Spending Analysis of Machine Learning-Based Communication Nudges in Oncology

Area of Science:

• Oncology
• Health Economics
• Behavioral Science

Background:

• Serious illness conversations (SICs) can improve cancer patient outcomes and reduce aggressive end-of-life care.
• Behavioral economics interventions may increase SICs, but their effect on end-of-life spending is not well understood.

Purpose of the Study:

• To evaluate the impact of a behavioral intervention on end-of-life spending in high-risk cancer patients.
• To determine if a machine learning-based intervention to increase clinician-initiated SICs affects healthcare costs.

Main Methods:

• Secondary analysis of a stepped-wedge cluster randomized trial with 1187 high-risk cancer patients.
• Intervention group received a behavioral intervention to increase SICs; control group received standard care.
• Spending data (acute care, outpatient, therapy, long-term care, hospice) analyzed using two-part models for the last 180 days of life.

Main Results:

• The intervention group had lower mean daily end-of-life spending (-$75.33), resulting in $13,747 savings per decedent.
• Savings were driven by reduced spending on systemic therapy (-$44.59) and outpatient care (-$9.62).
• No significant differences in spending on acute care, long-term care, or hospice were observed.

Conclusions:

• A machine learning-based, behaviorally informed intervention effectively increased SICs and generated significant end-of-life cost savings in cancer patients.
• The savings were primarily attributed to decreased systemic therapy and outpatient care utilization.
• This intervention offers a promising strategy for improving care quality and reducing healthcare expenditures.