An augmented illness-death model for semi-competing risks with clinically immediate terminal events

Area of Science:

• Biostatistics
• Perinatal Medicine
• Epidemiology

Background:

• Preeclampsia is a serious pregnancy-associated condition with significant fetal and maternal risks, resolving only after delivery.
• The timing of delivery relative to preeclampsia onset creates a semi-competing risks scenario, complicating risk assessment.
• Existing models do not fully capture the complex dependencies between preeclampsia development and delivery timing, especially distinguishing immediate vs. non-immediate deliveries.

Purpose of the Study:

• To develop a novel statistical model for semi-competing risks in preeclampsia research.
• To simultaneously characterize the risk of developing preeclampsia and the time to delivery post-diagnosis.
• To account for distinct delivery trajectories (clinically immediate vs. non-immediate) following preeclampsia onset.

Main Methods:

• Proposed a novel augmented frailty-based illness-death model incorporating a binary submodel.
• The model allows for direct and indirect dependencies between preeclampsia onset and delivery timing via shared frailty.
• Developed an efficient Bayesian sampler for model estimation and derived formulas for dynamic risk prediction.

Main Results:

• The proposed model effectively handles complex dependencies in semi-competing risks scenarios.
• Demonstrated applicability using electronic health record data on pregnancy outcomes.
• The model provides a framework for individualized risk prediction in preeclampsia.

Conclusions:

• The novel augmented frailty model offers a robust approach to analyzing preeclampsia and delivery timing.
• This framework enhances understanding of individualized risk prediction in clinical settings.
• The model is directly applicable to real-world clinical questions using electronic health record data.