Ethical AI-led Primary Care Infectious Disease Early Warning System

Our systematic review and research with end-users’ will inform the development of an ethically framed infectious disease (ID) early warning system (InDEW) using big data from patient’s electronic health records (eHR) and machine learning (ML) to detect early outbreaks of IDs in primary care in China.

Dr William Chi Wai WONG, Clinical Associate Professor, Li-Ka-Shing School of Medicine, The University of Hong Kong (HKU); Chief of Service, Department of General Practice (GP), HKU-Shenzhen Hospital, China

As of March 2021, WHO has reported over 126million confirmed COVID-19 cases with over 2.7million deaths. China has a highly dense population, combined with the cultural practices of consuming wild animals has led to serious and frequent ID pandemics such as SARS and COVID-19.  The COVID-19 outbreak highlighted the important role that primary care plays in screening and monitoring for COVID-19 as distinct from other upper respiratory diseases while maintaining routine primary care.

Over the past decade, China has made remarkable progress in its primary healthcare system with more than 34,000 community health clinics (CHCs) serving 706million people through a combination of western medicine and traditional Chinese medicine (TCM).

Our solution makes use of the Big Data in Patient’s eHR in Primary Care and the application of ML, a form of AI algorithms for data analysis, to develop a system that can identify and monitor emerging pattern of symptoms and presentations of common and novel IDs in REAL TIME, and enable early identification of community outbreaks.

To ensure our end solution is ethically sound, we have conducted a systematic review and a qualitative study to frame the inherent ethical issues, critical to medical surveillance systems using Big data and AI.

The innovation and uniqueness of in our solution lies not only in the creation of an ID early warning system at primary care level using eHR Big Data and ML technology, which has far-reaching benefits in ID control and prevention, but also in the rigorous ethical governance framework applied to ensure the integrity and transparency in the daily application of our ID early warning systems.

First, we undertook a systematic literature review on 29 peer-reviewed articles and secondly, a national qualitative study with 16 frontline general practitioners (GPs) and 32 patients was conducted in China to identify potential ethical implications of big data analytics. We found the access, use and sharing of eHR pose controversies regarding data security, individual’s privacy, informed consent, confidentiality, distributive justice and system governance. Regulations and guidelines on data privacy and security, confidentiality, transparency of AI-led surveillance system to protect individual’s rights are scant. Addressing issues, such as the publics mistrust and concerns on lack of governance framework to regulate reliability of the ML algorithms, personal data protection, potential resultant social stigma discrimination from the output results etc., is critical and high on our agenda.

Public health impact: In the short term, the system will monitor ID patterns, such as COVID-19 and other IDs in the community, enable a better understanding of ID evolution and early detection any potential “second waves” that will curb further outbreaks. In the medium-to-long term, the model will identify early stages of an outbreak to strengthen our defense against COVID-19 or other future ID pandemics.

Cultural and social impact: This is the first project to apply ML technologies in ID control in primary care in mainland China, where there is a unique mixture of TCM, western medicine, and local cultural practices towards IDs. Such data may yield previously unknown risk factors and new inter-relationships to improve understandings and prevention of IDs. The ethical governance framework will enhance public’s awareness, trust and acceptance to bring about attitudinal change on use of AI and personal data, providing an exemplar for future AI-led initiatives and regulations.  

Economic and political impact: Information derived from our system will be instrumental in guiding public health, economic and international policies on ID prevention and control, ranging from public communication on ID control, testing requirements, restrictions on certain social activities and movement, border control policies etc. 

The value of early warning, measured in hours or days, can be of immense significance in controlling ID spread and reducing mortality. Examples of how it works:

For URTI, our system can detect unusual ID patterns (increased frequency of cases or clinical symptoms) against baseline, alerting local CDCs in such activities/ suspected cases and their locations. It will enable health authorities to response promptly, to investigate, trace the source, and prevent a potential ID outbreak which may escalate into pandemic. For common influenza, the system may give a few days advanced notice for the Health Bureau to anticipate the peak of seasonal flu, and plan healthcare resource allocation accordingly.

In China, many STIs are treated based on Syndrome Approach and often blended under urethritis by GPs; hence, not reported to the local CDCs, leading to underreport and community outbreaks. Our system could detect increase in frequency of patient cases presenting with urethritis/STI-like symptoms or treatment alerting local CDCs to investigate into their causes and initiate corresponding measures, e.g. contract tracing, mandatory testing for at-risk groups.

