01 June 2023: Editorial
Editorial: Infectious Disease Surveillance Using Artificial Intelligence (AI) and its Role in Epidemic and Pandemic PreparednessDinah V. Parums1A*
Med Sci Monit 2023; 29:e941209
ABSTRACT: Artificial intelligence (AI), or machine learning, is an ancient concept based on the assumption that human thought and reasoning can be mechanized. AI techniques have been used in diagnostic medicine for several decades, particularly in image analysis and clinical diagnosis. During the COVID-19 pandemic, AI was critical in genome sequencing, drug and vaccine development, identifying disease outbreaks, monitoring disease spread, and tracking viral variants. AI-driven approaches complement human-curated ones, including traditional public health surveillance. Preparation for future pandemics will require the combined efforts of collaborative surveillance networks, which currently include the US Centers for Disease Control and Prevention (CDC) Center for Forecasting and Outbreak Analytics and the World Health Organization (WHO) Hub for Pandemic and Epidemic Intelligence, which will use AI combined with international cooperation to implement AI in surveillance programs. This Editorial aims to provide an update on the uses and limitations of AI in infectious disease surveillance and pandemic preparedness.
Keywords: Editorial, Artificial Intelligence, machine learning, infectious disease, Surveillance
Artificial intelligence (AI), or machine learning, is an ancient concept based on the assumption that human thought and reasoning can be mechanized. In 1950, Alan Turing published a paper describing how it might be possible for a machine to ‘think’ . Since the 1950s, AI research has become an established academic discipline within philosophy, mathematics, engineering, physics, and the biological sciences, with debate regarding its potential uses and hazards. Recently, the rise of large language models (LLMs), or chatbots, such as Chat Generative Pre-trained Transformer (ChatGPT) (www.chatgptonline.ai), developed by OpenAI and released in November 2022, has raised concerns about their effects on education, the arts, social media, and publishing . However, in a recent article on the history of AI in medicine, Haug and Drazen highlighted the practical importance and benefits of AI techniques, particularly in image analysis and clinical diagnosis . Notably, during the COVID-19 pandemic, AI was critical in genome sequencing, drug and vaccine development, identifying disease outbreaks, monitoring disease spread, and tracking viral variants .
One of the first routine applications of AI in clinical medicine was in diagnostic histopathology, which relies on pattern recognition of cells and tissue components, and immunohistochemistry, which can be quantified objectively, rather than subjectively, ‘by eye.’ Image analysis methods have been used in diagnostic histopathology for over 40 years. However, in the past decade, advances in personalized or precision oncology have required the support of predictive assays that can select patients for optimal treatment and have driven the use of AI in histopathology diagnosis . It is now possible to digitize images of whole histopathology slides of tissue and use AI to identify subtle cell and tissue characteristics and the expression of cell biomarkers to bring precision to diagnosing malignancy and inflammatory and infectious diseases . AI image analysis has also driven the development of regulatory-approved companion diagnostics .
Local hospital-based infection outbreaks are controlled by tracing and controlling the identified transmission routes. Hospitals use local and national guidelines and infection control procedures to detect local infection outbreaks. Localized hospital-based outbreak detection may require a manual review of patient notes and contact tracing, which can be labor-intensive, and early detection and control may not be possible. Whole-genome surveillance sequences can identify genetic similarities in infectious organisms but do not identify the source and spread of infection. In 2022, Sundermann and colleagues reported developing and applying the Enhanced Detection System for Healthcare-Associated Transmission (EDS-HAT) . EDS-HAT combines infection surveillance using whole-genome sequencing (WGS) surveillance with machine learning of patient electronic health record (EHR) data to identify outbreaks and transmission routes, which might otherwise be missed . Between November 2016 and November 2018, at nine hospital sites affiliated with the University of Pittsburgh, healthcare-associated bacterial pathogens were identified, which prevented multiple outbreaks, and their routes of transmission were detected . This form of AI learning highlights using electronic medical records to guide clinical algorithms that can link infections [6,7]. The findings from the EDS-HAT study showed that real-time machine learning based on the combination of electronic medical records and whole-genome sequencing prevented up to 40% of hospital-borne infections in the nine locations studied . Also, it was possible to identify similar infections at different locations and periods . The authors concluded that using EDS-HAT for infection surveillance could save hospital costs and improve patient safety .
