Marothi LETSOALO
  • Me
  • Research
    • Overview
    • PhDDesk
    • Collaborations
  • Trainings
  • News
  • Blog
  • Apps & Tools

On this page

  • Why This Work Matters
  • Scientific Problem
  • Research Question and Statistical Framing
  • Program Structure
  • Aim 1: Methodological Audit
  • Aim 2: Social Risk Phenotypes
  • Aim 3: Longitudinal Physiological Burden
  • Aim 4: Dynamic Translational Prediction
  • Expected Contribution
  • References

Other Formats

  • RevealJS

PhD

Precision Maternal Epidemiology in Obstetric HIV: A Biostatistical Framework for Latent Risk Stratification and Dynamic Prediction

Author

Marothi Letsoalo

Published

March 25, 2026

Why This Work Matters

  • Maternal outcomes in HIV remain shaped by interacting biological, clinical, and structural risk.
  • Existing population-average models can hide high-risk subgroups in routine antenatal care (Naicker et al. 2022; Serunjogi et al. 2025).
  • Multi-country longitudinal evidence supports the need for stronger phenotype-based risk stratification (Taha et al. 2022).

Scientific Problem

  • Severe maternal outcomes arise from layered biosocial mechanisms that are not well represented by static models.
  • Longitudinal antenatal records are often irregular in timing, complicating inference (Asparouhov and Muthen 2024; Wickrama et al. 2016).
  • Classification uncertainty in latent approaches can bias downstream associations if not explicitly corrected (Bauer and Curran 2003; Bakk and Vermunt 2014).

Research Question and Statistical Framing

Primary research question:
How can person-centered latent phenotype modeling improve individualized severe maternal outcome risk stratification in obstetric HIV under real-world, irregular longitudinal follow-up?

Supporting statistical questions:

  1. How can latent-variable methods identify unobserved maternal risk subphenotypes without overextraction bias?
  2. How can social risk phenotypes be linked to severe maternal outcomes while correcting for classification error and baseline confounding?
  3. How can we preserve valid longitudinal inference when antenatal visits are asynchronous across participants and sites?
  4. How can joint latent class prediction provide updated, clinically interpretable risk estimates at decision timepoints?

Program Structure

Aim Focus Statistical Core
1 Methodological audit GRoLTS + SMART-aligned reporting diagnostics
2 Social risk phenotypes Multilevel LCA, BCH correction, causal weighting
3 Longitudinal burden phenotypes Continuous-time mixture modeling
4 Dynamic prediction Joint latent class and landmark-style updating

References: (Van de Schoot et al. 2017; Lissa et al. 2024; Asparouhov and Muthen 2014; Rizopoulos et al. 2017)

Aim 1: Methodological Audit

  • Audit class extraction, random starts, fit diagnostics, and reporting quality.
  • Evaluate sensitivity to distributional misspecification and overextraction risk (Bauer and Curran 2003).
  • Enforce transparent latent trajectory reporting standards (Van de Schoot et al. 2017; Lissa et al. 2024).

Aim 2: Social Risk Phenotypes

  • Model latent social vulnerability classes from household and contextual indicators.
  • Apply three-step correction for class uncertainty and robust distal estimation (Asparouhov and Muthen 2014; Bakk and Vermunt 2014).
  • Integrate inverse propensity weighting for baseline-confounding control (Le et al. 2024).

Aim 3: Longitudinal Physiological Burden

  • Use continuous-time modeling to preserve asynchronous visit structure (Asparouhov and Muthen 2024).
  • Estimate multivariate trajectory phenotypes that reflect evolving clinical burden (Wickrama et al. 2016).
  • Connect biosocial profiles to phenotype evolution across gestational time.

Aim 4: Dynamic Translational Prediction

  • Use joint latent class modeling to combine marker history with time-to-event risk (Rizopoulos et al. 2017; Proust-Lima et al. 2023).
  • Produce updated individual risk profiles at clinical landmarks.
  • Evaluate practical utility for triage decisions and clinical benefit (Heagerty and Zheng 2005; Vickers and Elkin 2006).

Expected Contribution

  • A reproducible precision epidemiology framework for obstetric HIV.
  • Better early identification of high-risk maternal profiles.
  • A stronger bridge between methodological rigor and actionable maternal health decision support.

