Author(s): Wei Junyi, Tiffany Grace Sajoto, Zheng Peirong, Song Zhuowei, Lau Kai Heng, Kim Yu Ling

Background

In this paper, we will dive into the global engagement of nations in the realm of AI-driven healthcare and pinpoint areas of enhancement that can foster equity, fairness, and inclusivity in AI-based healthcare technology. Generally, the potential lack of representation of minorities in medical databases, and the other being misrepresentative or discriminatory data that is not addressed when developing AI models are two issues that potentially impact on AI-powered healthcare. These challenges give rise to biases in AI systems, which may exacerbate disparities rather than enhancing healthcare quality, particularly for marginalised populations. This research aims to provide an overview of the participation of various nations in the development and integration of AI models in medical applications and to highlight the apprehensions expressed by healthcare professionals regarding the sources of bias and inequity within these models.

Current Development of Medical AI in Singapore

Jarvis DHL

Jarvis DHL is an AI system used to transform chronic care for diabetes, hypertension and hyperlipidaemia. The development of this AI-based medical technology is contributed by National University of Singapore (NUS) and SingHealth Group (SingHealth). Their aim on this project is to integrate multiple solutions into a consolidated AI platform that designed to enhance the delivery process of healthcare, particularly in the context of the “3H” care framework. In the meantime, they also target to develop an AI system to gather local healthcare data to create AI algorithms and models, while facilitating evidence-based personalised care and shared decision making by primary care physicians.

Explainable AI

This AI technology which has NUS and National University Health System (NUHS) involved serves as a service for community healthcare. It focuses on advanced AI with prototype devices built for deployment and testing in a community setting. It aims to use an “AI as a service” approach to provide advice, tools (such as food logging) and lifestyle coaching.

An End To End Adaptive AI-assisted 3H Care

This adaptive model is built by Nanyang Technological University (NTU) together with National Healthcare Group (NHG), functioning to periodically assess the status of 3H patients, and identify pre-3H persons based on early behavioural patterns, health symptoms and other non-medical factors. This project is also aiming to develop approaches to create long-term behaviour change through gamification.

Applications of Medical AI in Treatment and Diagnosis

Early-stage detection and prevention[6]

SELENA+ system is applied for analysing eye retina images to detect three types of eye diseases in support of the national diagnostic retinopathy screening program. 

Targeting for clinical treatments as well as targeting for related special health support programs

  Medical AI can be used for prediction of hospital inpatients with a high risk of multiple readmissions in support of the hospital to home national program.

Resource optimization within hospitals and polyclinics, and across the entire public healthcare supply chain and service delivery network

Short term and near-term emergency department arrivals are able to be benefited from the application of medical AI. Moreover, number of people entering and leaving hospital via all entry points, and number of people in key hospital areas susceptible to crowd build ups via real-time video analytics (C3 centre uses various AI enhanced simulation models to support action planning and option assessment for dealing with current or impending congestion, load imbalances, and resource constraints across key hospital resources).

Administrative support

The use of robotic process automation and wireless sensors to facilitate covid patient discharge from special facilities set up as covid community care facilities.

Local AI Healthcare Assessment

Data on underrepresented groups and reasons for their underrepresentation

In Singapore, the largest ethnic minorities are Malays and Indians, making up 13.5% and 9.0% of the population in 2020. Although they make up a significant proportion of Singapore’s population, Singapore’s ethnic Chinese population take an even larger portion of it, making up 74% of the population. This causes Malays and Indians to be highly underrepresented in medical trials and studies. In fact, the first medical study across equally sized samples of all three minority groups is done in 2018. Given that fairer representation was only done recently in a few studies, the majority of medical data in Singapore would likely be oversaturated with data from the ethnic Chinese population, causing greater data bias in medical AI towards the Chinese population, reducing its effectiveness and fairness towards Malays and Indians.

Although the disparity in underrepresentation may not be as vast between biological men and women, it is nonetheless a significant problem. The Integrated Women’s Healthcare Programme has found that women tend to be an underrepresented demographic in global clinical trials, much less Asian women. In North America and Europe, where major medical agencies are located and conduct trials, the Asian population is a minority. Furthermore, while there is some improvement, most efforts to include sex and gender are inconsistent in nature. The approval of these major bodies influences the global medical sphere, including Singapore. This leads to a lack of medical data available on how the effects of chronic and acute illness and their various modes of treatment differ from Asian men. While existing clinical trials in Singapore would likely have more gender equity due to increased medical studies focused on women, this disparity in existing medical data no doubt still exists.

