perinatal mother to child transmission of HIV compared to standard logistic models?

Conclusion

Coarsened multinomial regression models are preferred to standard logistic models for estimating perinatal mother to child transmission of HIV, especially when data is missing or tests are off-schedule.

Supporting Evidence

  • The proposed cumulative and conditional models performed well compared to logistic models.
  • Power to estimate intrapartum and perinatal transmission was consistently higher for the proposed models.
  • Simulation results showed that the coarsened multinomial model reduced bias and mean squared error.

Takeaway

This study found a new way to better estimate how HIV is passed from mothers to babies, especially when some test results are missing.

Methodology

The study used coarsened multinomial regression models and compared them to standard logistic models through simulations.

Potential Biases

The approach relies on the assumption of non-informative missingness, which may not hold true for all endpoints.

Limitations

The model assumes that missing data is non-informative and may not separate intrapartum transmission from early breastfeeding transmission.

Participant Demographics

The study analyzed data from 2,052 liveborn infants born to HIV positive mothers, with 1,758 having complete data.

Digital Object Identifier (DOI)

10.1186/1471-2288-8-46

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication