Friday 31 January 2025
Session 1
13:00-13:15
AI-Powered Health Assistant for Conflict-Affected Regions in Myanmar: Leveraging Offline Technology to Enhance Healthcare Delivery
| Presenter | : | Zarni Lynn Kyaw |
| Abstract ID | : | A009 |
Access to quality healthcare in Myanmar's remote and conflict-affected regions is severely hampered by limited infrastructure, a scarcity of skilled professionals, and unreliable internet connectivity. Traditional clinical decision support systems (CDSS), while beneficial, are often impractical in these settings due to their reliance on stable internet access. This study explores the deployment of an offline, AI-powered health assistant utilizing a Retrieval-Augmented Generation (RAG) approach with the Meta Llama 3.1 Large Language Model (LLM) to address this challenge. The system, designed for affordability and offline functionality on single-board computing platforms, was fine-tuned with local medical guidelines to provide contextually relevant clinical recommendations. An initial pilot test with an English-only interface in a remote clinic demonstrated the system's feasibility and potential to support healthcare workers in the absence of internet connectivity. To enhance accessibility, real-time translation into Burmese, Shan, and Karen languages was integrated using the Google Translate API, with ongoing refinements in collaboration with native speakers. Results from the pilot indicated positive feedback from healthcare workers, who reported increased confidence in decision-making and improved quality of care. The offline capability was highly valued, ensuring uninterrupted access to clinical support. While preliminary data suggests improved patient outcomes, further quantitative analysis is warranted. This research underscores the potential of LLM-based RAG systems to overcome access barriers and empower healthcare providers in resource-constrained, conflict-affected regions. Future work will focus on large-scale evaluation, refining language models, and assessing the impact on patient health, thereby offering a scalable model for enhancing healthcare delivery in underserved communities through innovative technological solutions.
Poster Slot
A01
13:00-13:15
Sharing Artificial Intelligence (AI) Data in Healthcare—Challenges and Enablers in Low-and Middle-Income Countries (LMICs). A systematic review and case study.
| Presenter | : | Aprajita Kaushik |
| Abstract ID | : | A019 |
Health systems in low- and middle-income countries (LMICs) can greatly benefit from AI interventions in various use cases such as diagnostics, treatment and public health monitoring but face significant challenges in sharing data for developing and deploying AI in healthcare. This study aimed to identify barriers and enablers to data sharing for AI in healthcare in LMICs and to test the relevance of these in a local context. We identified 22 studies, mainly from Africa (n=12, 55%) and Asia (n=6, 27%). The most important data-sharing challenges were unreliable internet connectivity, lack of equipment, poor staff and management motivation, uneven resource distribution, and ethical concerns. Possible solutions included improving IT infrastructure, enhancing funding, introducing user-friendly software, and incentivising healthcare organisations and personnel to share data for AI-related tools. In Thailand, inconsistent data systems, limited staff time, low health data literacy, com-plex and unclear policies and cybersecurity issues were important data-sharing challenges. Key solutions included building a conducive digital ecosystem —having shared data input platforms for health facilities to ensure data uniformity and developing easy-to-understand consent forms, standardised guidelines for data sharing, and compensation policies for data breach victims. Though AI in LMICs has the potential to overcome health inequalities, these countries face technical, political, legal, policy, and organisational barriers to sharing data, which impede effective AI development and deployment. When tested in a local context, most of these barriers were relevant. Although our results might not be generalisable to other contexts, LMICs can use this study as a framework and leverage their individual health systems' barriers and strengths to devise local solutions for enhanced data sharing.
Poster Slot
B01
13:00-13:15
Artificial Intelligence Utilization in Cancer Screening Program across ASEAN: A Scoping review
| Presenter | : | Hein Minn Tun |
| Abstract ID | : | A021 |
Background: Cancer remains a significant health challenge in the ASEAN region, highlighting the need for effective screening programs. However, approaches, target demographics, and intervals vary across ASEAN member states, necessitating a comprehensive understanding of these variations to assess program effectiveness. Additionally, while artificial intelligence (AI) holds promise as a tool for cancer screening, its utilization in the ASEAN region is unexplored.
Purpose: This study aims to identify and evaluate different cancer screening programs across ASEAN and to assess the integration and impact of AI in these programs.
Methods: A scoping review was conducted using PRISMA-ScR guidelines to provide a comprehensive overview of cancer screening programs and AI usage across ASEAN. The search strategy involved searches through PubMed, Scopus, and Google Scholar with the inclusion criteria of only included studies that utilized data from the ASEAN region from January 2019 to May 2024.
Results: The findings reveal diverse cancer screening approaches in ASEAN. Countries like Myanmar, Laos, Cambodia, Vietnam, Brunei, Philippines, Indonesia and Timor-Leste primarily adopt opportunistic screening, while Singapore, Malaysia, and Thailand focus on organized programs. Cervical cancer screening is widespread, using both opportunistic and organized methods. Fourteen studies were included in the scoping review, covering breast (5 studies), cervical (2 studies), colon (4 studies), hepatic (1 study), lung (1 study), and oral (1 study) cancers. Studies revealed that different stages of AI integration for cancer screening: prospective clinical evaluation (50%), silent trial (36%) and exploratory model development (14%), with promising results in enhancing cancer screening accuracy and efficiency.
Conclusion: Cancer screening programs in the ASEAN region require more organized approaches targeting appropriate age groups at regular intervals to meet the WHO's 2030 screening targets. Efforts to integrate AI in Singapore, Malaysia, Vietnam, Thailand, and Indonesia show promise in optimizing screening processes, reducing costs, and improving early detection.
Poster Slot
C01
13:00-13:15
Readiness of managers and health care workers for e-Health: a cross-sectional study in Khartoum primary health care centers, Sudan
| Presenter | : | Ibrahim A. Atia |
| Abstract ID | : | A023 |
Background
e-Health is defined as “the use, in the health sector, of digital data— transmitted, stored and retrieved electronically—for clinical, educational and administrative purposes, both at the local site and at a distance”. In Primary health care (PHC), the role of e-Health in promoting PHC systems defines its need to achieve the PHC aims. This literary work aims to study the readiness of managers and healthcare workers for e-Health at Khartoum state PHC centers.
Methods
This was a facility-based cross-sectional study that took place between February and August 2022. A sample size of 327 was calculated, and multistage cluster sampling was used. A validated questionnaire was used, and the generated data were analyzed using the Statistical Package for Social Sciences (SPSS). Variables were described as n (%) and mean ± SD. Non-parametric tests and Spearman’s correlation were used to investigate the association of readiness scores with different categorical and numerical variables, respectively. A standard multiple regression model was used to model the associations.
Results
A total of 262 forms were completed. The overall readiness percentages were low for both managers (52.8%) and healthcare workers (55.3%). Factors associated with e-Health readiness included occupation, doctors’ level of expertise, PHC center, and PHC center type.
Conclusions
This study reports low levels of e-Health readiness as reflected by managers and HCWs in Khartoum State PHC. Project planners need to be aware of the obstacles and threats faced by e-Health initiatives if they are not carefully planned, executed, and monitored. Special attention must be given to addressing health inequities and inequalities to ensure that these projects will contribute to fostering accessibility to health services and narrowing the digital divide.
Poster Slot
D01
13:00-13:15
A Real-World Evaluation of an Innovative Artificial Intelligence Tool for Screening Breast Cancer at Population Level
| Presenter | : | Karthik Adapa |
| Abstract ID | : | A026 |
In resource-constrained countries like India, mammography-based breast screening is challenging to implement. This state-wide study, funded by the Government of Punjab, evaluated the use of Thermalytix, a low-cost, radiation-free AI tool, for breast cancer screening. Community health workers, trained to raise awareness, mobilized women aged 30 and above for screening. Thermalytix triaged women into five risk categories based on thermal images, with high-risk women recalled for diagnostic imaging. Over 18 months, 15,069 women were screened across 183 locations in Punjab. The median age was 41 years, and 69.9% were asymptomatic. Of 460 women testing positive (recall rate 3.1%), 268 underwent follow-up imaging, and 27 were confirmed with breast cancer, yielding a detection rate of 0.18%. The positive predictive value was 10.1%, and the median diagnostic interval was 21 days, with therapy initiation within 30 days. The study demonstrates the potential of Thermalytix for effective population-level breast cancer screening in low-resource settings.
Poster Slot
E01
Session 2
13:15-13:30
Evaluation and Stewardship of Artificial Intelligence Solutions for Health: Lessons Learned from USAID
| Presenter | : | Amit Chandra |
| Abstract ID | : | A030 |
Recent years have witnessed a proliferation of artificial intelligence (AI) solutions in the health sector. While AI solutions for health have the potential to generate novel insights for policymaking and support other health system goals, they also pose unique challenges including algorithmic biases and elevated cybersecurity risks. Ministries of Health, donor agencies, and other stakeholders must evaluate the potential benefits of AI solutions against their risks and pathway to scale. Drawing on lessons learned from case studies from India the authors present 5 overarching lessons learned. A framework of 7 questions serves to evaluate future potential AI for health applications.
Poster Slot
A02
13:15-13:30
Visual Narration in primary healthcare for COMPREHENSIV: A Multimodal Approach Using LLaVA and Whisper
| Presenter | : | Shreya Ramakrishnan |
| Abstract ID | : | A050 |
Subtheme- Technological Innovations to Strengthen Health Systems and Achieve Universal Health Coverage
Purpose: This study evaluates AI models' performance in enhancing primary healthcare systems and achieving Universal Health Coverage (UHC) through image-based Visual Question Answering (VQA) and local language translation tasks. It specifically focuses on understanding diseases such as leprosy and lymphatic filariasis and their associated environmental factors to improve patient education and healthcare delivery.
Methods: A total of nine AI models were assessed for their effectiveness in image-based Visual Question Answering (VQA): Gemini, Microsoft Phi-3, Llava v-1.6, CogVLM2, Qwen-VL-Plus, MiniGPT-4, Claude, Idefics2, and a custom Llava 1.5 + Whisper implementation, which features a Gradio interface with speech-to-text, image-to-text, and text-to-speech functions. IndicTrans2 and Indic-TTS from AI4Bharat were used for text and audio translation into local Indian languages.
Two tasks were designed:
1. Image-based VQA: Users can input an image along with audio or text queries to ask specific questions. Audio input benefits those who find typing challenging, streamlining interaction and enhancing user experience.
2. Translation: This task improves user’s understanding of healthcare content by allowing them to engage with the technology in their local language.
Results: AI models effectively support patient education and multimodal translation, benefiting both patients and providers. The Llava-v1.5 + Whisper implementation, Claude, Llava-v1.6-34b, and Microsoft Phi-3 demonstrated strong performance in VQA, while Gemini, CogVLM2, MiniGPT-4, Qwen-VL-Plus, and Idefics2 did not achieve satisfactory results. Additionally, IndicTrans2 and Indic-TTS successfully translated English text into 23 and 13 Indian languages, respectively, for reasonable usable text and audio outputs.
Conclusion: Multimodal LLMs show significant potential for enhancing primary healthcare and supporting UHC goals. However, varying performance in VQA highlights the importance of careful model selection. Exploring frameworks like RouteLLM could optimize AI model efficiency by routing tasks to the most suitable language models based on specific queries. This could reduce costs and improve model selection, ultimately enhancing healthcare delivery.
Poster Slot
B02
13:15-13:30
Study on private sector adoption of government-led digital health initiatives in India through a focused geographical approach of Microsite
| Presenter | : | Komal Malhotra |
| Abstract ID | : | A051 |
Ayushman Bharat Digital Mission (ABDM) envisions comprehensive digital health ecosystem in India by developing structures for integrated digital health. With citizen-centric focus, individuals are provided health account (ABHA). Verified registries are created for health professionals (HPR) and health facilities (HFR). Digital health record linking (HRL) enables consent-based sharing with desired stakeholders. National Health Authority (NHA) as implementing authority of ABDM, launched nation-wide initiative of 100 Microsites Project, to accelerate adoption in private sector. Microsite is a group of connected stakeholders, including providers, laboratories, and pharmacies within a geography. The project envisages active participation of state government, development partners (DP) and interfacing agencies (IFA) for its success.
Aims & Objectives: We aimed to study the implementation of Microsite project in Uttar Pradesh, India, to identify its critical success factors and assess the experiences of stakeholders in private sector adopting ABDM.
Methodology: A mixed-method study was utilized to analyze publicly available data as well as survey-based primary data from 7 healthcare facilities. 15 stakeholder interviews were conducted that included implementing authorities from DP, IFA and state government leadership.
Results: Successful adoption of ABDM observed in onboarding 1,285 health professionals and 2,385 health facilities against 750 and 600 targets respectively. HRL reached 68% target level. Motivation for ABDM adoption included compliance with government directives and monetary incentives. Technologically savvy doctors led adoption while government leadership spearheaded adoption strategies by close monitoring and evaluation. There were gaps in process awareness and satisfaction among private practitioners in Hospital/Laboratory Information Management system use. Technical issues, patient resistance, limited Information, Education and Communication material, and low awareness of ABDM and its benefits were noted to be challenges.
Conclusion: ABDM adoption in private health facilities has been successful through focused leadership initiatives. Addressing the identified challenges can potentially improve and sustain adoption process in private sector.
Poster Slot
C02
13:15-13:30
Bottlenecks of Leveraging Artificial Intelligence for Equitable Access in Rural Healthcare settings of Low and Middle-Income Countries
| Presenter | : | Manas Ranjan Behera |
| Abstract ID | : | A052 |
Introduction: Rural areas face inadequate healthcare services, including limited access to medical facilities and professionals, resulting in poorer health outcomes compared to urban populations. Artificial Intelligence (AI) presents a promising solution to bridge this gap by enhancing accessibility, efficiency and care quality.
Objectives: This comprehensive review aims to explore the bottlenecks of effective utilization of AI to address rural healthcare disparities.
Methodology: Four electronic international databases like PubMed, Scopus, Science Direct, and Google Scholar were searched to identify literatures related to health disparities and AI. Following a PRISMA guidelines, articles written in English and published between May 2013 and June 2024 was considered. Articles were eligible for inclusion if it identified at least one bottlenecks and a corresponding AI strategy to address it. Quality assessment and data extraction were conducted by two independent authors.
Results: A total of 613 articles were screened, and after removing duplication by Rayyan software, finally 12 articles were selected for full analysis. The findings were summarised under four key domains: (a) infrastructure development, (b) capacity building, (c) inclusive data quality and availability, and (d) regulatory support. Infrastructure development, especially robust internet connectivity is the key that can support in the establishment of AI in rural settings. Capacity building for healthcare professionals are highlighted for effective AI utilization. Inclusive data quality and its availability guarantee interoperability that can enhance rural healthcare data. Regulatory support particularly for designing of ethical algorithm, transparency and accountability throughout AI development and its implementation into rural healthcare practices are essential.
Conclusion: AI offers a critical tool in the pursuit of rural healthcare equity by alleviating health disparities. Overcoming initial implementation challenges requires collaboration among policymakers, healthcare providers, and technology developers.
Policy Recommendation: Ongoing research to develop cost-effective AI solutions tailored to real needs of the community is crucial to reducing disparities and achieving health equity.
Poster Slot
D02
13:15-13:30
Unveiling the Ethical Enigma: Navigating the Boundaries of Artificial Intelligence and Informed Consent for Patients and Healthcare Providers in India
| Presenter | : | Garima Singh Verma |
| Abstract ID | : | A055 |
The integration of Artificial Intelligence (AI) into India’s healthcare system offers a transformative opportunity. In fields such as Ophthalmology, Endocrinology, Radiology and Oncology to name a few, AI is being used for disease screening and diagnostic purposes. AI has the potential to bridge gaps and enhance healthcare equity. However, this advancement brings significant ethical concerns, notably regarding patient privacy, data security, trust and informed consent. These issues are especially critical for vulnerable populations such as women, children, and the elderly who may struggle to fully grasp the complexities of AI-driven medical treatments. Informed consent, a fundamental principle of ethical medical practice, is often inadequately addressed in India, with consent forms lacking the specificity required for both patients and healthcare providers to fully understand AI’s implications.
This research employs a qualitative methodology, including in-depth interviews with healthcare professionals, patients, and stakeholders, alongside a review of existing AI application and peer reviewed research articles. The aim is to develop a framework/ recommendation for the responsible integration of AI in Indian healthcare, prioritizing transparency, accountability, and fairness. Preliminary findings suggest that without proper regulation, up to 80% of AI systems could amplify existing biases, leading to unfair and potentially harmful outcomes. Significant gaps in safeguarding patient privacy and obtaining informed consent have been identified. The proposed framework underscores the importance of tailored informed consent practices, particularly for vulnerable groups, to ensure that the benefits and risks of AI are fully understood.
Therefore, implementing an ethical framework for AI in healthcare is crucial for protecting patient rights and promoting the responsible use of AI technologies. These framework / recommendations will enhance public trust and contribute significantly to India’s goals for Universal Health Coverage and the Sustainable Development Goals by 2030, ultimately leading to a more equitable and efficient healthcare system.
Poster Slot
E02
Session 3
13:30-13:45
A conceptual framework for the value and prioritization of of digital and data investments for health
| Presenter | : | Alex Fischer |
| Abstract ID | : | A058 |
As investments in digital and data technologies (DDTs) to support healthcare and health systems management have proliferated policymakers seeking to advance their country towards Universal Health Coverage are challenged in deciding which DDTs to invest in, be it artificial intelligence, telehealth, mobile health, health information systems, or other DDTs. Smart investment requires an understanding of the full value of each DDT for comparison and prioritization. But assessing and evaluating the full value of a portfolio of diverse DDTs remains elusive because DDTs differ in the gains for health, cost savings, efficiencies and other social benefits. In contrast, investing in traditional health interventions has long benefited from cost-effectiveness research, which makes transparent the costs and health benefits of different interventions and are synthesized by priority setting institutions such as United Kingdom's NICE and Thailand's HITAP.
In this conceptual framework, we review the landscape of methodologies for economic evaluation including extended cost-effectiveness analysis, multi-criteria decision analysis (MCDA), and cost-benefit analysis, and their strengths and weaknesses as well as their relevance and feasibility for DDTs. Next, we highlight the key principles of simplicity, diversity, collaboration, flexibility, and timeliness needed to improve the comparability of economic evaluations of DDT. Using these principles, we propose new possible methods for improving the comparability of DDTs, including a new digital-DALY metric, a time-based efficiency metric and a standardized dashboard approach, similar to MCDA. We also attempt to conceptually incorporate the challenges and risks of digital technology such as digital fragmentation and digital lock-in that can affect overall allocative efficiency and technical efficiency. We conclude with recommendations for policymakers and researchers to bridge the evidence-to-policy gap in priority setting of digital health interventions.
Poster Slot
A03
13:30-13:45
Promoting Scalable and Cost-Effective Integrated People-centered Eye Care (IPEC）for People With Diabetes in Rural China: Integrating Eye Care Services with Assistance of AI Technology to Primary Health Care (PHC) System
| Presenter | : | Phoebe Pu |
| Abstract ID | : | A061 |
China has the biggest population of people with Diabetes in the world. Diabetic retinopathy (DR) is the world’s leading cause of blindness in the working age population. DR-induced vision loss will further burden the under-resourced health system. Anhui, one of China's most populated yet underdeveloped provinces, has a significant unmet need for DR intervention. This project is working to integrate a holistic approach of DR care, including health promotion and education, diagnosis and treatment, referral and care follow-up, into the NCD management systems with focus on primary and secondary levels The project developed an integrated model to help prevent and manage diabetic retinopathy within broader diabetes care systems. It expanded DR screening and other eye care services at the primary level and cascaded the treatment of DR such as laser treatment down to the secondary facilities for the convenience of patients and reduced financial burden for both patients and the health care system. The innovative use of artificial intelligence DR grading technology delivers immediate screening results, enabling township-level referrals through a digital system to county hospitals for further diagnosis and treatment. Additionally, the project conducts regular health education sessions by county and township eye health workers for people with diabetes to enhance their knowledge of diabetes and eye complications. The project brings together health workers across different levels and community volunteers to provide holistic and comprehensive diabetes healthcare including prevention and treatment for diabetic eye disease. This includes health education, DR screening at the primary level using diagnostic artificial intelligence tools, establishment of two-way referral pathway and DR treatment in county level hospitals. The project also has a strong focus on people with special needs, such as women, left behind old people, people with disability or other vulnerabilities, to ensure they were not missed out by the existing care system.
Poster Slot
B03
13:30-13:45
Bridging the Health AI Divide in the Asia Pacific: Policy and Regulatory Recommendations for Lower Middle-Income Countries
| Presenter | : | Dion Nicole Seow |
| Abstract ID | : | A066 |
Background
Artificial Intelligence (AI) is revolutionizing healthcare, offering opportunities to enhance health systems and reduce global health disparities. Establishing robust regulatory frameworks for Health AI is essential for ensuring data privacy, security, accuracy, transparency, and accountability. This research identifies regulatory gaps in the responsible use of Health AI between high-income countries (HICs) and lower middle-income countries (LMICs) and proposes policy and regulatory recommendations for LMICs to strengthen their digital health systems.
Methodology
National and health-sector specific policies and regulations in twelve Asia Pacific countries — Australia, Japan, South Korea, Singapore, New Zealand, Taiwan, Malaysia, Thailand, Philippines, India, Indonesia, and Vietnam — were rigorously assessed against a Responsible Health AI Use framework. Its components include AI safety policies and programs, regulations, assurance mechanisms, research, and education. This research involved an in-depth literature review and expert interviews with regulators, academic experts, and policymakers.
Results
LMICs generally have weaker regulations, and Health AI policies, if they exist, do not address responsible use of AI. All four LMICs evaluated had issued limited or no ethical and regulatory guidance on Health AI use. This can be attributed to less mature digital health ecosystems, different health system priorities, and limited regulatory capacity in LMICs.
Conclusion
National Health AI plans must address responsible use supplemented by clear ethical guidelines and robust regulation. Appointing a national AI safety research institute can foster trust amongst users and beneficiaries. To ensure regulatory guidance keeps pace with technological advances, streamlining regulatory processes to be adaptive and agile along with enabling controlled testing of AI solutions is critical. High quality data infrastructure and governance are also paramount. Finally, national upskilling programs are required to build capacity amongst regulators of AI, users, and beneficiaries. These measures will help bridge digital health gaps between LMICs and HICs, ensuring safe, ethical, and effective AI use in healthcare.
Poster Slot
C03
13:30-13:45
Effectiveness of Telemedicine in Patients With Respiratory Failure: Systematic Review of Randomized Controlled Trials
| Presenter | : | Vether Fernhandho |
| Abstract ID | : | A079 |
Background:
Telemedicine refers to the practice of delivering medical care and services remotely through digital communication technologies. This approach allows healthcare professionals to evaluate, diagnose, and treat patients without the need for an in-person visit. It leverages a range of tools including video conferencing, mobile health apps, remote monitoring devices, and electronic health records (EHRs). Telemedicine potentially can be applied as a supporting care for patients with prolonged needs for hospital care. Yet, there have been very few studies on its effectiveness. This study aims to evaluate the effectiveness of telemedicine in patients with respiratory failure.
Methods: All studies were derived from PubMed, PMC, and ScienceDirect by keywords "Respiratory Insufficiency" AND "Telemedicine" AND "Home Care". The search was occurring from July 1ˢᵗ, 2024 until July 20th, 2024. Three authors searched, extracted, and evaluated the studies with inclusion criteria as RCTs within the last five years, Randomized Clinical Trial study, and Study conducted in adult patients. We excluded studies on patients under 18 years old, animal studies and irretrievable full text articles. The JADAD Scale was used to determine the quality of the included studies.
Results: Three RCTs were included after screening with a total of 122 respiratory failure patients with mean aged 55-63 years. All three studies showed that Telemedicine homecare, compared with patients in hospital, present as an effective treatment option, showed a significant improvement in overall survival, sudden patient’s hospitalization, and lower patient’s treatment cost. During quality assessment using the JADAD scale, all studies were of good quality.
Conclusion: Telemedicine is recommended as a novel option for respiratory failure patients due to its ability to significantly improve patients' experience doing their care and lowering patient's treatment cost.
Keyword: "Respiratory Insufficiency", "Telemedicine", "Home Care"
Poster Slot
D03
13:30-13:45
Evolving Arena of Adopting Innovative Digital Technologies to Enhance Health Financing Systems
| Presenter | : | Akihito Watabe |
| Abstract ID | : | A084 |
The Universal Health Coverage (UHC) service coverage index had a significant high-paced growth until 2015, but progress slowed to a 3-point increase from 2015 to 2021. In 2019, around 2 billion people faced financial hardship due to health expenses, and over 4.5 billion lacked complete access to basic health services by 2021. In this context, to advance SDG 3.8, effective, efficient, and equitable health financing is crucial. Global practices showcase that leveraging innovative technologies such as databases, mobile applications, web platforms, digital payments, big data analytics, and AI can significantly augment health financing functions.
For revenue-raising, mobile health insurance payment apps increase revenue collection and reduce costs. In pooling, interoperable information systems across health coverage schemes ensure seamless data exchange, creating a unified view of health financing. Innovative purchasing analytics transform purchasing into strategic purchasing. Cross learnings can be adopted from countries such as Korean NHIS’s AI-based fraud detection tool for insurance reimbursement, Kenya’s M-TIBA platform, which offers a mobile health wallet to manage healthcare payments, digital transaction management system of India’s Ayushman Bharat PM-JAY scheme, and Estonia’s digital invoicing system.
To adopt innovative digital health financing technologies, countries must strengthen their digital ecosystems, including governance, infrastructure, and human resource capacity. This involves establishing a dedicated digital health department, developing regulatory frameworks and data privacy laws, ensuring reliable internet connectivity and hardware availability, and implementing secure data exchange and interoperable systems. Additionally, continuous training of skilled human capital is essential for system efficiency and effectiveness.
In this context, the ADB has assessed the digital health financing landscape of the selected countries and, identified useful case studies and innovations in the Asia & Pacific Region, and proposed the integration of innovative digital technologies to leverage their health financing systems which emphasize the need to strengthen the digital ecosystem to achieve UHC.
Poster Slot
E03
Session 4
13:45-14:00
Digital Transformation in Public Finance: Empowering Jammu & Kashmir’s Health System Toward Universal Health Coverage
| Presenter | : | Rajeev Prasad |
| Abstract ID | : | A085 |
The Union Territory of Jammu and Kashmir faced significant challenges in monitoring the funds allocated to the Department of Health and Medical Education. The existing financial management system resulted in large sums of money remaining idle at implementing health facilities, leading to suboptimal resource utilization and impeding progress toward universal health coverage.
In response, the UT government embraced digital transformation, developing a new financial management model by integrating platforms such as the Budget Estimation, Allocation & Monitoring System (BEAMS), JK Payment System (JK PaySys), TreasuryNet, EMPOWERMENT, and Photographic Reporting of On-site Facilities (PROOF). This innovative approach not only addressed the existing issues but also established a new benchmark for public health administration.
BEAMS allowed the government to effectively monitor fund disbursement, ensuring optimal resource utilization. JK PaySys provided a centralized platform for Drawing and Disbursement Officers (DDOs) to manage all billing processes, while TreasuryNet facilitated online transactions and account compilation. EMPOWERMENT enhanced transparency by enabling citizens to oversee local projects, and PROOF provided real-time tracking of project progress with visual documentation and geospatial data.
The implementation of these digital tools significantly improved financial control and transparency. BEAMS and TreasuryNet enabled real-time budget monitoring and timely fund disbursement, while JK PaySys optimized payroll processing. PROOF and EMPOWERMENT enhanced performance tracking and reduced the risk of fraudulent activities.
This strategic advancement established a cohesive and transparent management system, positioning the UT of Jammu and Kashmir as a leader in technological innovation for effective public spending. These advancements have not only strengthened the health system but also made considerable progress toward achieving Universal Health Coverage.
We have thoroughly evaluated the financial model introduced by the UT of Jammu and Kashmir, and the results of this evaluation will be presented at the conference.
Poster Slot
A04
13:45-14:00
Strengthening Health Data Governance in India's Digital Health Landscape: Policy Pathways for Ethical and Protected Digital Health Practices
| Presenter | : | Oshia Garg |
| Abstract ID | : | A089 |
The rapid digital transformation of healthcare, particularly through initiatives like the Ayushman Bharat Digital Health Mission (ABDM), has revolutionized healthcare delivery in India. However, the fragmented regulatory landscape surrounding health data raises significant concerns related to patient rights, data security, and privacy. This study examines the regulatory environment for health data in India, specifically focusing on the Digital Personal Data Protection Act (DPDPA), 2023, to explore its potential for strengthening health governance and data protection in the digital health ecosystem.
The research adopts a mixed-methods approach, consisting of both quantitative and qualitative techniques, to assess India’s current health data protection frameworks, identify gaps, and analyze successful international models. Key findings highlight that while the DPDPA, 2023 introduces important provisions such as consent-based data collection, data portability, and data localization, it lacks sector-specific guidelines for healthcare and presents ambiguities in implementation. The study emphasizes the need for a unified framework that consolidates existing policies and proposes the introduction of a sector-specific Health Data Protection Act.
Barriers within India's digital health ecosystem include regulatory fragmentation, interoperability challenges, and cybersecurity risks, which undermine the effective use of digital health technologies. The study proposes several policy recommendations to address these issues, including strengthening privacy measures through encryption, enhancing interoperability with open standards, bridging the digital divide by improving access for marginalized populations, and fostering innovation through regulatory sandboxes. Additionally, the study advocates for creating a Data Protection Certification Program for healthcare providers to promote data security across the sector.
In conclusion, this research provides a comprehensive roadmap for strengthening health governance and data protection in India’s digital health landscape, aligning with the objectives of the DPDPA, 2023, and setting the foundation for a secure, interoperable, and patient-centered health ecosystem.
