Technological trends in epidemic intelligence for infectious disease surveillance: a systematic literature review

Data analysis and decision support

ML algorithms can estimate the size and location of infections and analyze key variables such as transmission methods, incubation periods, symptoms, and treatment resistance (Kaur et al., 2021). ML methods like convolutional neural networks, transfer learning, support vector machines, random forest, deep learning, and gradient boosting machine learning, are instrumental in predicting geospatial risks associated with outbreaks and the spread of epidemics. By analyzing multi-dimensional data that includes regional data on past outbreaks, environmental factors, travel patterns, social influences, vector distribution, and meteorological conditions such as temperature and rainfall, these models can accurately forecast the occurrence and timing of the outbreaks. This capability provides a valuable framework for early response and preparation in public health (MacIntyre et al., 2023).

ML has advanced greatly over the years and is now widely applied across various sectors. However, the success of ML depends on making optimal choices, including selecting the right algorithms, preprocessing methods, and hyperparameter settings for each dataset. Achieving adequate performance from a machine learning system requires expertise, yet even skilled professionals frequently rely on time-consuming and costly trial and error to identify the most effective methods for specific datasets (Karaarslan & Aydın, 2021).

Another successful evidence in the usage of AI in controlling epidemics is the outbreaks caused by mosquitoes such as Dengue, Chikungunya and Zika as human capabilities alone will not be enough to detect all the mosquito vectors in the affected area. During the outbreaks, AI assisted in identifying mosquito populations in the area to help improve epidemic management (Ganasegeran & Abdulrahman, 2020).

Predictive modeling for outbreak detection

AI has been helpful in handling various challenges faced by the healthcare sector such as the predictive capabilities for outbreak detection (Comito & Pizzuti, 2022). Predictive modeling uses mathematical models and formulas to calculate and predict the likelihood of the emergence and spread of epidemic (Firouzi et al., 2021; Raina MacIntyre et al., 2023). Chae, Kwon & Lee (2018) compared the long short-term memory (LSTM) model with deep neural network (DNN) models for predicting past outbreaks such as chicken pox and malaria. Additional factors that have been considered in the study include Naver search frequency, Twitter mentions, and average daily temperature and humidity. The findings show that the LSTM model performed better in disease outbreak while DNN models show consistent result whether there is an outbreak or not which makes the LSTM model is being widely used in predictive modeling (Zeng, Cao & Neill, 2020).

AI-based surveillance for early detection and prediction

AI assist health professionals by using advanced technological tools to analyze data and predict the spread of the outbreak. The COVID-19 pandemic accelerated the use of AI in epidemic surveillance. AI tools can help identify high risk locations and the duration of the outbreak such as the utilization of ML to detect outbreaks in the early stages. ML helps generate early warnings by analyzing open sources data which is beneficial in all places especially low-income countries that mostly depend on the traditional surveillance (Firouzi et al., 2021; Lalmuanawma, Hussain & Chhakchhuak, 2020; MacIntyre et al., 2023; Sim & Cho, 2021). The collection and analysis of the open sources data can quickly identify the patterns of the outbreak and take preventive measures to minimize the risk of the disease spreading (Firouzi et al., 2021; MacIntyre et al., 2023). Table 2 highlights the data analysis and predictive modeling approaches in various infectious disease surveillance systems.

EIOS uses NLP and text-mining to process millions of multilingual news to quickly identify health risks and improve communication between health professionals (MacIntyre et al., 2023). EIOS collaborates with the Index for Risk Management (INFORM) to identify and alert locations at risk to enable timely response and intervention. EIOS is also widely used for event-based monitoring for early detection and improve preventive measures (Spagnolo et al., 2020).

HealthMap was developed in 2006. It utilized Fisher–Robinson Bayesian filtering within a Linux/Apache/MySQL/PHP framework, integrating tools like Google Maps, GoogleMapAPI, Google Translate API, and a single AJAX library for efficient operation. It uses text processing algorithms to quickly identify, classify and overlay information on interactive map. HealthMap processed approximately 80 infectious disease alerts daily to provide real time insights for health professionals. HealthMap also successfully detect unidentified pneumonic cases in Wuhan Province outbreak which later widely known as COVID-19 a day before it was officially announced by the Chinese government (MacIntyre et al., 2023).

Next, BioCaster was launched in 2008. It utilized an ontology-based text mining system which is used to detect early outbreaks and disease monitoring (Collier et al., 2008). It then relaunched in 2021 during COVID-19 with improved features such as the inclusion of NLP to process multilingual news daily and was used for epidemic surveillance and predictive modeling. BioCaster is a web-based system with a friendly user interface which visualize real time statistics and outbreak trends (Gu et al., 2022; Meng et al., 2022).

