Integrating Causal Inference and Agent-Based Modelling to Assess the Impact of Clinicians' Guideline Adherence in Older Adults Hospitalized with Pneumonia.

Metadata-grounded summary

Citation abstract

This study integrates agent-based modeling (ABM) and causal machine learning (ML) to assess the impact of clinicians' adherence to antibiotic guidelines in older adults hospitalized with community-acquired pneumonia (CAP). Using a synthetic population and longitudinal data, we estimate individual treatment effects of vancomycin via the longitudinal modified treatment policy (LMTP) algorithm and simulate clinical decision-making and MRSA transmission dynamics in a NetLogo-based ABM. We evaluate how varying hospital-level propensities toward vancomycin prescribing influence 30-day mortality under different causal and behavioral scenarios. Results show high model repeatability and confirm increased mortality among mechanically ventilated patients treated empirically with vancomycin. Integrating causal inference within ABM supports evidence-based antibiotic stewardship and privacy-preserving simulation of clinical behaviors.