Poster Slot
B04
13:45-14:00
Advancing Tuberculosis Diagnosis in Vietnam: Implementation of AI-Integrated Chest X-Ray Interpretation in Health Facilities
| Presenter | : | Trang Thi Thu Le |
| Abstract ID | : | A094 |
In Vietnam, chest X-ray (CXR) triages GeneXpert testing to diagnose Tuberculosis (TB). The high variability in CXR interpretation quality across sites led to missing cases and a waste of GeneXpert tests. We integrated Computer-Aided Detection Artificial Intelligence (CAD) in health facilities to standardize the CXR interpretation and assure the accuracy of GeneXpert indication.
The USAID Support to End TB project introduced CAD into the clinical flow for TB diagnosis and treatment across seven provinces. The technology was integrated in stages, starting with analysis of 68,519 CXRs for AI calibration to determine the TB-presumptive threshold score, then gradual introduction into TB screening at 12 health facilities. An evidence-based protocol was developed to use CAD for CXR interpretation, applying the ‘CAD parallel’ model with a TB-presumptive threshold score of 0.6 (range of score: 0 - 1). CAD and humans read all CXRs, and individuals with TB-presumptive CXRs (determined either by human readers or CAD) were indicated for GeneXpert testing.
During 2022–2024, among 71,932 CXRs taken for TB screening in facilities with CAD-CXR integration, 19.2% (13,839) were TB-presumptive, as assessed by CAD or human readers. This led to 6,406 individuals tested with GeneXpert, resulting in a positivity rate of 23.5% and yielding 1,644 TB notifications (including 136 clinical diagnoses) with TB yield at 2,286/100,000 CXRs. Human reading concordance with CAD interpretations improved over time from 46.5% (July 2022) to 85.0% (March 2024). Accordingly, the number of GeneXpert tests among individuals with TB-presumptive CXR also rose from 30.0% to 63.6%.
CAD integration in health facility settings worked as a quality assurance tool for CXR interpretation. Using AI and human readers, CXR interpretation was standardized, enabling health facilities to optimize triage decisions for confirmatory TB testing using GeneXpert. The Vietnam National Tuberculosis Program intends to expand the CAD integration for facility-based TB screening.
Poster Slot
C04
13:45-14:00
Process and Outcome Evaluation of Deployment of an Artificial Intelligence-Based Hierarchical Diabetic Eye Care (AID-Eye) Model: A Quasi-experimental Study in Rural China
| Presenter | : | Xiaochen Ma |
| Abstract ID | : | A099 |
Objectives: To evaluate the implementation and effectiveness of the deployment of an Artificial Intelligence-Based Hierarchical Diabetic Eye Care (AID-Eye) Model in rural China
Methodology: We implemented the AID-Eye Model (Fig 1) between 2022 and 2024, including: (1) PHC Strengthening: equipping township health centers with portable fundus cameras and training general practitioners (GPs) for fundus image capturing, health education, referral and follow-up management, alongside training ophthalmologists at county hospitals to provide laser treatment; (2) Deployment of AI-based automatic grading algorithm for the Establishing an AI-based automated grading system for the detection of DR; (3) Establishing a vertical and horizontal integrated referral system. Implementation outcomes and facilitators/barriers were assessed using the Medical Research Council process evaluation framework. A quasi-experimental design with difference-in-differences approach evaluated the impact on eye health and diabetic eye care utilization. Quantitative and qualitative data were collected from three pilot counties in Anhui Province.
Results The AID-Eye project significantly enhanced diabetic eye care utilization, averting 27 cases of blindness and facilitating 4,300 additional DR screenings, 753 additional referrals, 829 additional county ophthalmology outpatient visits, and 167 additional DR treatments per 10,000 individuals with diabetes. The process evaluation demonstrated high intervention fidelity and reach, supported by local government and health system. Challenges included trust in AI, sustaining incentives, and integrating diabetic eye care into broader diabetes management.
Conclusions: The AID-Eye Model effectively improved diabetic eye care utilization in rural China, highlighting the potential for integrating diabetic eye care into real-world PHC settings with the assistance of AI. In resource-limited settings, technological innovations, when supported by organizational innovations within the healthcare system, play a critical role in the successful implementation and sustainability of complex interventions.
Poster Slot
D04
13:45-14:00
'Machine learning' learnings from the Brazilian Health System to address the global burden of noncommunicable diseases
| Presenter | : | Bruno Nunes |
| Abstract ID | : | A116 |
Noncommunicable diseases (NCDs), including multimorbidity, are major global health challenges, entirely influenced by social determinants of health. Addressing these challenges requires universal, comprehensive, and equitable actions in the health system and across sectors. The NCDs care requires a comprehensive life-course approach, which encompasses all levels of prevention. In this context, Machine Learning (ML) shows promise in tackling the burden of NCDs and contributing to achieving the Sustainable Development Goals (SDGs) in health, equity, and economic development. Notwithstanding the development of frameworks in the field, the principles of the Brazilian Unified Health System - SUS (universality, equity, comprehensive care, community participation, decentralized governance, access and right to information, non-discrimination, autonomy of individuals, use of epidemiology to establish priorities, and integrate health actions) align with the theoretical framework for guiding the appropriate development and adoption of ML models that consider the complexity of addressing NCDs. ML tools should, especially, adhere to three SUS's key principles: universality (addressing the needs of all individuals), equity (ensuring fairness care across diverse population groups), and comprehensiveness (covering all stages of NCD care). These principles should guide the tool's entire lifecycle, from development to post-deployment monitoring. Since its establishment in 1988, Brazil's health system has significantly improved the health of its population. Evidence demonstrates the positive impact of this system, particularly through its primary health care, based on the Family Health Strategy, in enhancing chronic condition management. The principles underlying SUS serve as a global exemplar, especially for the Global South. This study proposes adopting the Brazilian health system's principles to guide the development and monitoring of ML innovations for addressing NCDs worldwide. By creating models and solutions based on these principles, we can align with SDGs and promote equitable technology use throughout the entire health care continuum, reinforcing the concept of health as a universal right.
Poster Slot
E04
Saturday 1 February 2025
Session 5
10:00-10:15
Enhancing Vaccine Equity Through Iris-based COVID-19 Vaccination Verification for Undocumented Migrant Workers in Thailand: A Case Study in Research, Development, and Potential Extensions
| Presenter | : | Jessada Karnjana |
| Abstract ID | : | A117 |
During the COVID-19 pandemic, vaccination was one crucial measure for mitigating the impact and controlling the spread of the disease. Multiple doses of the vaccine were necessary to ensure its effectiveness, thus making vaccine recipient tracking essential. However, Thailand has at least 1,000,000 undocumented migrant workers who lack proper identification. These individuals often hesitate to receive vaccinations within the legally mandated tracking system. Driven by humanitarian concerns, the Thai Red Cross Society (TRC), in collaboration with NECTEC, has developed a biometric identification system (utilizing facial and iris recognition) specifically for vaccinating this population. The iris was chosen as one of the biometric modalities due to its several advantages, such as its uniqueness, stability over time, and high accuracy in matching performance. Initially, the iris recognition system developed by NECTEC is based on the framework proposed by John Daugman. This framework consists of the following steps: (1) identifying the iris region within an eye image, (2) transforming the iris texture into a binary code (called a template) using phase information obtained by projecting the iris onto complex-valued 2-D Gabor wavelets, and (3) matching templates for decision-making. The system has been utilized by TRC since 2021 and has been expanded for collaborative use with the Department of Disease Control (DDC). This expansion now includes 40 pilot hospitals across 15 provinces, with a total user base exceeding 40,000 individuals. The system maintains its high accuracy rate of 99%. Currently, the research team is enhancing the algorithm's precision by employing machine learning and information fusion techniques. There are also plans to broaden the collaboration with TRC and DDC to extend the scope of vaccination coverage and tracking for other critical diseases, such as tuberculosis, in the future. This system is a prime example of leveraging AI to advance equity following universal humanitarian principles.
Poster Slot
A05
10:00-10:15
Comparison of Health Information Systems Capacity among China and ASEAN Countries Based on the WHO SCORE Assessment Tool
| Presenter | : | Jing Kang |
| Abstract ID | : | A119 |
Objective: This study examines the current state of health information system (HIS) capacity in China and ten ASEAN countries, identifying successful practices and challenges to inform policy recommendations and support health information cooperation between China and ASEAN.
Methods: The WHO SCORE assessment framework was used to evaluate HIS capacity across five dimensions: Survey (S), Count (C), Optimize (O), Review (R), and Enable (E). Data were sourced from the WHO SCORE database.
Results: All countries achieved comprehensive and sustainable levels in data accessibility, indicating well-established data for health-related SDG and UHC indicators. 72% of countries have the capacity for comprehensive or sustainable health information data collection, and over 60% can review health sector performance. However, fewer than half of the countries achieved comprehensive or sustainable levels in Count, Optimize, and Enable dimensions. China and Malaysia scored highest across all five dimensions, with China reaching sustainable levels in Optimize and Review, and Malaysia achieving similar in Review. Brunei reached the highest level in Count but only the priority level in Enable. Singapore and Laos had lower scores in Enable and Count, respectively. No country reached sustainable levels across all dimensions, indicating room for improvement in HIS capacity across the region.
Conclusion: HIS capacities in Count, Optimize, and Enable are weak across China and ASEAN, suggesting the need for targeted efforts to improve health data collection, analysis, and policy application. Strengthening the link between HIS and national health planning is crucial for advancing HIS development. There is also a need to actively improve the cooperation and sharing mechanisms for health information, and promote the exploration of establishing cooperation agreements and memorandums between China and ASEAN countries to enhance the availability, reliability, and wide dissemination of shared data.
Poster Slot
B05
10:00-10:15
Establishing a Governance Framework to Leverage Artificial Intelligence in Forecasting TB Medication Demand
| Presenter | : | Taufiq Sitompul |
| Abstract ID | : | A121 |
Introduction
Efficient drug forecasting is crucial to optimize tuberculous (TB) care. However, in Indonesia, drug forecasting has relied on a labor-intensive process of manually estimating drug demand using historical data. .To improve efficiency and accuracy of forecasting, USAID’s Country Health Information Systems and Data Use (CHISU) program partnered with the Ministry of Health’s (MOH) Center for Data and Information Technology (Pusdatin) and National TB Program (NTP) to develop an artificial intelligence (AI) prototype using machine learning. A governance framework that addresses both social and technological dimensions of drug forecasting is foundational to this approach to effectively predict TB medication demand.
