Predicting Caregiver Communications in a Geriatric Clinic

Area of Science:

• Geriatric Medicine
• Artificial Intelligence in Healthcare
• Health Informatics

Background:

• Geriatric clinics face challenges in managing patient communication needs.
• Predicting communication requirements is crucial for efficient patient care.
• Machine learning offers potential for identifying key predictive factors.

Purpose of the Study:

• To evaluate a machine learning model for predicting geriatric clinic communication needs.
• To identify medical record variables that predict communication requirements.
• To assess the utility of patient and caregiver data in this prediction.

Main Methods:

• Extracted data from 557 patient records, including behavioral symptoms, cognition, medical history, and caregiver assessments.
• Utilized random forest models to predict incoming caregiver contacts, outgoing clinic contacts, and clinic communications.
• Employed permutation importance to identify significant predictive variables.

Main Results:

• Models explained 7.42% (incoming), 3.65% (outgoing), and 6.23% (training) of variance in communication needs.
• Out-of-sample test sets showed explained variance of 6.17% (incoming), 2.78% (outgoing), and 4.28% (clinic) communications.
• Key predictors identified: patient neuropsychiatric symptoms, global cognition, body mass index, caregiver burden, and patient/caregiver age.

Conclusions:

• Patient neuropsychiatric symptoms, caregiver burden, age, body mass index, and global cognition are potential predictors of communication needs in geriatric clinics.
• These findings can inform proactive patient care strategies.
• Further research should explore additional caregiver variables and modifiable factors.