AmarDoctor: An AI-Driven, Multilingual, Voice-Interactive Digital Health Application for Primary Care Triage and Patient Management to Bridge the Digital Health Divide for Bengali Speakers

TL;DR

AmarDoctor tackles the digital health divide for Bengali speakers by combining a Health Knowledge Graph with a multilingual, voice-enabled assistant and AI-driven symptom reporting to deliver dual-output clinical guidance for patients and clinicians. The system uses 1.4 million records to construct a comprehensive knowledge graph, supports 908 symptoms across Bengali dialects, and employs adaptive questioning to generate provisional diagnoses and specialist referrals, presented as SOAP notes for clinicians. In an independent evaluation of 185 clinical vignettes reviewed by five external physicians, AmarDoctor achieved $M1=81.08 ext{} ext{%}$ and $M3=87.57 ext{} ext{%}$ for provisional diagnoses and $M1=91.35 ext{} ext{%}$ for specialization, outperforming physician averages and illustrating the value of machine-in-the-loop decision support in resource-constrained settings. The work demonstrates a practical path toward inclusive AI in healthcare, with strong potential to scale to other languages and contexts while highlighting ongoing challenges around data scope, language coverage, and deployment ethics.

Abstract

This study presents AmarDoctor, a multilingual voice-interactive digital health app designed to provide comprehensive patient triage and AI-driven clinical decision support for Bengali speakers, a population largely underserved in access to digital healthcare. AmarDoctor adopts a data-driven approach to strengthen primary care delivery and enable personalized health management. While platforms such as AdaHealth, WebMD, Symptomate, and K-Health have become popular in recent years, they mainly serve European demographics and languages. AmarDoctor addresses this gap with a dual-interface system for both patients and healthcare providers, supporting three major Bengali dialects. At its core, the patient module uses an adaptive questioning algorithm to assess symptoms and guide users toward the appropriate specialist. To overcome digital literacy barriers, it integrates a voice-interactive AI assistant that navigates users through the app services. Complementing this, the clinician-facing interface incorporates AI-powered decision support that enhances workflow efficiency by generating structured provisional diagnoses and treatment recommendations. These outputs inform key services such as e-prescriptions, video consultations, and medical record management. To validate clinical accuracy, the system was evaluated against a gold-standard set of 185 clinical vignettes developed by experienced physicians. Effectiveness was further assessed by comparing AmarDoctor performance with five independent physicians using the same vignette set. Results showed AmarDoctor achieved a top-1 diagnostic precision of 81.08 percent (versus physicians average of 50.27 percent) and a top specialty recommendation precision of 91.35 percent (versus physicians average of 62.6 percent).

Figures (9)

• Figure 1: AmarDoctor's patient-centered medical triage and primary healthcare delivery. The workflow begins with a multilingual, voice-interactive chatbot that classifies a patient's intent to guide them to the appropriate service. For users reporting symptoms to book a doctor, the system initiates a six-stage process: (1) capturing chief complaints via text or voice input; (2) normalizing medical terminology using the MedAi Health Knowledge Graph; (3) suggesting related symptoms to the patient based on their initial input; (4) conducting an interactive diagnostic assessment with relevant medical questions leveraging the knowledge-graph; (5) generating specialist doctor suggestions for the patient and provisional clinical diagnoses for the physician ; and (6) facilitating virtual consultation booking, video consultation from a doctor and finally prescription delivery after the consultation.
• Figure 2: Illustration of intelligent symptom reporting and triage flow: (A) ASR-enabled voice-interactive symptom reporting interface; (B) and (C) show symptom suggestion and dynamic question flow to capture patient's symptoms in English and Bengali.
• Figure 3: Overview of the final recommendations received by physicians and patients: (A) displays two screens, illustrating structured SOAP note comprising patient's physiological and medical data and provisional recommendation with associated confidence score received by physician (B) displaying specialization recommendation with confidence score visualized in patient's side.
• Figure 4: Comprehensive symptom mapping workflow: curation of 908 English symptoms through medical team collaboration; followed by translation into standard Bengali, local variations, and Chittagonian and Sylheti dialects to capture linguistic diversity and ensure precise medical communication
• Figure 5: Visualizations of symptom distribution across diseases: a, illustrates the overall distribution, while b highlights the top five most frequently occurring symptoms for each disease.