Methods of medical data collection in Singapore

In Singapore, medical data collection primarily occurs through a combination of electronic health records (EHRs), healthcare information systems, surveys, and administrative databases. Here are some of the key methods of medical data collection in Singapore:

1. Electronic Health Records (EHRs)

The NEHR is a central repository of patient health records that allows authorised healthcare providers to access and share patient information electronically. It contains key medical information such as diagnoses, medications, allergies, and laboratory results.

2.  Healthcare Information Systems 

Healthcare Information Systems (HIS): Each hospital in Singapore has its own HIS to manage patient data within the institution. It includes patient demographics, medical history, treatment plans, and billing information.

Note: Clinic Management System (CMS): Similar to HIS, CMS is used in outpatient clinics to manage patient information and appointments.

3. Surveys and Studies

Health and Population-based Surveys: The Ministry of Health (MOH) and other agencies conduct periodic surveys to collect data on various health parameters, such as the Singapore Health Promotion Board’s (HPB) National Health Survey.

Clinical Trials and Research Studies: Academic institutions, hospitals, and research organizations conduct studies to gather specific medical data for research purposes.

4. Administrative Databases:

National Registry of Diseases Office (NRDO): NRDO maintains registries for chronic diseases like cancer, stroke, and heart disease. It collects data from public and private healthcare institutions to monitor disease trends.

Birth and Death Registration: The Singapore Registry of Births and Deaths collects demographic data, including information on births, deaths, and causes of death.

It’s important to note that data collection in Singapore is subject to strict privacy and confidentiality regulations outlined by the Personal Data Protection Act (PDPA) and other relevant laws. Patient consent and data security are paramount concerns in all data collection methods.

Involvement of Singapore in International AI research and open science efforts

1. Singapore and global medical AI

SEA-CoreNLP: It aims to promote the development of NLP in Southeast Asia and be the central hub for it. The current focus is on “core” NLP – tasks such as part-of- speech tagging, syntactic parsing or semantic role labelling etc. for Southeast Asian languages.

2. AISG x Egypt

AI Singapore and Egypt’s Ministry of Communications and Information Technology (MCIT) have signed a cooperation agreement to implement in Egypt two of AISG’s premier talent development programmes in AI, the AI Apprenticeship Programme (AIAP)® and AI for Everyone (AI4E) ®. This collaboration will help Egypt in its strategy to launch capacity-building programmes to propel the country’s AI capabilities.

3. GPAI

The Global Partnership on Artificial Intelligence (GPAI) is a multi-stakeholder initiative which aims to bridge the gap between theory and practice on AI by supporting cutting-edge research and applied activities on AI-related priorities.

Built around a shared commitment to the OECD Recommendation on Artificial Intelligence, GPAI brings together engaged minds and expertise from science, industry, civil society, governments, international organisations and academia to foster international cooperation.

AI Singapore is the co-chair of the working group on Innovation and Commercialization and the Broad adoption of AI by SMEs sub-committee.

4. AIRI

AI Singapore’s AI Readiness Index has been adopted by companies and associations globally. This includes GPAI within the workstream of the Broad Adoption of AI by SMEs, in which the SMEPortal embeds AIRI as a core feature.

Health conditions, complications, and symptoms that are specific or more common for people in Singapore and Southeast Asia

Singapore has the second-highest proportion of diabetics among developed nations, a report in 2015, by the International Diabetes Federation (IDF) revealed.[1] The recent National Population Health Survey showed an increase in the crude prevalence of diabetes from 8.8% in 2017 to 9.5% in 2020,[2] in compare with the latest and most comprehensive calculations show the current global prevalence rate of 6.1%.[3] In 2010, 1 in 9 Singapore residents aged 18 to 69 years were affected by diabetes. Indians and Malays consistently had higher prevalence of diabetes compared to Chinese across the years. An estimated 430,000 (or 14% of) Singaporeans aged 18-19 years are also diagnosed with pre-diabetes.1 in 3 individuals with diabetes do not know they have the condition. Among those diagnosed with diabetes/aware of their disease, 1 in 3 have poor control of their condition, which increases the risk for serious complications. Diabetes was the 4th and 8th most common condition of polyclinic attendances and hospitalisation respectively in 2014.

