Case study applications of the causal roadmap and targeted learning

Note all data is simulated and estimates of drug effects are for teaching purposes only and do not reflect real-world efficacy.

ART adherence and HIV virologic suppression

(Draft in progress) HIV adherence analysis plan and simulation results

This roadmap analysis plan defines a longitudinal causal estimand to evaluate how imperfect adherence to antiretroviral therapy affects HIV virologic suppression under realistic, observed treatment patterns. Using a structural equation model and directed acyclic graph, the plan formalizes the impact of intercurrent events such as treatment switching, disenrollment, and differential monitoring. An example analysis is implemented on HealthVerity claims data to estimate adherence–outcome relationships using longitudinal targeted maximum likelihood estimation (TMLE), adjusted for time-varying confounding and censoring. A simulation study then replicates key data-generating features—pharmacy fill patterns, lab engagement, and switching behavior—under varying degrees of adherence and confounding, to benchmark estimator performance and assess bias in standard Cox models versus TMLE under the causal roadmap framework

Causal Roadmap case study of an acute kidney injury safety analysis

https://bookdown.org/amertens/causal-roadmap-tutorial/

---