BlueDot was developed in 2009 for H1N1 influenza pandemic. It accurately predicts the movement of the virus by using air travel data. In 2014, BlueDot was utilized to predict the Ebola Outbreak in West Africa (Allen, 2016). BlueDot also predicted COVID-19 nine days prior to the official announcements of COVID-19 and successfully predicted high risk countries. It analyzes over 10,000 data using ML and over 60 languages using NLP daily to identify high risk areas and alert health professionals in real time, allowing them to respond quickly and take preventive measures before the outbreaks escalate further (Allam, 2020).

Metabiota utilized ML and NLP to predict infectious disease outbreaks. NLP is used to categorize data into clusters such as frequency, severity and duration of the outbreak. During COVID-19, it successfully predicts high risk neighboring countries such as Japan, Thailand and Hong Kong by taking human behavior and fear-driven responses into account. It also accurately foresees the spread of the COVID-19 virus 1 week before it was officially reported in these countries (Allam, 2020; Tong, 2020).

Data sharing

The existing systems manage unstructured COVID-19 data which leads to difficulties in data interoperability. Unstructured data hinders collaboration opportunities to fight against the pandemic (Kaur et al., 2021). Thus, implementing blockchain technologies can help reduce the unstructured data, as blockchain supports better information sharing and provide a structured data view, enhancing the efforts to control the spread of COVID-19 (Ahmad et al., 2023; Firouzi et al., 2021; Karaarslan & Aydın, 2021; Nguyen et al., 2021).

Blockchain technology creates a secure, constant record of transactions across a wide network across geographical settings through hashing algorithms and asymmetric cryptography (Ahmad et al., 2023). Each block in the blockchain is linked to the previous block by using cryptographic hashing, which makes the data unchangeable. There are two main types of blockchain which are permissionless and permissioned. In permissionless blockchain, anyone can participate in doing any transactions and mining data, while permissioned blockchain, is restricted to only a specific group of users chosen by the organization (Ahmad et al., 2023; Chattu et al., 2019).

Countering misinformation

Blockchain are being used widely in epidemic intelligence to combat misinformation by tracking the origin of the data and identifying unauthorized changes to the data. This ensures the reliability and accuracy of the data for public health professionals. One of the blockchain-based platforms, MiPasa supports sharing reliable data, identifying errors and misreporting by integrating data from various trusted open sources data such as the WHO and registered health organizations. It utilized data analytics and verified blockchain data to facilitate identifying hotspot locations and virus carriers while also protecting user anonymity and data privacy (Ahmad et al., 2023; Marbouh et al., 2020; Nguyen et al., 2021).

Blockchain can also verify news authenticity by registering, evaluating, and sorting news based on the credibility of the sources. Blockchain aids public health professionals in decision making which facilitates them to quickly foresee future outbreaks and implementing lockdown in affected areas (Ahmad et al., 2023; Comito & Pizzuti, 2022).

Contact tracing

A countless of mobile applications have been developed for infectious disease surveillance such as during the COVID-19 outbreak. However, the challenges arrive when using a centralized architecture as shown in Table 3, which means all the data is stored and managed in a single location. An example is Singapore’s TraceTogether, which uses Bluetooth to determine if someone has been in close contact with a COVID-19 positive person (Ahmad et al., 2023; Firouzi et al., 2021). Centralized systems raise concerns about data privacy, as authorities can access user information. They are also at risk of data tampering, fraud, or loss and are considered less reliable due to the potential for a single point of failure (Firouzi et al., 2021). Additionally, such systems may limit cooperation between different entities like healthcare, government, and law enforcement, and lack transparency, traceability, and permanence in data management during the pandemic (Ahmad et al., 2023).

Blockchain technology safeguards user data privacy through pseudo-anonymity (Behnaminia & Samet, 2023). Digital contact tracing can utilize regular expression matching techniques on a blockchain platform to store social interaction data securely, ensuring only authorized users can access it (Ahmad et al., 2023).

Outbreak detection

Blockchain technology facilitates immediate data exchange among healthcare providers, government bodies, and various stakeholders, enabling the quick identification of irregular patterns in disease occurrences (Hang et al., 2022; Upadrista, Nazir & Tianfield, 2023; Xie et al., 2021).

Bayesian change point (BCP) was used for outbreak detection. The BCP analysis is not initially designed for infectious disease outbreaks; however, it has been utilized to detect the beginning of influenza epidemics by identifying significant shifts in distributional parameters before and after certain points in time series data. The BCP algorithm breaks down the time series into blocks, then calculates the mean and variance, followed by estimating the probability of each break point being a change point (Ahmad et al., 2023).

IoT devices in disease surveillance

Remote monitoring devices like wearable devices, IoT sensors, and mobile apps are valuable tools for healthcare providers to remotely track a patient’s health status (Sim & Cho, 2021). These devices can gather data on vital signs, respiratory rate, and oxygen levels, aiding in the early detection of health deterioration. Table 4 presents a detailed overview of the devices used for data collection and their applications in epidemic intelligence. The availability of this data, along with advanced intelligent algorithms, has the potential to advance medical practices focusing on disease prediction, prevention, and treatment (Firouzi et al., 2021; Rahman et al., 2022b; Stanko, 2023). The market has seen a rise in wearable devices for personal health and fitness which cause healthcare professionals to explore their use in continuous monitoring, management, and accessing patient physiological information within remote health monitoring systems (Stanko, 2023). mHealth encompasses various medical and public health practices supported by mobile devices like smartphones, patient monitoring tools, and wireless gadgets (Chi Minh City et al., 2021).

