Long-Term Efficacy of an AI-Based Health Coaching Mobile App in Slowing the Progression of Nondialysis-Dependent Chronic Kidney Disease: Retrospective Cohort Study

Background Chronic kidney disease (CKD) is a significant public health concern. Therefore, practical strategies for slowing CKD progression and improving patient outcomes are imperative. There is limited evidence to substantiate the efficacy of mobile app–based nursing systems for decelerating CKD progression. Objective This study aimed to evaluate the long-term efficacy of the KidneyOnline intelligent care system in slowing the progression of nondialysis-dependent CKD. Methods In this retrospective study, the KidneyOnline app was used for patients with CKD in China who were registered between January 2017 and April 2023. Patients were divided into 2 groups: an intervention group using the app’s nurse-led, patient-oriented management system and a conventional care group that did not use the app. Patients’ uploaded health data were processed via deep learning optical character recognition, and the artificial intelligence (AI) system provided personalized health care plans and interventions. Conversely, the conventional care group received suggestions from nephrologists during regular visits without AI. Monitoring extended for an average duration of 2.1 (SD 1.4) years. The study’s objective is to assess the app’s effectiveness in preserving kidney function. The primary outcome was the estimated glomerular filtration rate slope over the follow-up period, and secondary outcomes included changes in albumin-to-creatinine ratio (ACR) and mean arterial pressure. Results A total of 12,297 eligible patients were enrolled for the analysis. Among them, 808 patients were successfully matched using 1:1 propensity score matching, resulting in 404 (50%) patients in the KidneyOnline care system group and another 404 (50%) patients in the conventional care group. The estimated glomerular filtration rate slope in the KidneyOnline care group was significantly lower than that in the conventional care group (odds ratio –1.3, 95% CI –2.4 to –0.1 mL/min/1.73 m2 per year vs odds ratio –2.8, 95% CI –3.8 to –1.9 mL/min/1.73 m2 per year; P=.009). Subgroup analysis revealed that the effect of the KidneyOnline care group was more significant in male patients, patients older than 45 years, and patients with worse baseline kidney function, higher blood pressure, and heavier proteinuria. After 3 and 6 months, the mean arterial pressure in the KidneyOnline care group decreased to 85.6 (SD 9.2) and 83.6 (SD 10.5) mm Hg, respectively, compared to 94.9 (SD 10.6) and 95.2 (SD 11.6) mm Hg in the conventional care group (P<.001). The ACR in the KidneyOnline care group showed a more significant reduction after 3 and 6 months (736 vs 980 mg/g and 572 vs 840 mg/g; P=.07 and P=.03); however, there was no significant difference in ACR between the two groups at the end of the follow-up period (618 vs 639 mg/g; P=.90). Conclusions The utilization of KidneyOnline, an AI-based, nurse-led, patient-centered care system, may be beneficial in slowing the progression of nondialysis-dependent CKD.