Ethics, Integrity, and Retributions of Digital Detection Surveillance Systems for Infectious Diseases: Systematic Literature Review

Introduction

In the wake of the global COVID-19 outbreak, there is growing pressure to improve our existing practice in the prevention and ongoing monitoring of emerging infectious diseases and the adoption of targeted interventions for emerging infectious diseases. Current infectious disease surveillance systems in most countries are remarkably similar. Once a case is clinically suspected and confirmed, there are multiple levels of reporting [1]. Then, the accumulated information from local institutions is aggregated, processed, and defined at the population level before actions are subsequently disseminated through the system from a “top-down” approach. The existing process carries an inevitable time lag that can result in both reduced effectiveness for responsive public health interventions [2] and opportunities for doctors and patients to negotiate reporting, which can have catastrophic results, as had been observed in the early COVID-19 outbreak in China [3] and the Ebola outbreak in West Africa [4]. The need for timely data collection or sharing, processing, decision making, and reporting in infectious disease surveillance has been identified as one of the main drivers for introducing artificial intelligence (AI) technology.

With the establishment of electronic health records (EHRs), big data have been acquired, making it possible to build data-intensive infectious disease surveillance or early warning systems, pre-empt emergency response, and strengthen infection prevention and control. Machine learning (ML) technologies, a multiplying form of AI, has shown considerable potential in tracing the source and detecting potential outbreaks or novel infectious diseases using patients’ EHRs. By utilizing real-time digital data analysis, a fully automated system could be built to transmit, through extraction, structured data and doctors’ medical records in text, while new technologies, such as named entity recognition, would allow extraction of patient-related features from the unstructured text into predefined categories to support future infectious disease monitoring and surveillance [5]. Examples of enhanced timeliness resulting from this approach in COVID-19 case tracing have been reported in China and several other Asian counties [3]. When AI is being employed for infectious disease control (eg, using mobile phone apps to trace COVID-19 cases), potentially infected patients, their close contacts, and, at times, larger communities can be tracked, tested, and, if necessary, quarantined to prevent further outbreaks.

Nonetheless, the advantages of big data and ML in infectious disease control need to be weighed against the considerable ethical and legal concerns pertaining to the protections and privacy of individuals and the public in respect to access, use, and sharing of large data sets of patients’ medical records. These kinds of AI interventions raise complex contemporary ethical questions regarding potential misuse of personal information and informed consent that have the potential to infringe on one’s human and civil rights. Furthermore, there is the heightened risk of patients’ personal information being leaked to social media when they have previously been assured of confidentiality and privacy [6]. Reidentification of named patients is a major concern when databases are hacked [7], and data custodians may sell data for financial gain to pharmaceutical, insurance, or software companies [8]. Furthermore, the routine use of big data analytics (BDA) or the ethics and widespread moral implications of ML continue to be vigorously debated around the accuracy of reporting [9,10], and the consequences of inaccuracies in the reporting of outbreaks are gaining considerable attention [11]. These developments, often in rapid response situations, have sparked issues between optimizing population health outcomes informed by epidemiology and public health, and societal or individual ethical rights and protections that inform human rights and freedom of choice.

Notwithstanding these debates, BDA is critical for managing communicable disease spread or outbreaks in today’s digital world where consistency in the application of regulations or rules of privacy, confidentiality, transparency, data handling, and security safeguards for ML transmission required to protect individuals’ rights is limited [12]. This study set out to examine the ethical issues of using AI and identify an intersection or balance between the protection of individuals’ human rights and patients’ autonomy, and common good for population-based public health outcomes specific to infectious disease control and prevention with the aim of investigating ethical issues identified from utilizing AI-augmented surveillance or early warning systems to monitor and detect common or novel infectious disease outbreaks.

Methods

Search Strategy and Selection Criteria

We searched broad search engines with no time restriction. CINAHL, PubMed, Science Direct, MEDLINE, Google Scholar, and Scopus, as well as legal and sociological databases were searched from database inception to December 8, 2020. Additionally, relevant articles from the documents’ bibliography and from other articles that cited the documents were retrieved. The following search terms and combination of terms were utilized: ethic* OR “data security” OR “data privacy” OR sensitiv* OR confidential* OR anonym* OR “personally identifiable information” OR privacy or “human right*” AND “electronic health records” OR ehr OR “Clinical decision support system” OR cdss OR “Artificial Intelligence” OR ai OR “augmented surveillance” OR surveillance OR “contact tracing” OR “Machine Learning” OR ml OR “deep learning” AND “Infectious Disease” OR “Communicable Diseases.” Prior to the searches, the authors reviewed all search terms. In order to identify all existing literature for this review, we considered all peer-reviewed empirical research articles, review reports, and grey reports. To enhance the rigor of the review, our approach followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Multimedia Appendix 1) [13].