References

Asparouhov, Tihomir, and Bengt Muthen. 2014. “Auxiliary Variables in Mixture Modeling: Three-Step Approaches Using Mplus.” Structural Equation Modeling: A Multidisciplinary Journal 21 (3): 329–41. https://doi.org/10.1080/10705511.2014.915181.
Asparouhov, Tihomir, and Bengt Muthen. 2024. “Mixture Modeling with Individual Measurement Times.” Mplus Technical Report. https://www.statmodel.com/download/Tscores.pdf.
Bakk, Zoltan, and Jeroen K Vermunt. 2014. “Robustness of Stepwise Latent Class Modeling with Continuous Distal Outcomes.” Structural Equation Modeling: A Multidisciplinary Journal 21 (1): 20–35. https://doi.org/10.1080/10705511.2014.955104.
Bauer, Daniel J., and Patrick J. Curran. 2003. “Distributional Assumptions of Growth Mixture Models: Implications for Overextraction of Latent Trajectory Classes.” Psychological Methods 8 (3): 338–63. https://doi.org/10.1037/1082-989X.8.3.338.
Heagerty, Patrick J, and Yingye Zheng. 2005. “Survival Model Predictive Accuracy and ROC Curves.” Biometrics 61 (1): 92–105. https://doi.org/10.1111/j.0006-341X.2005.030814.x.
Le, Khoa, Johannes Clouth, and Jeroen K Vermunt. 2024. “Causal Latent Class Analysis with Distal Outcomes: A Modified Three-Step Method Using Inverse Propensity Weighting.” Multivariate Behavioral Research 59 (5): 1006–26. https://doi.org/10.1080/00273171.2024.2367485.
Lissa, C. J. Van, M. Garnier-Villarreal, and D. Anadria. 2024. “Recommended Practices in Latent Class Analysis Using the Open-Source r-Package tidySEM.” Structural Equation Modeling: A Multidisciplinary Journal 31 (3): 526–34. https://doi.org/10.1080/10705511.2023.2250920.
Naicker, Nivashnee et al. 2022. “Pregnancy Rates and Outcomes in a Longitudinal HIV Cohort in the Context of Evolving Antiretroviral Treatment Provision in South Africa.” BMC Pregnancy and Childbirth 22 (1): 596. https://doi.org/10.1186/s12884-022-04829-2.
Proust-Lima, Cecile, Tiphaine Saulnier, Viviane Philipps, et al. 2023. “Describing Complex Disease Progression Using Joint Latent Class Models for Multivariate Longitudinal Markers and Clinical Endpoints.” Statistics in Medicine 42 (22): 3996–4014. https://doi.org/10.1002/sim.9844.
Rizopoulos, Dimitris, Geert Molenberghs, and Emmanuel MEH Lesaffre. 2017. “Dynamic Predictions with Time-Dependent Covariates in Survival Analysis Using Joint Modeling and Landmarking.” Biometrical Journal 59 (6): 1261–76. https://doi.org/10.1002/bimj.201600238.
Serunjogi, Robert et al. 2025. “Risk of Adverse Birth Outcomes and Birth Defects Among Women Living with HIV on Antiretroviral Therapy and HIV-Negative Women in Uganda, 2015–2021.” Journal of Acquired Immune Deficiency Syndromes 98 (5): 434. https://doi.org/10.1097/QAI.0000000000003596.
Taha, Taha E, Nonhlanhla Yende-Zuma, Stephen S Brummel, et al. 2022. “Effects of Long-Term Antiretroviral Therapy in Reproductive-Age Women in Sub-Saharan Africa (the PEPFAR PROMOTE Study): A Multi-Country Observational Cohort Study.” The Lancet HIV 9 (6): e394–403. https://doi.org/10.1016/S2352-3018(22)00037-6.
Van de Schoot, Rens, Marit Sijbrandij, Sonja D. Winter, Sarah Depaoli, and Jeroen K. Vermunt. 2017. “The GRoLTS-Checklist: Guidelines for Reporting on Latent Trajectory Studies.” Structural Equation Modeling: A Multidisciplinary Journal 24 (3): 451–67. https://doi.org/10.1080/10705511.2016.1247646.
Vickers, Andrew J, and Elena B Elkin. 2006. “Decision Curve Analysis: A Novel Method for Evaluating Prediction Models.” Medical Decision Making 26 (6): 565–74. https://doi.org/10.1177/0272989X06295361.
Wickrama, Kandauda AS, Tae Kyoung Lee, Catherine Walker O’Neal, and Frederick O Lorenz. 2016. Higher-Order Growth Curves and Mixture Modeling with Mplus: A Practical Guide. Routledge. https://doi.org/10.4324/9781315642741.

© 2020 - 2026

ORCID: 0000-0003-2170-6312

Quarto  |  VS Code  |  Codex  |  R