Effectiveness of Virtual Health Coaching Based on Artificial Intelligence to Improve Glycemic Control in Patients With Type 2 Diabetes Mellitus

Type 2 diabetes mellitus (T2DM) is a chronic metabolic disorder characterized by elevated blood glucose levels due to impaired insulin production or insulin resistance. Globally, the prevalence of T2DM continues to rise, and in Indonesia the number of cases has surpassed 10 million, placing the country among the top seven worldwide for diabetes prevalence. Poor glycemic control leads to serious complications such as kidney disease, cardiovascular disorders, and neuropathy, which reduce patients' quality of life and increase health care costs. One of the main challenges in diabetes management is the low level of patient adherence to recommended lifestyle changes, including diet, physical activity, and regular monitoring. Patients often struggle to maintain long-term self-management because of limited access to continuous education, motivation, and personalized support in primary health care settings.

Digital health innovations, including virtual health coaching, have emerged as promising solutions to address these challenges. Virtual health coaching provides structured, ongoing support by using digital platforms to deliver education, reminders, and behavioral feedback. The integration of artificial intelligence (AI) into virtual coaching systems allows the intervention to be more adaptive and personalized, responding to the unique behaviors and needs of each participant. The DIACOACH application has been specifically developed to support diabetes self-management in Indonesia. It offers adaptive education modules, lifestyle modification reminders, diet and exercise monitoring, medication tracking, and a chatbot interface for personalized interaction. Early community service activities using DIACOACH have shown improvements in patient knowledge, motivation, and preliminary indicators of glycemic control such as fasting blood glucose and HbA1c levels. However, systematic research is required to evaluate its effectiveness in a controlled study, particularly in primary care contexts with limited resources.

This study is designed as a quasi-experimental trial with a non-equivalent control group design. Participants will be adults aged 30 to 56 years who have been diagnosed with T2DM for more than three months, are clinically stable, able to read, and have access to a smartphone or digital device. Individuals with type 1 diabetes, severe complications, ongoing steroid therapy, or those enrolled in other structured interventions will be excluded. Eligible participants will be recruited from primary health care centers (Puskesmas) in Jeneponto and Bantaeng, Indonesia. After baseline assessments, participants will be allocated to either the intervention group or the control group.

The intervention group will use the DIACOACH application for twelve weeks. Through this platform, participants will receive adaptive education tailored to their daily routines, reminders about diet and physical activity, and continuous encouragement to maintain self-care practices. The app integrates principles of Social Cognitive Theory and Self-Determination Theory to build self-efficacy and intrinsic motivation. The AI-based system provides feedback based on real-time data entered by participants, such as blood glucose values, diet logs, exercise activities, and medication adherence. The control group will continue with standard diabetes education typically offered at health centers, which usually consists of routine counseling sessions and printed health information.

Primary outcomes will include fasting blood glucose (FBG), HbA1c, and random blood glucose (RBG), measured at baseline and after the twelve-week intervention. Secondary outcomes will assess changes in dietary behavior, physical activity, adherence to self-care, satisfaction with the intervention, and overall quality of life. Data will be collected through laboratory tests, validated questionnaires, and digital records from the application. Quantitative data will be analyzed using appropriate statistical tests such as paired t-tests and analysis of variance to evaluate differences between groups and across time points. In addition, a qualitative component will be conducted by interviewing selected participants from the intervention group to explore their perceptions, experiences, barriers, and facilitators in using AI-based virtual coaching for diabetes self-management.