Results
The AI prototype leveraged patient-level data from Indonesia's TB information system to predict the annual demand for fixed-dose combination drugs for drug-sensitive TB patients up to district level. Through the process of developing the prototype, the following AI governance framework was defined to include six aspects, which are Organizational Framework, Technological Preparedness, Capacity Strengthening, User-Centered Solution, Ensure AI Safety and Ethics, and Sustainable Design
Conclusion:
As part of the AI model development phase, the AI Governance Framework served as an important reference in development and testing. In the future, organizational standard operating procedures to implement, maintain, and monitor the AI model, and AI safety and ethics, need to be reinforced before the AI model can be fully implemented in forecasting the TB medication demand.
Poster Slot
C05
10:00-10:15
Applying an integrative suite of AI tools for understanding and enhancing healthy food systems
| Presenter | : | Jody Harris |
| Abstract ID | : | A138 |
Food systems, and the food environments where supply meets demand, are fundamental in shaping diets and health. These are hugely complex and dynamic systems, and traditional research methods have struggled to capture this complexity. AI is inherently more capable at understanding and interpreting complexity and can be applied to strengthen food system research and policy.
The FEED-MU project is taking an existing AI platform for understanding global conflict (Fenix Insight Online) and adapting it to understand food systems. This includes AI-assisted and machine learning-powered scanning, curation, mapping, and geospatial analysis of open-source intelligence in 26 languages from over 100 countries; entity recognition from words or images; recognizing text networks and sentiment analysis; vectorizing, similarity and semantics analysis; and anomaly detection using a multiplicity of data sources in real time.
This platform has numerous applications to food systems that the project will apply, including comparing conventional vs. AI mapping to understand food accessibility and convenience; recognizing and categorizing foods and outlets through visual or semantic analysis for better understanding of local food availability and dietary patterns; sentiment analysis for better understanding of food desirability; capturing of large-scale vendor data to understand price and affordability; and undertaking these analysis at a huge scale and in close to real time, to better understand change and drive transformation.
AI has the potential to answer some of the biggest remaining food system questions, such as how the food supply and nutrition ends of the system connect, and how different social, economic and political drivers affect food system change across contexts. Understanding the process of using AI for food system research is an important emerging field, and adapting and applying an existing suite of AI-powered analytic tools can leapfrog years of expensive development, allowing faster, more accurate, and more useful results for improving food systems.
Poster Slot
D05
10:00-10:15
A Study on the Construction and Preliminary Evaluation of a Self-Help Online Intervention Model Among Newly Diagnosed People Living With HIV/AIDS
| Presenter | : | Xinya Tu |
| Abstract ID | : | A151 |
Background: Previous literature has consistently confirmed that HIV-infected individuals often suffer from enormous psychological stress in the early stages of diagnosis.
Objective: This study aimed to develop and evaluate a mobile mini-program designed to deliver psychological interventions to newly diagnosed individuals living with HIV/AIDS. The program is also intended to integrate artificial intelligence in the future to enhance personalization and the effectiveness of stress reduction strategies.
Methodology: The intervention was developed through a comprehensive literature review, expert consultations, and qualitative interviews to identify the primary stressors and psychological needs of newly diagnosed individuals. It includes basic stress reduction modules adapted from the World Health Organization's "Self-Help Stress Reduction Program" and enhanced modules tailored specifically for this population. Feasibility and preliminary efficacy were assessed using a self-controlled before-and-after research design.
Results: The intervention demonstrated high usability and led to a statistically significant reduction in perceived stress among participants. Expert consultations further validated the program’s feasibility and alignment with the needs of the target population. The intervention also showed potential in reducing depression and anxiety, although changes in other psychological measures, such as resilience, social support, and sleep quality, were not statistically significant.
Conclusion: This study successfully developed a mobile-based Self-Help Stress Reduction mini-program tailored to newly diagnosed individuals living with HIV/AIDS. The program, consisting of both basic and enhanced stress reduction modules, effectively reduced perceived stress in initial trials. Future integration of artificial intelligence is anticipated to further enhance the program by offering intelligent companion chat services through natural language processing. This innovation aims to provide a more natural and supportive communication channel, particularly benefiting users who experience stigma, through advanced speech recognition and sentiment analysis.
Poster Slot
E05
Session 6
10:15-10:30
Women entrepreneurs using AI to simplify access to health and non-health social entitlements among the urban poor
| Presenter | : | Suma Pathy |
| Abstract ID | : | A159 |
Covid underscored the need for social security for India's urban poor. Health-linked entitlements, including health insurance and cash transfers for pregnant women / TB patients are programmed under the health department and not integrated with other entitlements which reach the user through separate channels. Access to social protection is increasingly become more digitized, with at least a part of the delivery pathway built on a digital infrastructure. However, marginalised groups, especially women, face barriers in accessing due to the overall complexity of the process, the need for familiarity with technology and linkages with diverse government departments.
Population Services International (PSI) partnered with Haqdarshak, a social enterprise, to create a single delivery window for health and non-health entitlements and to enable doorstep facilitation of these benefits for women, persons-with-disabilities and other vulnerable populations in low-income urban settlements of Indore. The single-window delivery was made possible through a tech and last-mile support system. An artificial-intelligence (AI) powered app was used to map users’ demographics to their eligible entitlements and a network of women entrepreneurs from the community simplified the access pathway. AI system built in predictors for health entitlements and was informed by vulnerabilities, including benefits for persons-with-disabilities and transgenders. This allowed entrepreneurs to effectively use the app for these groups and to increase the overall uptake of health-linked entitlements.
The intervention provided 226,776 persons with 283,259 entitlements, with health-related schemes contributing 48%. 70% of the beneficiaries were women and 1700 were persons-with-disabilities. 440 women entrepreneurs were trained to use technology and leverage government linkages to provide doorstep assistance for social entitlements. As next steps, it will be useful for this app to build-in interoperability with government apps. enhance its utility as a business and customer-relationship management platform for the entrepreneur and provide them with safety features such as number masking.
Poster Slot
A06
10:15-10:30
SoSafe: A Life-Cycle Digital Platform Leveraging Technology to Address Social Challenges and Enhance Well-being and Resilience for People of All Ages and Genders in Thailand
| Presenter | : | Adhipat Warangkanand |
| Abstract ID | : | A167 |
As we address Thailand’s complex social challenges such as ageing populations, teenage pregnancy, gender-based violence, mental health issues, and support for vulnerable groups of people —the SoSafe: Life-Cycle Digital Platform represents a significant technological innovation. Developed by UNFPA Thailand and its key partners, SoSafe builds on the TraffyFondue city management system to create a comprehensive digital tool designed to address a wide range of social issues across different stages of life.
Launched in 2024, SoSafe utilizes the popular Line application to offer a user-friendly, confidential, and effective alternative digital means for citizens to access essential information, support and government services. The platform enables individuals to report and seek assistance for unintended pregnancies, sexual harassment, domestic violence, mental health concerns, and elder services. By facilitating seamless inter-ministerial and inter-agency collaboration, SoSafe breaks down operational silos, streamlining support from government and local authorities to ensure timely and relevant assistance. It also serves as a friendly tool for all groups of population to start constructive conversations around rights and choices on their body.
One of SoSafe’s key innovations is its integration with existing government structures from multiple ministries i.e. the Ministry of Public Health, the Ministry of Social Development, the Ministry of Interior and civil society networks, which enhances cross-ministerial coordination and policy development through data analysis. This approach not only addresses immediate needs but also provides valuable insights for long-term socio-economic strategies, supporting sustainable development and universal health coverage utilisation.
The two-hour session at PMAC 2025 will showcase SoSafe as a prime example of how technological advancements can transform health systems and tackle diverse social challenges. We will explore how digital innovations can enhance public health, ensure equitable access, and foster collaborative solutions. Participants are invited to engage in dialogue, share experiences, and consider how similar innovations can be applied globally to address the pressing needs of vulnerable populations. Together, we can build a more inclusive and resilient health system that meets the diverse needs of communities worldwide.
Poster Slot
B06
10:15-10:30
Integrating participatory mapping and advanced analytics to enable healthier food environments in Bangkok
| Presenter | : | Sabri Bromage |
| Abstract ID | : | A174 |
The food environment (FE) is a confluence of personal and external dimensions influencing consumers’ food choices. Epidemiologic studies implicate the FE as a key determinant of diet quality and nutrition, but utility of this evidence-base for enabling nutrition-sensitive FEs remains limited given challenges inherent in operationalizing such a heterogeneous and spatiotemporally-dynamic exposure. Artificial intelligence (AI) is increasingly important in augmenting FE mapping and spatial analyses. Integrating AI in collection and analysis of complex spatiotemporal FE data – particularly open data (OD) – can enable extremely powerful insights into the patterning of FE dimensions across space, time, and populations.
This study integrates participatory mapping with AI and advanced geospatial analytics of OD to interpret and inform healthier FEs in inner Bangkok, focusing on areas surrounding Phayathai Campus, Mahidol University, as a starting point for methods development. We are collecting objective and subjective ground-based data on characteristics of food outlets and consumer-food outlet interactions using a SurveyCTO-based FE mapping instrument as part of a large-scale student-led “mapathon”. Ground-based measurements are compiled with OD on food outlets, acquired via cloud-based platforms and processed using deep and machine learning tools, Google Maps Platform analytics, and Geographic Information Systems to validate a software framework for geospatial prediction of multifaceted metrics describing FE-related risks. Spatiotemporal patterns and densities of these predictions are presented across inner Bangkok in the form of a user-friendly map that periodically harvests and re-analyzes ground- and cloud-based data to enhance expansiveness and performance of predictions.
