JMIR Publications

ABSTRACT

Background:

Chronic obstructive pulmonary disease (COPD) requires continuous self-management, particularly during the transition from hospital to home, when follow-up support may be fragmented. Artificial intelligence–enabled mobile health interventions may extend individualized support beyond hospitalization, but evidence regarding their clinical effects in patients with COPD remains limited.

Objective:

This pilot randomized controlled trial aimed to evaluate the preliminary effects of the Artificial Intelligence Health Education Accurately Linking System (AI-HEALS), a supervised, knowledge-guided intervention delivered through WeChat, on COPD-related health status, cough-related burden, and self-management–related outcomes during the hospital-to-home transition.

Methods:

We conducted a single-center, parallel-group, open-label pilot randomized controlled trial with blinded outcome assessment and data analysis. Fifty hospitalized adults with COPD were randomly assigned in a 1:1 ratio to receive usual care alone or usual care plus AI-HEALS for 3 months. AI-HEALS integrated knowledge-guided question answering, self-monitoring and weekly check-ins, personalized reminders, and tailored educational content, with daily review of interaction logs by research staff. The co-primary outcomes were COPD-related health status, assessed using the COPD Assessment Test (CAT) and the Cough Evaluation Test (CET). Secondary outcomes included dyspnea, psychological distress, perceived stress, perceived social support, self-efficacy, medication-taking behavior, dietary self-regulation, and health-related quality of life. Outcomes were assessed at baseline, discharge, and 3-month follow-up. Linear mixed-effects models and analysis of covariance were used, and analyses followed the intention-to-treat principle.

Results:

All 50 participants were randomized, with 25 assigned to each group, and no participants were lost to follow-up for the co-primary outcomes. At 3 months, CAT scores were lower in the intervention group in the unadjusted analysis (mean difference −4.44, 95% CI −8.04 to −0.84; P=.01). The covariate-adjusted estimate remained favorable but uncertain (mean difference −3.42, 95% CI −7.06 to 0.22; P=.06). CET scores also favored the intervention group, but the differences were not statistically significant in either the unadjusted analysis (mean difference −2.32, 95% CI −4.90 to 0.26; P=.07) or the adjusted analysis (mean difference −2.35, 95% CI −5.02 to 0.33; P=.08). In adjusted analyses, perceived stress was lower at 3 months (mean difference −2.03, 95% CI −3.86 to −0.21; P=.02), and self-efficacy was higher at discharge (mean difference 1.57, 95% CI 0.38 to 2.76; P=.01). No clear adjusted between-group differences were observed for the other secondary outcomes. The intervention group generated 991 chatbot interactions during the study period.

Conclusions:

AI-HEALS was associated with a greater unadjusted reduction in CAT scores at 3 months, although the adjusted estimate remained uncertain. CET scores showed a favorable but nonsignificant pattern. Preliminary signals were also observed for perceived stress and self-efficacy. These hypothesis-generating findings support further evaluation of supervised, knowledge-guided AI–enabled COPD self-management support in a fully powered, multicenter trial with longer follow-up and objective clinical outcomes. Clinical Trial: Chinese Clinical Trial Registry ChiCTR2400092829.