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].

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.

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).

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.

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.

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.

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.

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.

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].

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.

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].

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.