Inclusion and Exclusion Criteria

There were no language restrictions placed on the literature search. To be included for further review, the collected articles must have addressed the ethical issues of using AI, digital surveillance systems, early warning systems, and/or BDA technology for detecting, monitoring, or tracing specifically infectious diseases.

Literature Selection

We selected literature for inclusion in 2 stages to ensure rigor. In the first stage, 2 authors (IYZ and YXM) independently screened the titles and abstracts of all citations for potentially relevant articles. In the second stage, the same 2 authors independently examined the full texts of these papers against prespecified inclusion criteria. Any discrepancies were resolved with input from a third author (EH). Of the 6714 titles and abstracts reviewed, we excluded 6668 articles that did not meet our eligibility criteria, resulting in 46 articles for full-text review. Additionally, 17 articles were excluded because they (1) were not related to digital surveillance systems, (2) did not focus on infectious diseases, or (3) lacked focus on ethical concerns. In all, 29 papers satisfied our eligibility criteria and were included by consensus agreement (Figure 1).

‎View this figure

Data Extraction, Quality Assessment, and Data Analysis

All authors read and reviewed the 29 articles in order to summarize the approaches, methodologies, samples, and findings. Two authors (IYZ and YXM) undertook literature quality assessment, extracted data from the literature, included them into a spreadsheet, and analyzed the data independently. Discrepancies were resolved, and data were confirmed in several rounds of discussions with other team members. We evaluated study quality and methodological rigor for 5 empirical studies by using the modified mixed-methods appraisal tool (MMAT) [14], which was not applicable for the remaining discussion papers. Lower quality scores did not result in exclusion of any articles. However, the findings of articles with lower quality scores were given less weight during data analysis.

We structured our identified ethical issues using a theoretical framework developed by Asadi et al [15], in which key concepts of prior BDA had been identified, defined, and examined using stakeholder theory and discourse ethics [16]. Key themes from our review were discussed and summarized, and the gaps in the literature and methodologies were identified.

Results

Overview and Participant Demographics

Among the 29 included articles, 2 qualitative studies [17,18], 2 quantitative studies [19,20], 1 mixed methods study [21], and 24 discussion papers were identified (Table 1). All of the selected articles were published between the years 2015 and 2021. The 5 empirical studies [17-21] adopted a combination of focus groups, a Delphi approach, a database digital ethnography, and surveys as research methods, and they were all identified as high quality. Together they included 7331 participants. Of these 5 studies, 1 was conducted in South Korea [21] and 4 in Australia [17,19-21]. Four studies [17-19,21] reported gender ratios and 3 [17,19,21] demonstrated age ranges. Research participants were the general public; policymakers; and experts in infectious diseases, epidemiology, food safety, health informatics systems, and health and technology law.

Ethical Issues

Textbox 1 shows the key ethical issues identified in the review summarized and clustered under individual, organizational, and societal levels with key themes presented under the corresponding concepts. Six domains, namely awareness of implementing digital infectious disease surveillance, digital integrity, trust, privacy and confidentiality, civil rights, and governance were highlighted.

Awareness of Implementing Digital Infectious Disease Surveillance

The informants reported insufficient understanding at every stage of data collection and distribution of digital infectious disease surveillance systems. They felt they were not informed of their rights to refuse or their ability to withdraw consent. Irresponsible decision making was identified by infectious disease patients in relation to insufficient information provided about contact tracing technology. At the societal level, digital infectious disease surveillance based on BDA was poorly accepted by the general public due to their uncertainty about its purpose and the risks posed from the potential mitigation of data sharing. Consequently, they wanted more information about the digital infectious disease surveillance systems and wanted their associated ethical concerns and consequences addressed. They also expected organizations to make public and transparent the algorithms and data processes used, and use plain language when explaining infectious disease surveillance. Simultaneously, both the general public and field experts emphasized that governments or institutions should convey the importance of infectious disease outbreak control to communities without violating ethical principles.

Data Integrity

Data integrity weaknesses were identified as common in BDA. Unreliable or invalidated data sourcing or algorithms were seen to lead to inaccurate identification of outbreaks or infected individuals, false predication of an event’s trajectory or the likelihood of reoccurrence, and inaccurate notifications. The outcomes, such as inadequate data integrity, were further seen to continuously intensify economic losses to trade, tourism, and health services, causing unnecessary panic and the loss of public trust in health authorities. Some organizations and experts also worried that digital infectious disease surveillance systems would widen existing health care disparities or digital divides by underrepresenting vulnerable and at-risk populations such as older adults, children, and people in economically underresourced areas; for instance, data of COVID-19 hotspots influenced the allocation and distribution of resources [25]. It was recommended that digital infectious disease surveillance applications should consider ethical requirements and the rights of people from diverse regions and communities.