Our future ambition is to understand how spatiotemporal variation in food environment interactions – operationalized as a suite of metrics derived from ground-based participatory monitoring and cloud-based prediction – influence nutrition of urban consumers, by establishing a long-term cohort study integrating these metrics with diverse measures of diet, nutrition status, and contextualizing socioeconomic, lifestyle, and behavioral factors in the Bangkok population.
Poster Slot
C06
10:15-10:30
Harnessing AI to Increase the Efficiency of Outbreak Verification in EpiCore
| Presenter | : | Nomita Divi |
| Abstract ID | : | A177 |
Background
Innovative methods of disease surveillance are allowing earlier detection of potential threats. The key, however, is to rapidly determine which early warning signals are real and which may be false alarms. Ending Pandemics created EpiCore, a robust global network of human, animal, and environmental health professionals committed to verifying disease outbreaks. Requests for Information (RFI) are sent by human curators to volunteer experts in the geography of the event through EpiCore’s online secure platform. While RFIs are verified on average within 24 hours, creating RFIs on EpiCore requires hours of manual event information curation including entering of event details such as location, date, population affected, size of population affected, syndromes etc.
Method
AI technology was incorporated in July 2024 to reduce the person hours spent on entering RFIs on the system. EpiCore has recently adapted OpenAI’s Large Language Model (LLM) for information extraction and AI generated Requests (aka. Requests for Information (RFIs)). Accuracy of information extraction and person hour reductions after incorporation of AI systems was captured.
Findings
The harmonic mean of the AI’s precision and recall were calculated for an overall F1 score of 98.95%. F1 score balances sensitivity and specificity into a single metric. The F1 scores for each epidemiologically relevant data field were:
Event Location: 98.65%
Event Date: 96.92%
Population Affected (Human, Livestock, or Wildlife): 100%
Disease Syndrome (distinct from individual symptoms): 96.97%
Confirmed or Suspected Disease: 96.24%
Number of Deaths or Illnesses (Case Count): 95.38%
Conclusions
EpiCore is an innovative and scalable approach to improve signal quality and reduce false information supporting decision makers when used complementarity with official early detection and verification systems. These results suggest that AI technology represents an effective way to make such systems more efficient, reducing human burden and errors
Poster Slot
D06
10:15-10:30
Leveraging AI to Enhance Adolescent and Young People Access to Sexual and Reproductive Health Information and Services: Insights and Policy Implications from JustAsk! AI Chatbot in India
| Presenter | : | Sahil Kapoor |
| Abstract ID | : | A188 |
JustAsk!, is an AI driven chatbot designed for adolescents and young adults to improve access to information and services related to sexual and reproductive health and rights (SRHR), mental health and internet safety. Addressing critical knowledge gaps and reducing the stigma associated with the topics, the platform provides a personalised, non-judgmental and safe space to young people. Launched in 2023 in India in collaboration with the State Governments of India and Bayer AG, the platform also facilitates access to nearby health facilities and helplines. JustAsk!, integrates AI with behavioral science to empower the youth with essential knowledge, promotes open dialogue paving the way for a future free of stigma.
JustAsk! has registered over 130,000+ users across India and facilitated approximately 1.9 million conversations. This extensive engagement has provided anonymous critical data, enabling UNFPA and partners to address the growing need for accessible and reliable SRHR information and mental health issues. The platform is designed on an open-source technology and is compliant with Global Data Protection Regulation (GDPR) and GoI’s Digital Personal Data Protection Act 2023 guidelines to ensure privacy of the users. The data intelligence gathered through user engagement metrics, highlights the potential of AI-driven tools to bridge information gaps and create a more inclusive health ecosystem. By integrating such technological solutions and leveraging this data intelligence, policymakers can develop targeted and effective health strategies that address the unique SRHR needs of youth, contribute to universal health coverage and inform national and regional policy decisions. This approach is set to highlight regional disparities, identify trends, and drive actionable insights for responsive interventions.
Poster Slot
E06
Session 7
12:30-12:45
Evaluating Trustworthiness in AI-Based Diabetic Retinopathy Screening: Addressing Transparency, Consent, and Privacy Challenges
| Presenter | : | Anshul Chauhan |
| Abstract ID | : | A195 |
"Evaluating Trustworthiness in AI-Based Diabetic Retinopathy Screening: Addressing Transparency, Consent, and Privacy Challenges"
Background
Artificial intelligence (AI) offers significant potential for enhancing individual well-being and societal progress but also introduces complex ethical, legal, social, and technological challenges. To fully harness AI's benefits while minimizing risks, the concept of trustworthy AI (TAI) emphasizes that the successful development, deployment, and use of AI depend on establishing trust among individuals, organizations, and societies. This study explores healthcare providers' perspectives (HCP), policymakers, AI developers, and ethical/legal experts on implementing AI-based diabetic retinopathy screening, focusing on trust and trustworthiness concerns.
Methods
Fourteen semi-structured interviews were conducted with ophthalmologists, program officers, AI developers, bioethics experts, and legal professionals. Thematic analysis of the transcribed interviews was performed using Atlas. Ti, identified key recurring themes. These themes were the foundation for determining the values this group considers crucial in evaluating AI trustworthiness. The major recurring themes were instrumental in guiding the identification of quotes that this population group considers crucial for evaluating the trustworthiness of AI.
Results
Three key themes emerged regarding the perceived trustworthiness of AI from the interviews: (1) the need for trustworthy AI applications for wider use, (2) the importance of reliable data in AI development, and (3) adherence to ethical, legal, and privacy principles. The study highlighted shortcomings in the AI company's data collection practices. These include insufficient transparency, inadequate patient consent, neglect of data privacy, and the absence of clear regulatory frameworks, leading to unchecked data breaches and data colonialism.
Conclusion
Keywords: Artificial intelligence, Diabetic retinopathy screening, trustworthiness
Poster Slot
A07
12:30-12:45
Leveraging AI-Driven Psychotherapy to Strengthen Mental Health Resilience in Post-Pandemic Africa
| Presenter | : | David Kennedy Okello |
| Abstract ID | : | A206 |
Mental healthcare in Kenya remains critically underfunded, particularly in rural and informal urban settlements. This study evaluates the impact of AI-powered mental health screening tools, rooted in Cognitive Behavioral Therapy (CBT), and culturally relevant alternative therapies to improve access for underserved populations. The Healthy Minds Africa Collective Initiative (HMACI) integrates these technologies and therapies to address gaps in mental health services across Africa.
Using a mixed study design, researchers analyzed app data from 2020 to 2024 for 1.6 million users and conducted follow-ups with 15 participants in music therapy, 500 in culinary therapy, and 300 in creative therapy. Participants were selected via purposive sampling, with criteria including completed screenings. Thalia's AI-based Android app provided mental health screening and therapy options, while an AI voice tool piloted early detection of anxiety and depression.
Key findings reveal significant participation among younger demographics: 47% of users were aged 18–24, and 34% were 25–34, with males comprising 55% of total users. For 400,000 therapy participants, similar trends were observed. AI screenings increased access to mental health services by 60%, with 70% of participants reporting improved emotional well-being and social connectedness, and 65% noting reduced anxiety. AI voice analysis achieved a 25% reduction in acute mental health crises.
This study highlights the potential for scaling AI-driven tools and alternative therapies, recommending integration into public health systems to enhance universal mental health coverage. Policymakers are urged to formalize alternative therapies and adopt AI-based solutions to address mental health needs effectively.
Poster Slot
B07
12:30-12:45
Utilizing AI (Artificial Intelligence) for Predicting and Intervening in Adolescent Suicide and Self-injury through a mobile-based EMA and EMI Application
| Presenter | : | Dong Hun Lee |
| Abstract ID | : | A207 |
Despite increased awareness and interest in mental health, limitations and inefficiencies in terms of time, spatial, and economic aspects pose challenges to effective management. Especially, adolescents experiencing suicidal and self-injurious behavior encounter significant barriers as the traditional approaches may not be well-suited to their needs in an era where they are highly proficient with mobile phones. Therefore, to enable early intervention, the current study describes a mobile application that can efficiently evaluate risk and facilitate self-monitoring through Ecological Momentary Assessment (EMA) which captures the momentary process of real-time sampling. EMA is implemented using a self-injury and suicide risk scale based on Item Response Theory (IRT) and self-report mental health monitoring items including physical information, lifestyle patterns, and diary methods. Additionally, digital biomarkers and comparable devices enable the real-time collection of user information, such as heart rate, oxygen levels, and social media patterns. By leveraging real-time response data from users, an algorithm that predicts the risk of self-injury and suicide classifies behavior patterns into distinct groups (e.g., high-risk multifunction self-injury group, suicide planning group, co-occurrence of suicidal ideation and self-injury group, etc.). Then, based on the biopsychosocial-pathway model which emphasizes the importance of the interaction between biological, psychological, and social factors in understanding self-injury and suicide, an Artificial Intelligence (AI) deep learning algorithm is employed to provide evidence-based Ecological Momentary Intervention (EMI) based on each group’s identified risk and protective factors. EMI encompasses delivering interventions (e.g., body-based therapies and cognitive behavioral therapies, etc.) in real-time during individuals’ daily lives. Furthermore, by accumulating big data, the algorithm is continuously refined to offer personalized intervention. Although such support via mobile application may come with limitations in methodological, technical, and ethical aspects, it may be especially effective for global younger populations at risk of suicide and self-injury, offering versatile options for crisis intervention.
Poster Slot
C07
12:30-12:45
Mapping Healthcare Facility Access Times: Leveraging High-Resolution Population Data & Routes API
| Presenter | : | Reyhan Alemmario |
| Abstract ID | : | A222 |
Background: Accurate assessment of healthcare accessibility using reliable data sources is
crucial for ensuring equitable access to medical services and improving public health outcomes.
