Development and Validation of Dynamic Prediction Model for Thromboembolic and Bleeding Risk in Patients With Atrial...

Area of Science:

• Cardiology
• Pharmacology
• Medical Informatics

Background:

• Existing risk assessment tools for atrial fibrillation (AF) patients primarily focus on warfarin users and initial treatment decisions.
• There is a lack of dynamic models for reassessing thromboembolic and bleeding risks in patients using direct oral anticoagulants (DOACs).

Purpose of the Study:

• To develop a novel tool for the dynamic evaluation of thromboembolic and bleeding risks in AF patients receiving DOAC therapy.
• To create models that allow for regular risk updates based on a patient's current clinical status.

Main Methods:

• A landmarking approach was applied to a national claims database of 42,450 AF patients newly prescribed DOACs.
• Dynamic prediction models were developed using time-dependent variables and the Fine-Gray subdistribution hazard model to account for competing risks like death.
• The study included development (2018) and temporal-validation (2019) cohorts.

Main Results:

• Dynamic models were created to estimate 1-year thromboembolic and major bleeding risks over a 2-year period post-DOAC initiation, incorporating factors like DOAC adherence.
• The thromboembolic risk model (concordance index: 0.715) and major bleeding risk model (concordance index: 0.697) outperformed established scores like CHA2DS2-VASc and HAS-BLED.
• Calibration plots demonstrated good agreement between predicted and observed risks, with nomograms developed for clinical utility.

Conclusions:

• The developed dynamic prediction models provide a superior and novel method for reassessing thromboembolic and bleeding risks in AF patients on DOACs.
• These models facilitate personalized and dynamic risk management by enabling regular updates based on evolving patient status.
• The findings support improved clinical decision-making for AF patients on DOAC therapy.