Paper Number
ICIS2026-2532
Paper Type
Short
Abstract
Internet hospitals increasingly serve as an entry point into integrated online-to-offline care pathways, but hospitals often lack timely tools to anticipate subsequent offline demand. Drawing on Protection Motivation Theory, this study develops PACC-Net, a theory-guided deep learning framework that predicts whether patients will seek offline outpatient care within 14 days after an online consultation. The model separately represents threat appraisal and coping appraisal by integrating consultation dialogues, patients’ prior healthcare utilization, and physician credential signals. Using linked online consultation and offline outpatient records from a large public hospital in China, we show that PACC-Net outperforms traditional machine learning, neural dialogue, and LLM-based baselines across multiple metrics. The findings demonstrate the value of appraisal-aware dialogue modeling for predicting short-term offline care-seeking and supporting proactive outpatient resource coordination.
Recommended Citation
Zhou, Yang; Li, Juan; Feng, Jiahui; and Liu, Hefu, "Predicting Online-to-Offline Care Conversion in Internet Hospitals: A Theory-Guided Deep Learning Approach" (2026). ICIS 2026 Proceedings. 15.
https://aisel.aisnet.org/icis2026/health/health/15
Predicting Online-to-Offline Care Conversion in Internet Hospitals: A Theory-Guided Deep Learning Approach
Internet hospitals increasingly serve as an entry point into integrated online-to-offline care pathways, but hospitals often lack timely tools to anticipate subsequent offline demand. Drawing on Protection Motivation Theory, this study develops PACC-Net, a theory-guided deep learning framework that predicts whether patients will seek offline outpatient care within 14 days after an online consultation. The model separately represents threat appraisal and coping appraisal by integrating consultation dialogues, patients’ prior healthcare utilization, and physician credential signals. Using linked online consultation and offline outpatient records from a large public hospital in China, we show that PACC-Net outperforms traditional machine learning, neural dialogue, and LLM-based baselines across multiple metrics. The findings demonstrate the value of appraisal-aware dialogue modeling for predicting short-term offline care-seeking and supporting proactive outpatient resource coordination.
Comments
19_Healthcare_and_wellbeing