AI has supported infectious disease surveillance methods to identify pathogens and their variants and allows for a rapid response with the most effective treatments or public health measures . However, surveillance data from clinical symptoms and signs (syndromic data) alone can be unreliable. For example, the initial symptoms and signs of COVID-19 can be similar to rhinovirus, enterovirus, or influenza virus infections . Infectious disease surveillance also includes the detection of antibiotic resistance. As with histopathology image analysis, AI image analysis has been applied to determine antibiotic resistance . Microbiology disk-diffusion tests determine bacterial susceptibility to antimicrobials but rely on subjective analysis . Infection surveillance includes the identification of bacteria as susceptible or resistant to antimicrobials . Automated readers can be costly and unavailable to laboratories operating on a small budget . AI algorithms can now determine the antibiotic susceptibility of bacteria using image-processing algorithms of the disks to determine the required antibiotic type and measure the growth inhibition zones . The results can be forwarded automatically to the Global Antimicrobial Resistance Surveillance System of the World Health Organization (WHO) .
The COVID-19 pandemic spread rapidly worldwide and required a rapid response from various public health tools, including new and long-standing AI solutions for the surveillance of infectious diseases. Lessons learned have shown that AI may be used in early-warning systems for infection [10,11], detection of new disease outbreaks [12,13], epidemiological tracking, tracing, and forecasting [14,15], and the allocation of healthcare resources . The volume of health communications from the news, local reports, and online publications on infections from outbreaks of new infectious diseases is too large and diverse for groups of epidemiologists to analyze. However, AI algorithms and analytics can be used to track emerging pathogens with possible pandemic potential . For example, for more than a decade, the internet-based infectious disease surveillance system, HealthMap (www.healthmap.org/) has identified emerging infectious disease outbreaks, including influenza A (H1N1) . In real-time, HealthMap uses a continuously expanding dictionary in more than nine languages to extract geographical data that links multiple reports to identify clusters of infectious diseases that global public health authorities in different areas may have missed . On December 30, 2019, HealthMap raised a warning of – “a cluster of cases of pneumonia of unknown cause” – just days after the first case of COVID-19 due to infection with SARS-CoV-2 was identified .
During the COVID-19 pandemic, epidemiological metrics at the population level had limited predictive value for the true prevalence of SARS-CoV-2 in asymptomatic travelers between countries. Also, there were differences in the profiles for infection risk in people from different countries . In 2021, Bastani and colleagues reported designing and using a reinforcement AI learning system called Eva, developed in Greece in 2020, to identify individuals at high risk of infection when traveling across national borders . During the summer of 2020, Eva was deployed across all Greek borders to reduce the number of asymptomatic SARS-CoV-2-infected individuals . Due to the country’s limited viral testing resources, Eva assigned infection risk for travelers coming into Greece based on demographic information and testing results from previous travelers . Eva, the AI risk prediction method identified 1.85 times as many asymptomatic, infected travelers as random surveillance testing . Border control policies are an important part of infectious disease surveillance, requiring analysis of population-based metrics . Eva was one of the first examples of how AI reinforcement learning and real-time data that could adapt and improve could be applied to safeguard public health during a pandemic.
Therefore, AI-driven approaches complement human-curated ones, including traditional public health surveillance. The field of AI in clinical medicine is evolving. However, despite its many limitations and concerns, AI may bring the most apparent medical benefits in diagnosis and infectious disease surveillance. AI methods can help identify infection outbreaks from reports of symptoms, online posts on social media, contact with healthcare professionals, and evaluate electronic medical health records . Passive AI surveillance methods can be used to identify a lack of adherence to pharmaceutical and nonpharmaceutical interventions, including facial image-recognition algorithms to monitor compliance with mask-wearing . Facial image-recognition algorithms can also be used to monitor population movements and monitor social distancing .
In 2022, a machine-learning/AI study was undertaken using data from the Taiwan National Health Insurance Research Database (NHIRD) . The study aimed to identify the cause of influenza-like illnesses in 83,227 hospital admissions and the effects of age, gender, and 19 comorbidities and outcomes . A risk analysis was undertaken for severe disease and patient mortality and supported the role of AI/machine learning technologies in emerging infectious disease outbreaks . The prediction models based on decision-tree analysis were comparable with deep neural network models in predicting influenza-like illness severity . This study supported that a machine-learning/AI prediction decision tree could be used to facilitate clinical decisions and the allocation of medical resources at different stages during epidemics of emerging infectious diseases .
