Risk-Based Decision Making: Estimands for Sequential Prediction Under Interventions.

Area of Science:

• Causal inference
• Medical decision-making
• Health informatics

Background:

• Standard prediction models may not adequately inform intervention choices, as they don't account for the intervention's effect.
• Clinical decisions often involve repeated, deferred, or reevaluated interventions, necessitating dynamic risk assessments.

Purpose of the Study:

• To outline key considerations for formulating estimands in sequential prediction under interventions.
• To guide the development of prediction models that support sequential medical decision-making.

Main Methods:

• Formalizing prediction tasks within a sequential, causal, and estimand framework.
• Illustrating considerations with a case study on delivery method choices (vaginal vs. cesarean section).

Main Results:

• Highlights the importance of defining estimands relevant to specific intervention strategies in sequential settings.
• Provides a framework for addressing the complexities of time-varying intervention decisions.

Conclusions:

• Emphasizes the need for sequential prediction models that provide reconsiderable risks at multiple decision points.
• Offers guidance for future research to ensure causal estimation approaches align with sequential intervention decision needs.