Earlier studies in Indonesia utilized healthcare travel time estimates were derived from surveys,
which had limitations in comprehensively assessing accessibility disparities across entire
regions due to sampling methodologies that may not accurately reflect the population's spatial
distribution. Objective: This research aims to identify specific areas (i.e., at the village unit)
where residents face prolonged travel to the nearest primary healthcare facilities in Bekasi
Regency, a rapidly developing suburban area. Methods: First, to establish starting points for
accessing facilities, our approach leveraged Meta's high-resolution population density dataset,
initially aggregated into kilometer-square grid units. Within each grid cell, we applied the
k-means algorithm to identify population clusters, then downsampled by using the centroid of
each cluster as a representative point, weighted by the total population of that cluster. Second,
we estimated the travel time from each population cluster centroid to the closest facility. The
methodology integrates Google Maps Route API for driving travel time estimation with Voronoi
partition to optimize the assignment of representative points to the nearest primary healthcare
facility. Results: The analysis identified five distinct zones where travel times exceed 30
minutes, a threshold surpassing 99.03% of the population's access times. These areas are
predominantly characterized by remote locations without nearby primary healthcare facilities,
geographical barriers, and generally lower population densities. This study enables more
informed decision-making for prioritizing healthcare accessibility improvement in certain areas
of Bekasi, potentially guiding infrastructure development and healthcare facility placement. This
method can be extended to assess travel times to advanced health facilities offering specific
disease treatments, which further fill the gap in the current understanding of specialized
medical access in the region and supporting more targeted health policy interventions.
Poster Slot
D07
12:30-12:45
Enhancing Healthcare Access for Urban Low-income Gig Workers in India using a Technologically-enabled Solution: eKure
| Presenter | : | Harsha Tomar |
| Abstract ID | : | A225 |
India's expanding gig economy, marked by informal employment and fluctuating incomes, has exposed a significant social protection gap, particularly in healthcare. Millions of low-income gig workers, often the primary earners for their families, are digitally connected to employer and aggregator platforms but remain underserved due to inadequate social security and financial constraints. This disparity persists even as the country undergoes a digital transformation, with healthcare services like teleconsultation, doorstep delivery of drugs or diagnostics and health insurance being increasingly accessible online, especially in urban areas. Though digitally savvy, this vulnerable workforce continues to face significant barriers in accessing essential healthcare services.
Recognising this, PSI and Entitled Solutions have developed eKure, a pioneering WhatsApp-based digital health platform, designed to make healthcare both affordable and accessible for low-income gig workers. eKure utilizes existing technology ecosystems and strategic partnerships with healthcare providers and employer networks to offer personalized and affordable healthcare plans for this segment. Since its inception, eKure has provided 150,000 gig workers with digital access to essential and cost-effective healthcare services, including preventive, primary care and health insurance.
The platform’s architecture is built on a robust technology that ensures data privacy and security through advanced encryption, addressing key concerns in digital healthcare delivery. Its intuitive WhatsApp interface supports multiple Indian languages, making it accessible to a broad and diverse userbase. This user-centric design, coupled with strong collaborations with employers, has driven widespread adoption, with approximately 80,000 gig workers actively enrolled.
Currently, the platform is being enhanced to diversify its offerings and refine its user interface, further establishing its role as a critical tool in closing healthcare gaps and ensuring sustainable impact. By meeting the immediate healthcare needs of gig workers and emphasizing preventive care, eKure demonstrates the transformative power of technology in delivering healthcare to this vital segment of India’s workforce.
Poster Slot
E07
Session 8
12:45-13:00
UTILISATION OF ARTIFICIAL INTELLIGENCE IN MATERNAL CARE: A SCOPING REVIEW IN LOW- AND MIDDLE-INCOME COUNTRIES
| Presenter | : | Muhammad Rizky Widodo |
| Abstract ID | : | A231 |
Background: Maternal health remains a significant public health issue in low- and middle-income countries (LMIC). Achieving the Sustainable Development Goals, particularly target 3, is challenging. However, advancements in Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) are transforming health data management and maternal healthcare delivery. Despite a surge in research, a comprehensive review of AI, ML, and DL use in LMIC maternal health is lacking. This scoping review aims to (a) collate existing research in LMIC maternal health using these technologies and synthesise findings, and (b) identify limitations faced in integrating these technologies into maternal health research.
Methods: PubMed, Scopus and Web of Science databases were searched for studies from January 1, 2000, to July 9, 2024. Inclusion criteria were: (a) studies using AI, ML, and DL with maternal health datasets, (b) studies focused on LMIC maternal health issues, and (c) original research articles in peer-reviewed journals and conference proceedings in English.
Results and Discussions: The initial search identified 99 studies, with 20 articles meeting inclusion criteria after screening. Most employed technology domain is ML models (n=17) with accuracy ranges of 60-90%. Twelve studies employed retrospective designs, and 55% (n=11) used secondary data. Primary research areas were antenatal care (n=9) and combined stages of maternal care (n=8).
Conclusions: This review highlights significant progress in using digital technology to enhance maternal healthcare in LMIC. The focus on ML models reveals opportunities for advancements in predictive modelling, indicating a promising future for digital maternal health.
Poster Slot
A08
12:45-13:00
Digital health governance in China and cross-cutting collaboration of enabling ecosystem
| Presenter | : | Minmin Wang |
| Abstract ID | : | A232 |
Background: Robust governance is crucial for effective digital health development. However, clear identification of the current and future directions of the digital health governance framework in China remains unclear.
Methods: A scoping review was conducted focusing on digital health governance in China. A total of 78 national and regional legislation, policies, and practices from January 1st 2015 to August 1st 2024 were included, and the contents were synthesized based on World Health Organization framework.
Results: In the last decades, China has been working on the digital health governance and digital ecosystem. China has established the leadership and governance mechanism, and also a digital health strategy, and China benefits from robust infrastructure including reliable electricity access, internet connectivity, and advanced devices and instruments, facilitating its digital health transformation. China proposed a digital health strategy in 2018 to foster an enabling environment that supports healthcare's digital transformation. Central to China’s digital health strategy is the integration of health information systems across primary public health services, electronic medical records, and health insurance data. A multisectoral collaboration framework involving the health sector and ministries of science and technology, healthcare financing, and human resources has been established. Concerns persist regarding legislation, policy and compliance; standards and interoperability; and workforce training. Broader collaboration between public and private sectors especially value the role of civil society were considered the key to scale up digital transformation.
Conclusion: This is the first study exploring the China’s digital health governance structure. China’s experience outlines a blueprint for an enabling ecosystem through collaborative efforts across sectors, these can provide implementation strategy for scale up digital health transformation in low- and middle-income countries.
Poster Slot
B08
12:45-13:00
Strengthening Neonatal Healthcare System in underserved geographies: Leveraging Advanced Telehealth (Tele-SNCU (Special Newborn Care Unit)) Hub-and-Spoke Models to achieve Universal Health Coverage (UHC) and Neonatal Mortality Sustainable Development Goal
| Presenter | : | Roshan Mendhe |
| Abstract ID | : | A235 |
Background: Every day, around 6,300 newborns die, making up nearly 47% of under-5 child deaths. Progress since 2015 has stalled, with 64 countries, including India, likely to miss the SDG target for neonatal mortality without urgent health service improvements. Newborn survival rates vary greatly, especially in remote areas with limited access to specialized care leading to inequity. Digital technologies and telehealth can bridge this gap, providing timely, lifesaving interventions and better care to reduce neonatal mortality and help meet SDG targets.
Objective & Methodology: PATH, in collaboration with Neonatology department of All India Institute of Medical Sciences, (AIIMS) Nagpur, is implementing a Tele-SNCU initiative in SNCU Dharni, located in a remote tribal area in Maharashtra, India to reduce neonatal deaths. The hub and spoke setup use innovative technologies like 360-degree cameras, high-speed internet, Internet of Thing (IoT), and electronic health records for real-time remote monitoring of newborns, through dashboards by neonatologists at the hub facility for immediate action. AIIMS Nagpur, as hub, providing expert consultation, virtual rounds, onsite skill-based training, and monitoring to save newborns at the spoke facility in Dharni.
Result & Discussion: Dharni's infant mortality rate of 54 per 1,000 live births is significantly higher than national and state averages. Geographic challenges and a poor referral system limit access to specialized care. Since the tele-SNCU project's start, neonatal deaths and referrals from Dharni's SNCU have decreased by 60% (April 2021 to July 2024), advancing towards SDG targets. The initiative has improved care quality, with reduced rates of respiratory distress syndrome (41.4% to 1.5%) and sepsis (21% to 14.1%), better management of very low birth weight babies, and decreased out-of-pocket expenses step towards UHC. Insights from this project, including challenges like recurring costs and EHR development, can inform scaling telehealth initiatives to other regions, supporting SDG achievements.
Poster Slot
C08
12:45-13:00
Constructing Care Cascades for Hypertension and Diabetes Management Using Health Big Data in China: A Cross-Sectional Study
| Presenter | : | Tingzhuo Liu |
| Abstract ID | : | A238 |
Background: China’s National Essential Public Health Service Package (NEPHSP) aims to promote health for all at the primary health care level and includes a focus on hypertension and type-2 diabetes mellitus (T2DM). Care cascades, which track patient retention through clinical stages, are crucial for evaluating disease management outcomes. However, there are limited contemporary data to quantify the care cascades of hypertension and T2DM in primary health care. Traditional methods for constructing these cascades are resource-intensive, but the health information systems now enable the use of routinely collected big data for this purpose.
Methods: This cross-sectional study linked individual-level data from NEPHSP, health insurance claims, and hospital records across four regions in China: Xiling District (central), Wenchuan County (western), Acheng District, and Jiao District (northern). We compared diagnosed cases of hypertension and T2DM among individuals aged ≥35 against expected totals from epidemiological data and constructed care cascades to evaluate patient retention at key stages: (1) enrolled in the NEPHSP, (2) adherent to the follow-up care of NEPHSP, (3) receiving medication treatment, and (4) having hypertension and/or T2DM controlled.
Result: The study identified 149,176 individuals with hypertension and 50,828 with T2DM, representing 46.0% and 45.6% of expected totals, respectively. Of those diagnosed, 65.4% with hypertension and 66.1% with T2DM were enrolled in NEPHSP, with adherence rates of 54.8% and 64.7%. Treatment rates were 70.8% for hypertension and 82.2% for T2DM, with control rates of 80.9% and 73.9%. Regional variations showed higher enrollment in the northern regions and better control outcomes in the central regions.