HVB represents a huge disease burden in China. Early identification of potential HBV outbreaks could alert GPs to test HBV in at-risk groups and linkage-to-care.

In the first 12 months we intend to build early warning prototype models for 16 common IDs (Figure 1) based on the 300,000 patients’ eHR from the Department of GP of HKU-SZH. The accuracy and sensitivity of these models will be tested using historical and prospective eHR from the hospital.  Once required accuracy is achieved, we will pilot the early warning models at the Department of GP at HKU-SZH prospectively and at least one CHC in the Shenzhen. This will involve major changes to the existing consultation system and connect it initially to the research team for data analysis.

The Department of GP is the design leader of the International GP Integrated Project by the Shenzhen Health Bureau to raise the quality of primary care benchmarked against international standard. The highly acclaimed Key Clinical Discipline of GP was conferred to the department for its academic excellence by the Shenzhen Health Bureau in 2020. The research team is thus in an excellent position to work with the local CDC. We will further develop early warning models for other common/known IDs as well as novel IDs and will apply the early warning systems to CHCs in the rest of Shenzhen and China. 

We will adopt UNPD framework to measure the social impact of our work. The primary outcome is the number of common (and novel) ID outbreaks identified in the first year of implementation of early warning system at Department of GP at HKU-SZH and one selected CHC (pilot phase). Other secondary outcomes are within 5 years post-pilot: Number of CHCs adopted our ID early warning system in CHCs in China and elsewhere; Number of common (and novel) ID outbreaks identified; Number of contact tracing; Number of ID case prevented; Number of training workshops/ peer-reviewed conference/ manuscripts to promote the system and raise awareness. Satisfaction of the end-users; Number of complaints or ethical-related issues addressed this project; and Number of other guidelines/ systems that would adopt our ethical approach.

During the model training stage, we will build an iterative system to update the structures and the parameters of this early warning model by the update of eHR information, to make sure it could be automatically upgraded and optimised as data grows. Our initial goal is to build a prototype early warning model for 16 different IDs with 80-90% prediction accuracy within 95% CI. 

(1)Operational/ infrastructure challenge in implementing expansion, such as the quality and availability of patient’s her: our solution relies on a universal healthcare system with an IT infrastructure where most of the patients’ record are in electronic format. This may be difficult to introduce our system in countries where there is no fully implemented eHR.

(2)Legal barrier in the use of personal data: ID control often involves contract tracing. There is a potential risk of exposing patients’ information against local regulations on confidentiality and privacy. Furthermore, these is data security risk of information leakage from the system.

(3)Ethnics, social and culture barriers:  This is particularly relevant for socially stigmatised IDs, such as STIs and HBV. The lack of transparency of how their personal data is being used in the process, lack of an ethical governance framework, mistrust in the authorities, the fear of exposure, potential social discrimination, penalties in case of contract tracing- all contribute to public reluctance to accept this technology. This could be addressed with a combination of the establishment of a clear ethnic governance framework and public communication raising their understanding on how their data are being used.

Additional affiliations

• Co-Director, Guangdong Provincial Train-the-Trainer Centre (Since Jul 2019)
• Leader, International GP Integrated Program (Since Dec 2019
• Committee Member of Institute for Global Health and Sexual Transmitted Diseases, Southern Medical University (Since 2019)
• Training Faculty, South China STI/HIV supported by NIH, USA (2013-2019) Advisory Professor, Fudan University, China (2017-2020)

The HKU-SZH has been the forerunner in raising standard in China healthcare reform. In this project, we pride ourselves in setting up Primary Care ID early warning system which serves its purpose in an ethical manner. Gaining recognition in the Trinity Challenge will serve to reaffirm the international benchmark and our commitment to upholding a high ethical standard in the use Big Data Analytics in healthcare. This could help the team to provide further evidence that such approach is needed and convince our partners to implement it in a wider context. We also hope to get reviewers’ comments to further improve our work.

The University of Hong Kong (HKU), the primary applicant belongs to HKU and could leverage on their expertise and advise in infectious diseases and microbiology to ensure quality of the project.