Trust

Some people questioned the necessity and effectiveness of using contact-tracing technologies in an AI-assumed situation. They mistrusted the AI-augmented systems’ capacity to send correct notifications to infected individuals to instruct them to quarantine during an infectious disease outbreak in a timely manner. Emerging technology risks, such as data breaches or data set biases and government strategies of mandatory application of digital contact tracing, could further undermine public trust. Individuals and relevant experts called for an open debate or scrutiny, transparent procedures for data usage, and public consultation plans and privacy regulations.

Privacy and Confidentiality

Digital infectious disease surveillance was identified to pose considerable risks to an individual’s rights to privacy and confidentiality. Contact tracing that linked a potential infectious disease with patients’ movements, locations, or social networks was seen as a considerable threat to further disclosure of sensitive information. For example, personal social interactions and contact history, especially in the case of sexually transmitted infections or HIV, could be revealed. Individuals were concerned that third parties or malicious users might access large health data sets for profit and/or abuse. Disclosure of private information was seen to lead to business depletion, privacy invasion, and public discrimination and stigmatization. Data anonymity, robust encryption processes, and deidentified aggregate data were contended to be crucial to all data privacy and security procedures in compliance with data protection regulations.

Civil Rights

Data ownership rights for public health surveillance were regarded as a major ethical challenge. Some reported that authorities or institutions limited their rights to decide the adoption of digital infectious disease contact tracing or surveillance systems. Personal liberties were also impacted by movement mapping, cross-border sharing of personal health information, and frequent security checks using QR codes for filling in personal information. Countries using centralized contact-tracing apps and privacy-by-design apps were seen to have the potential to impose restrictions on civil liberties, which, in turn, impacted ethical engagement in digital health. Data co-ownership and strengthening of transparency were seen as helpful to encourage individuals to participate in data visualization, analysis, and knowledge translation, and balance the power dynamics among decision makers, researchers, and citizens.

Governance

Technical and ethical standards, as well as legislation and guidelines for infectious disease outbreaks, data security, protection of individual privacy, and avoidance of discrimination were considered poorly developed and incomplete. Technology companies were recommended to establish a mechanism to deal with inaccurate epidemic reporting and dissemination of misinformation, and were expected to develop rapid ethical assessment, training, and disciplinary measures for data collection or sharing. Governance institutions or bodies, such as national health commissions, medical councils, and company boards, were further recommended to provide appropriate oversight on the performance of algorithms and data usage. Some literature argued that an independent privacy audit was needed to secure a transparent approach for the public [33,36]. Both individuals and relevant experts also supported data sharing protocols to specify the scope and granularity of disclosure. An example provided was that personal names should not be publicly available, and personal information should only be gathered and shared within a period of time regulated by law and clearly justified based on population health needs.

Discussion

Challenges of Using AI Surveillance Systems for Infectious Diseases

The emergence of AI surveillance systems for infectious diseases promises tangible global public health benefits, but these are accompanied by significant ethical, political, and legal challenges, which span over a wide spectrum on 3 levels. Six main themes were generated from this systematic review, ranging from people’s awareness and knowledge of digital infectious disease surveillance systems and personal privacy on the individual level to organizational issues of maintaining data integrity and security, and the lens was extended upward to the societal level, involving public trust, civil rights, and the need for a governance framework with ethical oversights. Some of the challenges are inherent to public health practice and only heightened by the use of digital tools, and others, such as public trust and awareness of digital surveillance, algorithmic decision making, and data security, are specific to AI approaches and largely unprecedented. It is vital to consider these challenges to enhance individuals’ rights, privacy, public responsibilities, and optimal population health outcomes so that digital surveillance can tackle pandemics ethically.

Strengths and Limitations

Studies and position papers on the ethical implications of AI surveillance for infectious diseases have gained momentum since 2019 in response to the outbreak of the COVID-19 pandemic; the dearth of studies prior to this makes it challenging to reveal any time trends. Furthermore, most articles included in the review were discussion papers, limiting the ability to evaluate generalizability, transferability, and rigor or multiple stakeholders’ perspectives on the use of AI-augmented infectious disease surveillance. The 5 empirical studies had restricted sampling approaches both in terms of size and stakeholder representation (health professionals and community groups), with 4 of the studies being conducted in Australia alone. Few papers involved vulnerable or diverse populations, and none involved specific cultural or socioeconomic groups, further limiting the scope of the review. Furthermore, articles acknowledged the inherent strengths and weaknesses of the public health system operating at the location of the study.