Discussion: Detection and control rates for hypertension and T2DM are suboptimal in these regions of China. Strategies are needed to enhance enrollment, adherence, and care delivery processes in the NEPHSP. Using health big data to construct care cascades offers a cost-effective and timely approach to monitor and improve primary health care delivery in China.
Poster Slot
D08
12:45-13:00
Enhancing Primary Care in Timor-Leste with Medibot: An AI-Powered Clinical Decision Support Chatbot
| Presenter | : | Chi Ling Chan |
| Abstract ID | : | A243 |
The quality of primary care in Timor-Leste is constrained by limited clinical training, particularly in remote areas where physicians lack access to clinical supervision and professional development. While clinical guidelines exist, they are often in English instead of the native language Tetun, difficult to access physically or digitally, and not well-integrated into medical practice. Consequently, providers rely on informal advice from peer networks, over Whatsapp chats. Globally, this is a common challenge among lower resource health systems with limited medical training capacity.
Medibot is the first AI-driven clinical decision support chatbot in Timor-Leste enabling primary care providers to use AI as a co-pilot in diagnosis and treatment. Medibot is a hyper local AI bot: it is trained on national Ministry of Health (MOH) guidelines to provide locally contextualized recommendations, enables doctors to query in Tetun, and integrates with WhatsApp and Telegram to reduce user adoption barriers. For added reliability, Medibot employs robust validation mechanisms, including self-critique, role assignment, peer-review, question tagging and community moderation.
Initial ground user tests are in progress to test Medibot's efficacy in promoting evidence-based practices, improving clinical decision-making, ameliorating rural patient outcomes, reducing treatment times, and enhancing provider and patient satisfaction, with prospects of expanding AI4All and achieving ‘Sustainable Development Goal 3: Good Health and Well-being’ in the long-term. In tandem with deploying Medibot, grassroot doctors are trained on effective AI prompting and digital adoption to ensure change is systemic and cost is optimized by using low-code, low-cost tools to enable scalability and sustainability. In partnership with Timor-Leste’s MOH, Medibot aims to scale nation-wide to 1,200 doctors serving a target population of 1.5 million people between 2024-2026. Beyond Timor-Leste, Medibot is currently working with the World Health Organization to scale access to AI-enabled clinical decision support to other countries in Asia.
Poster Slot
E08
Session 9
13:00-13:15
Initial incremental cost-effectiveness of AI-driven CXR screening for tuberculosis among prisoners in Southern Thailand
| Presenter | : | Nyi Nyi Zayar |
| Abstract ID | : | A248 |
Background: Thailand is a lead in Universal Health Coverage (UHC) for tuberculosis (TB), ensuring care to vulnerable populations including prisoners. Mobile chest X-rays (CXR) interpreted by both AI and radiologists has been used for TB screening among prisoners in Southern Thailand. This study aimed to estimate an incremental cost-effectiveness of AI-driven CXR compared to the radiologists’ readings.
Methods: A model-based simulation was used to compute an additional cost per disability-adjusted life years (DALYs) averted, an additional year live with full health due to detection of TB, using AI-driven CXR compared to radiologists. Cost parameters were based on the TB screening program implemented in the prisons including 2.7% prevalence of TB. The sensitivity (71%-97%) and specificity (69%-88%) of AI were referenced from previous studies. The effects of TB transmission from the false negative inmates were, however, not considered due to the limited data. The twice of gross domestic product per capita, was taken as willingness to pay (WTP) threshold to determine the cost-effectiveness.
Results: The estimated cost and DALYs averted were USD 79,379.4 and 86.2 for AI-driven CXR, and USD 184,277.0 and 87.2 for radiologists. AI-driven CXR saved additional USD 104,897.6 but resulted in 1.0 additional DALYs averted lost. The incremental cost-effectiveness ratio of AI-driven CXR screening over CXR interpreted by radiologists was USD 104,897.6 per DALY averted loss. There was 49.2% probability that losing one DALY averted by AI-driven CXR could save the cost more than WTP threshold. Sensitivity analysis showed AI-driven CXR would be cost-effective when the sensitivity of AI was more than 82.1% and specificity remains at the reference level.
Conclusion: AI-driven CXR screening saves costs, but lost DALY averted compared to radiologist interpretation. Further development of AI to have higher sensitivity is essential to reduce transmission and ensure cost-effectiveness of this innovation to strengthen UHC.
Poster Slot
A09
13:00-13:15
Mental Health Support with Machine Learning: A Prediction-Based Intervention Chatbot for Mental Health Support
| Presenter | : | Rizma Adlia Syakurah |
| Abstract ID | : | A249 |
Mental health problems and illnesses affect millions globally, including college students. They are still faced with stigmas, peer pressures, and socio-cultural barriers that hinder them from seeking the proper, adequate, safe, affordable, accessible, and professional mental health support that they need. Early screening and prompt, continuous, and personalized support are critical in improving their mental health status and academic performance. Machine learning offers promising solutions by enabling the development of more personal, accessible, cost-effective, safe, and scalable tools for mental health support. This study aims to design and develop a prediction-based chatbot for mental health support for college students. It leverages machine learning algorithms to identify emotional states, predict mental health risks, and deliver real-time tailored support. We are building a mental health chatbot for college students using four main phases; preparation, data collection, chatbot training and fine-tuning, and validating, testing, and integration phase. The first and second phases are already conducted, resulting in a set research model and data collection protocol approved by a research ethical committee and experts from mental health professionals and bioethics, five batches of data collection sessions with 125 counseling session recordings transcribed, coded, categorized, labeled, and formatted into a structured dataset containing over 7,800 data. The next phase is currently starting, with the dataset being processed by our Large Language Model (LLM) to train the chatbot, which is fine-tuned to enhance its ability to understand user inputs and generate more accurate, empathetic, and context-aware responses that address mental health concerns relevant with students' needs. Once the prototype chatbot is ready, it will be validated by mental health professionals and further tested for its usability.
Poster Slot
B09
13:00-13:15
Study design and baseline characteristics for a cluster randomized controlled trial of a mobile health-based primary care program for Type 2 Diabetes in rural Thailand – SMARThealth Diabetes program
| Presenter | : | Renu John |
| Abstract ID | : | A254 |
Study design and baseline characteristics for a cluster randomized controlled trial of a mobile health-based primary care program for Type 2 Diabetes in rural Thailand – SMARThealth Diabetes program Background: The existing primary health care system in Thailand faces challenges in providing optimal care for NCDs due to an inadequate primary care workforce. This paper presents the design and baseline characteristics for a study evaluating a technology-based solution to enhance NCD management through task-sharing among nonphysician health care workers in rural Thailand.
Methods: A pragmatic, type 2 hybrid effectiveness or implementation, parallel-group cluster randomized controlled trial of 12 months duration involving 51 subdistrict health offices in rural communities of Kamphaeng Phet province, Thailand, is currently being conducted. The intervention arm utilizes a technology enabled program (SMARThealth Diabetes), including workforce restructuring, clinical decision support system, and continuous performance monitoring, while the control arm is continuing with usual practice. Baseline characteristics are presented by comparing sociodemographic characteristics and primary as well as secondary outcomes for the trial for intervention and control groups. The trial is registered under The Thai Clinical Trials Registry TCTR20200322006; The primary outcome measure will be the change in mean HbA1C measured at randomization and 12 months from randomization, between the intervention and control clusters.
Results: The data collection commenced in November 2022. Among 1593 enrolled, baseline characteristics were comparable between the intervention and control groups. The intervention group had a lower mean BMI (25 ±8 versus 26 ± 4.6), a higher waist-to-hip ratio (0.98 ± 0.3 versus 0.89 (±0.1) and a lower eGFR (85 ± 13 versus 98 ± 16). The mean HbA1c was similar between the groups, with 7.3 (± 1.4) in the intervention group and 7.2 (± 1.4) in the control group.
Conclusions: Baseline characteristics were comparable for intervention and control groups. The trial will explore a novel digital health intervention to enhance diabetes management in rural Thailand, providing insights for improving NCD care in low-resource settings globally.
Poster Slot
C09
13:00-13:15
AI in Healthcare: Transforming Bangladesh's Urban and Rural Landscapes
| Presenter | : | Mahbubur Rashid Ories |
| Abstract ID | : | A255 |
Artificial Intelligence (AI) is advancing rapidly and can potentially revolutionize healthcare. The study employed both qualitative and quantitative methodologies to examine primary data collected based on the health program of a non-governmental organization named Centre for Development Innovation and Practices (CDIP), as well as its rural and urban patients and beneficiaries in five districts of Bangladesh: Dhaka, Feni, Noakhali, Gazipur, and Cumilla. The sample size for the primary data collection survey was determined based on five significant study areas in Bangladesh: two city corporations, three pourashavas, and ten urban villages. This calculation was done using ideal-wise research methods, ensuring a perfect confidence level and margin of error. This study examines the impact of artificial intelligence (AI) on Bangladesh's healthcare system, considering both urban and rural perspectives. This paper examines the potential advantages and disadvantages of AI in healthcare delivery. According to the study, most community and urban medical officers believe AI can enhance patient outcomes by tailoring treatment plans, optimizing hospital operations, and improving urban diagnostics. Nevertheless, rural areas encounter obstacles such as inadequate digital literacy, infrastructure, and healthcare accessibility when integrating AI. Notwithstanding these obstacles, AI has the potential to enhance healthcare accessibility, facilitate remote diagnostics, and alleviate the workload of medical personnel. This study analyzes the socio-economic, moral, and practical impacts of AI on Bangladesh's healthcare system. The study presents a comprehensive plan for the implementation of an artificial intelligence (AI) healthcare strategy at a national level in Bangladesh. The plan prioritizes the establishment of regulatory frameworks, development of infrastructure, and promotion of AI education. The study emphasizes the importance of fair access, ethical debates, and public-private partnerships. Pilot projects and continuous evaluation need to be utilized to maximize the advantages of artificial intelligence and tackle both urban and rural difficulties.
Poster Slot
D09
13:00-13:15
Evaluation of Hainan’s Multifaceted Approach to Promote System-level Digital Therapeutics Development: A Mixed-Methods Study
| Presenter | : | Jin Xu |
| Abstract ID | : | A135 |
Poster Slot
E09
