Development of Deep Learning Models for AI-Enhanced Telemedicine in Nursing Home Care

Area of Science:

• Gerontology
• Health Informatics
• Artificial Intelligence in Healthcare

Background:

• Acute health events in institutionalized older adults frequently result in unnecessary hospitalizations.
• Telemedicine offers a solution for remote assessment and continuity of care in nursing homes.
• Effective remote decision-making is crucial for managing acute conditions in residential care facilities.

Purpose of the Study:

• To evaluate the outcomes of telemedicine-supported acute event management in older adults residing in care facilities.
• To develop and assess a deep learning model for classifying acute episodes and predicting hospital referrals.
• To enhance clinical decision-making and reduce avoidable hospital transfers.

Main Methods:

• A quasi-experimental study involving 5202 acute events managed via a 24/7 telemedicine system.
• Data preprocessing included imputation, normalization, encoding, and dimensionality reduction (Truncated SVD).
• A deep feedforward neural network was trained and optimized using resampling techniques (SMOTEENN) and cross-validation.

Main Results:

• Telemedicine facilitated the resolution of approximately 90% of acute events within the nursing home setting.
• The deep learning model achieved an AUC of 0.91 and accuracy of 0.88, outperforming traditional algorithms.
• The model demonstrated improved performance in predicting hospital referrals, with an F1-score of 0.63.

Conclusions:

• Telemedicine, enhanced by AI, provides a reliable strategy for accurate triage and reduced hospital transfers in nursing homes.
• AI-assisted telemedicine systems can optimize clinical decision-making and improve care continuity for institutionalized older adults.
• Integration of these systems into long-term care workflows is recommended to improve patient outcomes and healthcare efficiency.