The ability to provide trustworthy maternal health information using phone-based chatbots can have a significant impact, particularly in low-resource settings where users have low health literacy and limited access to care. However, deploying such systems is technically challenging: user queries are short, underspecified, and code-mixed across languages, answers require regional context-specific grounding, and partial or missing symptom context makes safe routing decisions difficult. We present a chatbot for maternal health in India developed through a partnership between academic researchers, a health tech company, a public health nonprofit, and a hospital. The system combines (1) stage-aware triage, routing high-risk queries to expert templates, (2) hybrid retrieval over curated maternal/newborn guidelines, and (3) evidence-conditioned generation from an LLM. Our core contribution is an evaluation workflow for high-stakes deployment under limited expert supervision. Targeting both component-level and end-to-end testing, we introduce: (i) a labeled triage benchmark (N=150) achieving 86.7
Context: A poor-quality diet with low diversity is central to the high prevalence of malnutrition among young Indian children. The Indian Integrated Child Development Services Scheme (ICDS) addresses this nutritional gap through its Supplementary Nutrition Program (SNP), which provides energy, protein, fat, and essential micronutrients to children via Take-Home Rations (THR, for children aged 6-36 months) and Hot Cooked Meals (HCM, for children aged 37-72 months). Aims: To examine the diet diversity of THR and HCM provided under ICDS-SNP across Indian states and identify strategies for improvement. Settings and Design: A pan-India study analysing 12 diverse food groups in the ICDS-SNP: cereals and millets, pulses and legumes, green leafy vegetables, other vegetables, roots and tubers, nuts and seeds, edible oils and fat, eggs, milk and milk-based products, other non-vegetarian foods, fruits, and sugar across states. Methods and Material: Diet diversity scores for THR and HCM were calculated based on the provision of these food groups. Regional variations and inclusion of locally available foods were examined. Statistical analysis used: Descriptive analysis was conducted to assess diet diversity scores across states. Regional patterns and variations in the SNP food composition were analysed. Results: The study revealed significant variability in diet diversity across states, with scores ranging from 3-11. Southern and Eastern states showed higher diversity, incorporating local fruits, vegetables, and animal-source foods like milk and eggs. Both THR and HCM were predominantly cereal based. Findings suggest that achieving diet diversity in ICDS-SNP is feasible but lagging states need tailored, geographically specific strategies.
Type 2 Diabetes (T2D) is a major public health concern in India, with suboptimal outcomes despite established guidelines and improved healthcare facilities. Understanding the barriers and facilitators influencing diabetes care and self-management is essential for effective context- specific intervention design. We conducted semi-structured, in-depth interviews in July 2023 with 18 adults (aged 35–65) with T2D from low, middle, and high socioeconomic groups in Delhi, India, using purposive sampling. The interview guide explored demographics, health awareness, and barriers and facilitators to behavior change and healthcare use. Interviews were audio-recorded, transcribed verbatim, and analyzed using thematic analysis (Braun and Clarke). Framework development was informed by grounded theory principles. Rigor was ensured through independent coding, triangulation, and adherence to COREQ guidelines. Three interrelated themes emerged: (1) personal attributes (self-experience and social influences), (2) healthcare access (availability, affordability and provider interaction) and (3) enabling factors (financial resources, knowledge, and education). Key barriers included poor risk perception, fear of medication dependence, inconsistent provider communication, and out-of-pocket costs compounded by overcrowded facilities and limited access to reliable information. Key facilitators included family support, positive provider engagement, and access to free/subsidized medicines, tests, and counseling. These findings were integrated into the ‘Determinants of Diabetes Care and its Management (DDM)’ framework, which highlights the complex interplay of individual, socio-cultural, and health system factors influencing care-seeking and adherence. This study provides a comprehensive understanding of the interconnected individual, social and systemic determinants affecting diabetes care in India. The proposed DDM framework provides actionable insights for designing patient-centered interventions, strengthening provider communication and improving equitable access to care, particularly in resource-constrained settings, with implications for program implementation and policy. Trial registration CTRI/2023/10/058452 dated 7/July/2023.
Maternal nutritional status is the key determinant of a newborn’s nutritional health at birth. Therefore, the present study aimed to examine the correlation between maternal anthropometric measurements and neonatal nutritional status. This was a longitudinal study of mother–infant dyads in a slum of West Delhi. Maternal data and anthropometric measurements (height, weight, and MUAC) were collected. Pregnant women were followed until delivery, and nutritional assessment of neonates was conducted within 72 h of birth using anthropometric measurements (weight, length, head circumference, and MUAC). Pregnant women (n = 178) having gestational age ≥ 28 weeks and their neonates (n = 133). Nutritional assessment of pregnant women revealed that 19.4
Background:Assam, India, has the country's highest maternal mortality ratio (195 per 100,000 live births), mainly due to poor access to and quality of maternal health (MH) care. Many women receive inadequate antenatal and postnatal services, made worse by isolation, socioeconomic barriers, and weak health care infrastructure. Digital tools like mobile messaging and chatbots have improved antenatal care (ANC) and facility-based deliveries in similar settings. The e-SAATHI (Strengthening ANC/PNC via AskNivi Tailored Health Information, Referrals, and Follow-Up) project aims to provide personalized, stage-specific maternal health support through a chat-based system in Assam. Objective:This study assesses the acceptability, feasibility, and effectiveness of the e-SAATHI chatbot in increasing women's access to MH information and improving ANC and PNC service uptake across public and private facilities. Objectives include increasing ANC/PNC use (eg, ≥4 ANC visits and timely PNC), promoting respectful care, and gathering insights for scaling digital health in high-burden regions. Methods:Phase 1 (0-3 mo) involves co-designing and pilot testing aligned with World Health Organization and national guidelines. Phase 2 (4-24 mo) involves enrolling pregnant and postpartum women via health facilities and social media. The chatbot sends 2-3 messages weekly from 10-week pregnancy to 15 weeks postpartum. About 300 health care providers will be trained and engaged for onboarding and feedback. Phase 3 (25-36 mo) involves scaling up across districts, reaching 225,000 women. Data collection includes interviews, surveys, facility assessments, and chatbot analytics. Qualitative analysis will explore experiences; quantitative data (ANC completion, facility delivery, PNC follow-up, and satisfaction) will compare pre- and post-interventions. Ethical approvals, informed consent, and data confidentiality are observed. Results:The study was funded in September 2022. As of August 2025, 210 facilities have been onboarded, and 201,813 women were enrolled. Chatbot-based data collection began in April 2023 and will continue through the study period. Qualitative and quantitative evaluation data collection started in November 2023 and is expected to complete in June 2027. Interim analyses will be conducted after midline data collection in 2026; final analyses will be performed after endline data collection in 2027. The primary outcome will be the change in the composite quality score of maternal and newborn care. Secondary outcomes will include service uptake indicators, user-reported knowledge and self-care practices, and satisfaction with care. Operational feasibility-including provider integration and barriers such as digital literacy and connectivity-will also be assessed. Ongoing collaborative learning and adapting cycles are expected to capture intervention adaptations and inform optimal strategies for scale-up. Conclusions:e-SAATHI offers a scalable digital approach to improve MH across a variety of sociodemographic, linguistic, and risk settings. By delivering timely, personalized support, the chatbot may enhance health-seeking behavior and outcomes in Assam and in similar low-resource areas globally.