Retributions Versus Common Good for Public Health and Long-term Impacts

The tensions between human and civil right discourses and the need for rapid public health responses are exacerbated with the use of BDA/AI in the context of a pandemic [2]. There are medicolegal and moral retributions and concerns arising from utilizing available epidemiological information through a highly effective and responsive infectious disease surveillance method that could protect local and, potentially, global communities from serious infectious disease outbreaks [3,18]. Conversely, routine collection and linkage of detailed personal information pose considerable risks for the violation of individual and civil freedoms of choice and privacy [19].

As of July 21, 2021, there have been 191 million confirmed cases and 4.11 million deaths globally from COVID-19 [43]. Given that sound public health interventions are predicated upon promoting and protecting the health of communities, timely, cost-effective, and socioculturally informed primary care interventions, advocacy, and empowerment with long-term impact evaluations are required. There are strong reasons to develop an ethical governance framework to support AI-augmented infectious disease surveillance to achieve these outcomes. The question of how to balance what is needed for the “good of public health outcomes” and human rights in pandemic crisis situations is highlighted in the literature reviewed [33,37].

Trust Building and Privacy Protection

Detection and notification of infectious disease outbreaks requires prompt accurate disease diagnosis and follow-up of infected individuals and their close contacts. An early AI-augmented warning system has exponential potential for implementing real-time, responsive, and adaptive calculations. This means that substantial personal information, such as names, ages, locations, and relevant heath data, will be accessed in a timely manner and collected by AI systems for calculation, analysis, and notification. Public opt-in and well-founded trust in the digital system, its implementation, and the governance framework are therefore essential factors regarding ethical issues at the organizational and societal levels for infectious disease surveillance, which could enable the acceptability and effectiveness of the system.

Even with optimal technical standards to maximally reduce the risks and consequences of data misuse, data safety, security, and integrity cannot be guaranteed. The role of AI-augmented infectious disease surveillance might only be supplementary for public health, while ethical issues are carefully observed. One of the most important challenges facing AI is to design and develop appropriate methods to deidentify personal information and protect privacy, yet a greater risk of false positive and false negative notifications exists with higher-level deidentified data [33].

Moreover, well-founded trust and confidence vary in different countries and between individuals. People in democratic countries tended to distrust AI-augmented surveillance systems by challenging them when personal information is reported to health authorities, often without appropriate informed consent [3]. In contrast, Chinese citizens expressed their trust in the Chinese government’s response to COVID-19, which they felt had been highly successful in controlling the spread of the virus through the use of mobile phone data combined with intensive testing and restriction programs [44]. Although there is no unified standard to establish trust, the need for effective, transparent, accountable, and independent oversight is very important.

Generation of an Ethical Framework and Global Governance Structures

Digital standards and guidelines for developing and evaluating the performance of infectious disease surveillance alone are insufficient. The scope of the ethical framework in Textbox 1 needs to be expanded globally. Cross-national and national governance structures; institutional systems with regulatory, medical, ethical, and legal frameworks; and benchmarking standards have essential roles to play in the development and deployment of these new health technology systems. However, this review identified the urgent need for an ethical framework to underscore all AI-augmented infectious disease surveillance systems [18,19,30]. Considering the rapid development of global trade supply chains, mass gatherings, and international travel, World Health Organization’s International Health Regulations in 2005 [45] outlined the cross-border implications of a pandemic response and provided a framework for sharing, monitoring, and evaluating information from the sources of infections [37]. In response to a pandemic emergency, cross-border sharing of personal health data is essential for tracing infectious disease patients and their contacts. At the same time, cross-border sharing further raises the heightened and unique risk for individual privacy and security breaches. In turn, a spectrum of actions regarding ethical, political, and legal implications must be framed within strict safeguards and needs to be mandated globally [37].

Implications

A systematic medical, ethical, and legal framework is necessary for governance of AI-augmented infectious disease surveillance and the protection of personal privacy and data integrity. Public health systems should maximally increase the social awareness of AI surveillance and BDA for infectious diseases, and implement new technologies for infectious disease surveillance in a more person-centered and humane manner. Future research needs to focus on the setting up and implementation of an AI-augmented infectious disease surveillance system underscored by an ethical framework based on universal human rights. Decision makers should take into account varying and diverse population needs, sociocultural status, and regulatory and legal governance in order to promote trust building between end users, including infectious disease patients, doctors, and AI system implementers. This systematic review is intended to contribute to the development of a more comprehensive and concrete ethical framework for AI-augmented infectious disease surveillance, which will enable it to ultimately maximize public health responsiveness synergized within an ethical context.