Joint modeling of correlated binary outcomes using multivariate logistic regression: contraception and HIV knowledge in Sri Lanka

Wijesekara, NM, Withanage, N and Abeynayake, NR ORCID: https://orcid.org/0000-0003-1738-9553 (2024) Joint modeling of correlated binary outcomes using multivariate logistic regression: contraception and HIV knowledge in Sri Lanka. Journal of Applied Statistics. ISSN 0266-4763

Accepted Version File will be available on: 7 June 2025. Available under License Creative Commons Attribution Non-commercial. Download (312kB)

Abstract

Reproductive health significantly contributes to the overall well-being and social welfare of women. Within the spectrum of modern and traditional contraceptive methods in use, condoms have been strongly advocated by numerous HIV programs as a primary means of preventing HIV infection in Sri Lanka. Given the intrinsic relationship between contraceptive utilization and HIV awareness, our study aims to concurrently analyze the patterns of contraceptive usage and HIV knowledge, while accounting for their potential correlation. In this study, we introduced the application of the Gumbel type II distribution to effectively capture the interdependence of these joint probabilities, accounting for various covariates. The outcome of simulation studies demonstrated the superior performance of the integrated joint model, in comparison to the separate univariate models. Our findings highlighted several noteworthy risk factors associated with both contraceptive usage and HIV prevention knowledge. These included variables such as residence, level of education, wealth quintile, husband's education level, number of children, engagement with the newspapers, television viewership, and mobile phone usage. The Result indicates a positive association between the adoption of contraception and the awareness of HIV prevention measures, suggesting that individuals who actively embrace contraception are more likely to possess knowledge about preventing HIV transmission.