However, there are many limitations and concerns with using AI in infectious disease surveillance, and AI disease-tracking systems must be supplemented by molecular testing . Constant recalibration of AI monitoring will be required as new pathogen variants emerge, and variables such as vaccination will modify demographic characteristics and disease presentation . Also, these AI systems will never be completely accurate and may result in false alarms or failure to identify important epidemiological signals . Also, AI and machine-learning algorithms based on low-quality data could be misleading or result in harmful clinical recommendations due to systematic data selection bias or underrepresentation of selected populations . For example, in an analysis of COVID-19 mortality data from the US, a lack of adequately encoded racial information in surveillance databases resulted in disparities in mortality rates between Black and Hispanic patients, which resulted in underreported mortality rates by up to 60% .
Data privacy is a significant concern regarding using AI in infectious disease surveillance. More individuals are now using wearable health technology, connected health devices, and smartphones that may be linked to open social media . As more individuals interact with the digital world, approaches to the preservation of privacy should be a priority. One suggested approach to controlling this limitation is AI-powered, privacy-preserving technology, or federated learning, which can be used with smartphones . Federated learning ensures that participants’ data remain on their own devices, which avoids the privacy risks associated with centrally stored data .
There will likely be future pandemics of infectious diseases. Experience with AI during the COVID-19 pandemic has shown that although AI can contribute to infectious disease surveillance, it still cannot replace the cross-jurisdictional collective intelligence required to prevent and combat novel and emerging infectious diseases. Preparation for future pandemics will require the combined efforts of collaborative surveillance networks, which currently include the US Centers for Disease Control and Prevention (CDC) Center for Forecasting and Outbreak Analytics and the World Health Organization (WHO) Hub for Pandemic and Epidemic Intelligence, which will use AI combined with international cooperation to implement AI in surveillance programs [27,28].
AI has now become an established technology in all areas of medicine. The limitations and privacy concerns should be recognized and addressed with the development of guidelines for AI use. The recent COVID-19 pandemic and infectious disease outbreaks have highlighted the importance of this technology. Predictions for future developments in infectious disease surveillance will include emerging technologies, such as quantum computing, biosensors, augmented intelligence, and large language models, which can analyze large amounts of unstructured text, streamline labor-intensive processes, and identify trends in emerging infectious diseases.
1. Turing AM, Computing machinery and intelligence: Mind, 1950; 59; 433-60 Available at: https://academic.oup.com/mind/article/LIX/236/433/986238
2. Sallam M, ChatGPT utility in healthcare education, research, and practice: Systematic review on the promising perspectives and valid concerns: Healthcare (Basel), 2023; 11(6); 887
3. Haug CJ, Drazen JM, Artificial intelligence and machine learning in clinical medicine, 2023: N Engl J Med, 2023; 388(13); 1201-8
4. Abubaker Bagabir S, Ibrahim NK, Abubaker Bagabir H, Hashem Ateeq R, COVID-19 and artificial intelligence: Genome sequencing, drug development and vaccine discovery: J Infect Public Health, 2022; 15(2); 289-96
5. Bera K, Schalper KA, Rimm DL, Velcheti V, Madabhushi A, Artificial intelligence in digital pathology – new tools for diagnosis and precision oncology: Nat Rev Clin Oncol, 2019; 16(11); 703-15
6. Sundermann AJ, Chen J, Kumar P, Whole-genome sequencing surveillance and machine learning of the electronic health record for enhanced healthcare outbreak detection: Clin Infect Dis, 2022; 75(3); 476-82
7. Yu K-H, Beam AL, Kohane IS, Artificial intelligence in healthcare: Nat Biomed Eng, 2018; 2; 719-31
8. Maharaj AS, Parker J, Hopkins JP, The effect of seasonal respiratory virus transmission on syndromic surveillance for COVID-19 in Ontario, Canada: Lancet Infect Dis, 2021; 21; 593-94
9. Pascucci M, Royer G, Adamek J, AI-based mobile application to fight antibiotic resistance: Nat Commun, 2021; 12; 1173
