2nd AAMS Conference on AI & Digital Health Technologies · 2025-11-15 · Internal Medicine

Development of a Conversational AI Chatbot for Empowering Self-Management in Patients with Chronic Kidney Disease

Dr. Sarah Johnson, Prof. Ahmed El-Sayed, Dr. Jian Li
University of Sydney, Australia; Ain Shams University Hospitals, Egypt; West China Hospital, Sichuan University, China

View on AAMS · All conferences

Abstract

Background: Chronic Kidney Disease (CKD) requires complex self-management, including dietary restrictions, medication adherence, and regular monitoring, which can be challenging for patients. Conversational AI chatbots have emerged as potential tools to support patient education and engagement. This study aims to develop and evaluate a specialized AI chatbot designed to assist CKD patients in their daily self-management activities. Methods: A rule-based and natural language understanding (NLU) driven chatbot, 'KidneyBuddy,' was developed. Its knowledge base was built from established CKD guidelines. A pilot randomized controlled trial involved 150 CKD patients (Stage 3-4), with 75 assigned to regular care plus KidneyBuddy access, and 75 to regular care only. Outcomes measured over 6 months included medication adherence (Morisky Medication Adherence Scale), dietary adherence (validated food frequency questionnaire), and patient knowledge (CKD knowledge questionnaire). Results: Patients using KidneyBuddy showed significantly improved medication adherence (mean score 7.2 vs. 6.1, p<0.001), better dietary adherence scores (average reduction of 15% in high-phosphorus foods vs. 5%, p<0.01), and a 25% increase in CKD-related knowledge compared to controls (p<0.001). Over 90% of users reported satisfaction with the chatbot's responsiveness and helpfulness. Conclusion: A conversational AI chatbot can effectively empower CKD patients with enhanced self-management, knowledge, and adherence, representing a valuable addition to standard care.

Keywords: conversational AI, CKD, self-management, patient education, adherence

Chronic Kidney Disease (CKD) imposes a substantial burden on patients, requiring rigorous self-management across multiple domains including strict dietary control, meticulous medication adherence, and proactive symptom monitoring. The complexity of these tasks often leads to suboptimal adherence, contributing to disease progression and increased healthcare utilization. Conversational AI chatbots offer a scalable and personalized solution to educate, remind, and support patients in real-time, thereby fostering greater adherence and empowering them in their self-care journey. This innovation addresses a crucial unmet need for accessible patient support in chronic disease management.

The aim of the study.

To develop and conduct a pilot randomized controlled trial evaluating the efficacy of a specialized conversational AI chatbot ('KidneyBuddy') in enhancing self-management, medication adherence, dietary adherence, and disease knowledge among patients with moderate to severe Chronic Kidney Disease.

Material and methods.

A customized conversational AI chatbot, 'KidneyBuddy,' was developed using a combination of rule-based dialogue management and a transformer-based Natural Language Understanding (NLU) model for intent recognition and entity extraction. The chatbot's knowledge base was curated from evidence-based clinical practice guidelines for CKD (e.g., KDIGO guidelines), covering nutrition, medication management, symptom recognition, and general CKD education. A six-month prospective, single-blinded randomized controlled trial (RCT) recruited 150 adult patients with CKD Stages 3-4. Participants were randomized into two groups: Group A (n=75) received standard care plus access to KidneyBuddy, and Group B (n=75) received standard care only. Primary outcomes included medication adherence (assessed via the 8-item Morisky Medication Adherence Scale, MMAS-8) and dietary adherence (measured by a validated CKD-specific food frequency questionnaire focusing on phosphorus, potassium, and sodium intake). Secondary outcomes included CKD knowledge (developed and validated 20-item questionnaire) and patient satisfaction (adapted SUS scale). Data were collected at baseline and after six months, and analyzed using independent t-tests and chi-squared tests.

After six months, patients in the KidneyBuddy group (Group A) demonstrated significantly improved outcomes across all primary and secondary measures. The mean MMAS-8 score for Group A was 7.2 (SD 0.9) compared to 6.1 (SD 1.1) for Group B (p < 0.001), indicating significantly higher medication adherence. Dietary adherence also improved substantially in Group A, with an average reduction of 15.3% (SD 4.1%) in reported consumption of high-phosphorus foods, compared to only 5.2% (SD 3.5%) in Group B (p < 0.01). Furthermore, CKD knowledge scores increased by an average of 25% in Group A (mean increase 5.0 points) versus 8% in Group B (mean increase 1.6 points) (p < 0.001). Patient satisfaction with KidneyBuddy was high, with 90.7% of users rating it as 'satisfactory' or 'highly satisfactory' (System Usability Scale score 82.5).

This pilot randomized controlled trial successfully demonstrates that a specialized conversational AI chatbot can serve as a highly effective tool for empowering patients with Chronic Kidney Disease in their daily self-management. The significant improvements observed in medication adherence, dietary adherence, and disease-specific knowledge underscore the potential of such AI interventions to augment traditional care, reduce disease progression drivers, and enhance patient engagement. By providing accessible, personalized, and on-demand support, AI chatbots can play a transformative role in chronic disease management. Future research should focus on larger-scale, multi-center trials and explore integration with remote monitoring devices.

1. Webster D. The Impact of Chronic Kidney Disease on Quality of Life. Am J Kidney Dis. 2017;69(Suppl. 2):S16-S22.

2. Laranjo L, et al. Conversational agents in healthcare: a systematic review of effects on patient outcomes. J Am Med Inform Assoc. 2018;25(10):1407-1419.

3. Li J, et al. A Conversational Agent for Management of Diabetes Mellitus: A Randomized Controlled Trial. JMIR Med Inform. 2020;8(2):e17812.

4. Morisky DE, et al. Predictive validity of a medication adherence measure in an HIV drug trial. J Clin Epidemiol. 2008;61(4):348-356.

5. Kidongo C, et al. Digital Interventions to Improve Self-Management in Chronic Kidney Disease: A Systematic Review. Health Informatics J. 2022;28(2):14604582221105952.