Expanding care coordination in an integrated health system through causal machine learning

Area of Science:

• Health Services Research
• Artificial Intelligence in Healthcare
• Clinical Informatics

Background:

• Hospital readmission is a critical healthcare quality indicator.
• Post-discharge interventions show inconsistent effectiveness in reducing readmissions.
• Optimizing patient selection for interventions is essential for resource allocation.

Purpose of the Study:

• To evaluate the effectiveness of a novel causal machine learning model, the Predicted Benefit Intervention (PBI) score, in identifying low-risk patients who would benefit from post-discharge care coordination.
• To assess the feasibility of implementing causal machine learning at scale within a large health system.

Main Methods:

• A large-scale randomized trial involving 9959 low-risk patients across 19 Kaiser Permanente Northern California (KPNC) hospitals from May to December 2022.
• Patients were randomized to either usual care or the Transitions Program, featuring medication reconciliation, appointment scheduling, and 30-day follow-up calls.
• The PBI score was used to guide the targeting of post-discharge care coordination.

Main Results:

• The 30-day hospital readmission rate showed a non-significant decline in the intervention group (7.7% vs. 8.2%).
• A statistically significant decline in the observed-to-expected readmission ratio was observed post-randomization in the intervention group.
• The study demonstrated the feasibility of large-scale implementation of causal machine learning for targeted care delivery.

Conclusions:

• Causal machine learning, exemplified by the PBI score, offers a feasible approach to enhance targeting and resource allocation for post-discharge care coordination.
• While direct readmission rates did not significantly decrease, the model's impact on the readmission ratio suggests potential benefits in optimizing intervention delivery.
• This study highlights the potential of AI in improving healthcare quality metrics and operational efficiency.