 Problems and potential improvements that could be made for more inclusive medical AI in Singapore

1. AI tools are yet to be perfect, through machine learning and deep learning algorithms, Selena+ reads these digital retinal photography, identifying diabetic damage to the eye with an accuracy rate of over 95 to 97 per cent.[4]

2.  The Multiple Readmission Predictive Model identifies high-risk patients for better care intervention to reduce their risk of being readmitted to hospital. This AI model analyzes multiple facets of a patient via hundreds of indicators, and has an accuracy of 7 in 10 patients correctly predicted.[5]

References

[1] Jessie, Lim.(2022, August 12). Singapore is no. 2 nations with the most diabetics: 5 things about diabetes. The Straits Times. https://www.straitstimes.com/singapore/singapore-is-no-2-nation-with-most-diabetics-5-things-about-diabetes#:~:text=Singapore%20is%20No.,about%20diabetes%20%7C%20The%20Straits%20Times 

[2] News highlights. Ministry of Health. (n.d.-c). https://www.moh.gov.sg/news-highlights/details/result-from-government%27s-five-year-war-against-diabetes-effort/#:~:text=Hence%2C%20while%20the%20recent%20National,the%20same%20period%20at%207.9%25. 

[3] Global diabetes cases to soar from 529 million to 1.3 billion by 2050. The Institute for Health Metrics and Evaluation. (n.d.). https://www.healthdata.org/news-events/newsroom/news-releases/global-diabetes-cases-soar-529-million-13-billion-2050#:~:text=The%20latest%20and%20most%20comprehensive,jump%20to%2016.8%25%20by%202050. 

[4] Auto, H. (2021, January 27). Diabetes Singapore deploys AI technology to screen patients for early signs of diabetic eye conditions. The Straits Times. https://www.straitstimes.com/singapore/diabetes-singapore-deploys-ai-technology-to-screen-patients-for-early-signs-of-diabetic 

[5] Ai futures: How ai is Augmenting Singapore’s Healthcare. GovInsider. (n.d.). https://govinsider.asia/intl-en/article/ai-futures-how-ai-is-augmenting-singapores-healthcare-goh-han-leong-ihis 

[6] Ta, A. W. A., Goh, H. L., Ang, C., Koh, L. Y., Poon, K., & Miller, S. M. (2022). Two Singapore public healthcare AI applications for national screening programs and other examples. Health Care Science, 1(2), 41-57.

[7] Advisory guidelines for healthcare sector – PDPC. (n.d.-a). https://www.pdpc.gov.sg/-/media/Files/PDPC/PDF-Files/Sector-Specific-Advisory/advisoryguidelinesforthehealthcaresector28mar2017.pdf 

[8]  Ai Grand Challenge. AI Grand Challenge. (n.d.). https://www.synapxe.sg/healthtech/health-ai/ai-grand-challenge 

[9] International. AI Singapore. (2022, November 15). https://aisingapore.org/home/international/ 

[10] Medicine.nus.edu.sg. (n.d.). https://medicine.nus.edu.sg/obgyn/research/reproductive-development-biology-research-program/iwhp.html 

[11] Publications. Bioethics Advisory Committee. (n.d.). https://www.bioethics-singapore.gov.sg/publications/ 

[12] Ravindran, T. S., Teerawattananon, Y., Tannenbaum, C., & Vijayasingham, L. (2020, October 27). Making pharmaceutical research and regulation work for women. The BMJ. https://www.bmj.com/content/371/bmj.m3808 

[13] Singapore Department of Statistics | Census of Population 2020 … (n.d.-b). https://www.singstat.gov.sg/-/media/files/publications/cop2020/sr1/findings.pdf 

[14] SPH.NUS.EDU.SG. (n.d.). https://sph.nus.edu.sg/2018/04/examining-chronic-disease-risk-and-outcomes-among-singapores-major-ethnic-groups/#:~:text=This%20is%20the%20first%20cohort,more%20prone%20to%20insulin%20resistance