Effectiveness of UrDiaCare

This study evaluates UrDiaCare, an Urdu-language, knowledge-grounded conversational assistant developed to support diabetes self-management among adults with type 2 diabetes in Pakistan. The intervention is designed to complement routine diabetes care by providing accessible, culturally appropriate education and ongoing support through WhatsApp.

UrDiaCare focuses on five areas of diabetes self-management: medication adherence, healthy eating, physical activity, blood glucose monitoring, and prevention of diabetes-related complications. Its development includes translating and culturally adapting diabetes education materials, expert review, and pilot testing with adults with type 2 diabetes to assess clarity, usability, cultural appropriateness, and technical functionality before finalizing the intervention.

The intervention is guided by Social Cognitive Theory (SCT) and the Health Belief Model (HBM). SCT emphasizes self-efficacy and supports mastery experiences, encouragement, and contextualized feedback. HBM informs messages addressing perceived susceptibility, severity, benefits, and barriers related to diabetes self-management. The conceptual framework proposes that UrDiaCare may improve self-efficacy and health beliefs, leading to better medication adherence and self-care behaviors and, ultimately, improved glycemic control.

UrDiaCare uses a knowledge-grounded, retrieval-augmented approach. Its responses are restricted to a predefined, clinician-reviewed knowledge library based on established diabetes education resources. The system is not a diagnostic or prescribing tool and does not provide individualized medical recommendations. Questions outside its approved educational scope trigger a standard response directing participants to their treating physician or an appropriate health facility. The knowledge library and safety rules are version-locked during the trial to maintain consistency.

The trial will use a parallel-group randomized controlled design. Participants will receive either UrDiaCare in addition to standard outpatient diabetes care or standard care with a static Urdu educational pamphlet covering the same five self-management areas. The study will follow participants for four months, with assessments conducted at baseline, two months, and four months.

The primary outcome is change in HbA1c from baseline to four months. Secondary outcomes assess medication adherence, diabetes self-management behavior, diabetes self-efficacy, self-care behaviors, and the usability and acceptability of UrDiaCare. The study will also examine factors associated with glycemic improvement among participants receiving UrDiaCare.

Intervention engagement and technical performance will be monitored throughout the study. UrDiaCare will use predefined safety checks to identify emergency or out-of-scope requests, while system information will be used to assess engagement and intervention delivery. Participant confidentiality will be protected through secure data-handling procedures, with identifying information and communication records stored separately and access restricted to authorized study personnel.

Overall, the study will determine whether adding a theory-informed, Urdu-language, knowledge-grounded conversational intervention to standard diabetes care can improve glycemic control and diabetes self-management. It will also provide evidence on the usability and safe implementation of an AI-assisted approach to diabetes self-management in Pakistan.

Inclusion criteria

Adults aged 30-70 years with T2DM confirmed ≥ 6 months (clinical record) and HbA1c ≥ 7.5% at screening

• Currently prescribed oral antidiabetic medication(s) for ≥ 3 months
• Personal ownership and regular daily use of a functioning smartphone (Android/iOS)
• Able to read and understand Urdu
• Available for the full 4-month study period and willing to provide written informed consent -

Exclusion criteria

Type 1, gestational, or secondary diabetes (e.g., steroid-induced, MODY)

• Current insulin therapy (to avoid confounding from differing self-management complexity)
• Severe, uncorrectable comorbidity likely to materially affect glycemic control or survival within 4 months (e.g., end-stage renal disease, active malignancy, liver failure)
• Significant cognitive impairment or psychiatric disorder precluding informed consent or technology use
• Participation in another diabetes clinical trial within the preceding 3 months
• Insufficient Urdu reading proficiency, as the UrDiaCare interface is Urdu-script based