10. Freifeld CC, Mandl KD, Reis BY, Brownstein JS, HealthMap: Global infectious disease monitoring through automated classification and visualization of Internet media reports: J Am Med Inform Assoc, 2008; 15; 150-57
11. Lim S, Tucker CS, Kumara S, An unsupervised machine learning model for discovering latent infectious diseases using social media data: J Biomed Inform, 2017; 66; 82-94
12. Al Hossain F, Lover AA, Corey GA, FluSense: A contactless syndromic surveillance platform for influenza-like illness in hospital waiting areas: Proc ACM Interact Mob Wearable Ubiquitous Technol, 2020; 4; 1-28
13. Mollalo A, Mao L, Rashidi P, Glass GE, A GIS-based artificial neural network model for spatial distribution of tuberculosis across the continental United States: Int J Environ Res Public Health, 2019; 16; 157
14. Reich NG, McGowan CJ, Yamana TK, Accuracy of real-time multi-model ensemble forecasts for seasonal influenza in the U.S: PLoS Comput Biol, 2019; 15(11); e1007486
15. Liu D, Clemente L, Poirier C, Real-time forecasting of the COVID-19 outbreak in Chinese provinces: Machine learning approach using novel digital data and estimates from mechanistic models: J Med Internet Res, 2020; 22(8); e20285
16. Dantas LF, Peres IT, Bastos LSL, App-based symptom tracking to optimize SARS-CoV-2 testing strategy using machine learning: PLoS One, 2021; 16(3); e0248920
17. Alavi A, Bogu GK, Wang M, Real-time alerting system for COVID-19 and other stress events using wearable data: Nat Med, 2022; 28; 175-84
18. Brownstein JS, Freifeld CC, Madoff LC, Influenza A (H1N1) virus, 2009 – online monitoring: N Engl J Med, 2009; 360; 2156
19. Bhatia S, Lassmann B, Cohn E, Using digital surveillance tools for near real-time mapping of the risk of infectious disease spread: NPJ Digit Med, 2021; 4; 73
20. Bastani H, Drakopoulos K, Gupta V, Efficient and targeted COVID-19 border testing via reinforcement learning: Nature, 2021; 599; 108-13
21. Guo P, Zhang J, Wang L, Monitoring seasonal influenza epidemics by using internet search data with an ensemble penalized regression model: Sci Rep, 2017; 7; 46469
22. Loey M, Manogaran G, Taha MHN, Khalifa NEM, A hybrid deep transfer learning model with machine learning methods for face mask detection in the era of the COVID-19 pandemic: Measurement (Lond), 2021; 167; 108288
23. Chiu HR, Hwang CK, Chen SY, Machine learning for emerging infectious disease field responses: Sci Rep, 2022; 12(1); 328
24. Antonelli M, Penfold RS, Merino J, Risk factors and disease profile of post-vaccination SARS-CoV-2 infection in UK users of the COVID Symptom Study app: A prospective, community-based, nested, case-control study: Lancet Infect Dis, 2022; 22; 43-55
25. Labgold K, Hamid S, Shah S, Estimating the unknown: greater racial and ethnic disparities in COVID-19 burden after accounting for missing race and ethnicity data: Epidemiology, 2021; 32; 157-61
26. Sadilek A, Liu L, Nguyen D, Privacy-first health research with federated learning: NPJ Digit Med, 2021; 4; 132
27. Center for Forecasting and Outbreak Analytics: 2023 Annual Report Available at: https://www.cdc.gov/forecast-outbreak-analytics/pdf/cdc-cfa-annual-report-2023.pdf
28. Morgan O, Pebody R, The WHO Hub for Pandemic and Epidemic Intelligence; Supporting better preparedness for future health emergencies: Euro Surveill, 2022; 27(20); 2200385
28 Sep 2023 : Clinical ResearchTrends in Hospitalization Duration for Rare Diseases: A Retrospective Analysis of Sarcoidosis, Still's Dise...
Med Sci Monit In Press; DOI: 10.12659/MSM.941536
27 Sep 2023 : Laboratory ResearchEnhanced Proliferation and Adhesion Marker Gene Expression in Fibroblast Cells: Evaluating the Efficacy of ...
Med Sci Monit In Press; DOI: 10.12659/MSM.941641
26 Sep 2023 : Clinical ResearchHypothalamic Injury and Clinical Outcomes in Patients with Post-Traumatic Hyponatremia: A Diffusion Tensor ...
Med Sci Monit In Press; DOI: 10.12659/MSM.942397
Most Viewed Current Articles
13 Nov 2021 : Clinical ResearchAcceptance of COVID-19 Vaccination and Its Associated Factors Among Cancer Patients Attending the Oncology ...
Med Sci Monit 2021; 27:e932788
30 Dec 2021 : Clinical ResearchRetrospective Study of Outcomes and Hospitalization Rates of Patients in Italy with a Confirmed Diagnosis o...
Med Sci Monit 2021; 27:e935379
14 Dec 2022 : Clinical ResearchPrevalence and Variability of Allergen-Specific Immunoglobulin E in Patients with Elevated Tryptase Levels
Med Sci Monit 2022; 28:e937990