Using a Machine-Learning Algorithm to Identify Palliative Care Needs in a Primary Care Population: A Pilot Study

Area of Science:

• Health Informatics
• Palliative Care Medicine
• Machine Learning Applications

Background:

• Early palliative care improves end-of-life outcomes.
• Referrals to palliative care specialists are often delayed in primary care settings.

Purpose of the Study:

• To assess the impact of a machine-learning algorithm on the timeliness of palliative care referrals.
• To evaluate the integration of a machine-learning tool into primary care for identifying palliative care needs.

Main Methods:

• A pilot study utilized a machine-learning algorithm to evaluate electronic health records of primary care patients.
• Patients identified by the algorithm as having high palliative care needs were reviewed by specialists.
• A stepped-wedge randomization was used to send referral notifications to primary care providers (PCPs) for verified cases.

Main Results:

• The algorithm evaluated 127,080 patients, presenting 934 for review.
• The time to 0.1% of the population receiving a palliative care consultation was 60.9 days in the intervention arm versus 71.8 days in the control arm.
• A machine-learning approach demonstrated an 15% reduction in time to palliative care consultation.

Conclusions:

• A machine-learning algorithm was effectively integrated into a primary care practice to identify palliative care needs.
• The study demonstrated a potential to reduce delays in accessing palliative care.
• Further refinement of the workflow is necessary for optimal implementation.