Advanced tracking devices like AI-based apps in bracelets and rings can effectively trace infected individuals, aiding in outbreak detection and control measures for diseases like COVID-19 and dengue (Rahman et al., 2022b). Various tracking applications and GPS systems are used to identify patients and implement efficient control measures. Additionally, technologies such as Radio Frequency Identification (RFID), Near-Field Communication System (NFC) and AI significantly enhance disease and vector surveillance, monitor environmental and social risks related to COVID-19 and dengue, and (Alo et al., 2022; Chi Minh City et al., 2021; Rahman et al., 2022b; Sim & Cho, 2021).

Garg et al. (2020) developed a model using passive RFID technology to anonymously track the movement and interaction of objects until someone tests positive for an infectious disease like COVID-19. This system doesn’t require mobile phones; instead, individuals and animals will wear RFID tags. Their innovative approach also incorporates blockchain technology to store the data securely, ensuring privacy through decentralized control. The RFID readers are placed in buildings or vehicles to collect proximity information, which is then recorded on a blockchain-based smart contract.

Wearables

Before and during the pandemic, Shausan, Nazarathy & Dyda (2023) mentioned that a few wearables such as the TempTraq skin patch, Fitbit smartwatches, and Oura Rings have been utilized for health monitoring and early detection of conditions such as COVID-19. These devices, which track temperature, heart rate, and other physiological data, have helped healthcare workers and research studies to monitor and analyze health trends. The integration of wearable device data with self-reported symptoms and other information has enabled the development of algorithms to detect and track the progression of COVID-19 (Shausan, Nazarathy & Dyda, 2023; Sim & Cho, 2021). EPIWATCH employs two AI systems; one uses NLP for named entity recognition to identify and extract details like disease, date, and location from reports. The second AI system helps in sorting outbreak reports by relevance, filtering out non-pertinent articles, to streamline the analysis and presentation of data to users and analysts (Firouzi et al., 2021; MacIntyre et al., 2023; Raina MacIntyre et al., 2023). Proximity and vision sensors are incorporated into wearables, robots, and smart devices to ensure adherence to social distancing measures. These sensor-based systems detect proximity and alert users with a warning alarm if social distancing rules are breached (Alo et al., 2022; Firouzi et al., 2021).

Mobile phone

Mobile applications help combat infectious disease by providing information on symptoms, treatment and preventive measures to their user (Firouzi et al., 2021; Kondylakis et al., 2020; Ye, 2020). It allows individuals to monitor their own health status, receive reminders for medication and vaccination and alert health professionals about potential outbreaks which allow them to take immediate action to areas that have limited access to medical care (Alo et al., 2022; Dami, 2022; Firouzi et al., 2021; Stanko, 2023). Improvement in mHealth technology and the widespread use of mobile phones have eased the monitoring of individual’s health behaviors. Individuals with long-term illnesses can now also stay at home instead of staying in the hospitals (Saheb & Izadi, 2019). These advancements also allow for more reliable applications that collects data from various sources for further research and practice in travel medicine (Lai et al., 2019).

Smart thermometer

In the initial stages of COVID-19 pandemic, governments used traditional thermometers for temperature checks at specific sites, but this method had limitations in consistently identifying infected individuals. Advancements led to the exploration of smart thermometers and wearable devices for continuous health monitoring (Tripathy et al., 2022). These devices, which track physiological changes to signal potential illness, had been previously used in the U.S. for flu surveillance and were adapted to detect COVID-19 by analyzing data trends and physiological markers. For example, data from smart thermometers helped predict COVID-19 hotspots, while wearables like smartwatches and fitness trackers were employed to detect early signs of the virus, utilizing metrics like respiratory and heart rate changes (Shausan, Nazarathy & Dyda, 2023).

Robotics

The integration of 5G communication technology with various technologies such as drones and aids in monitoring legal compliance within the transport system. It also supports thermal screening of individuals through thermal cameras, facilitating the instant sharing of date and time-specific information (Moglia et al., 2022; Rahman et al., 2022b). Given the rapid spread of viruses, manual contact tracing is often insufficient. However, the speed and efficiency offered by 5G technology can significantly improve control measures (Pillai & Kumar, 2021).

AI-based imaging and robotic applications have become essential tools for diagnosis, treatment, and daily life support during the pandemic (Firouzi et al., 2021; Jiao et al., 2023). Imaging techniques in AI play a vital role in detecting and managing COVID-19, with contributions like masked face recognition and fever detection aiding in tracing cases and contacts (Firouzi et al., 2021; Jiao et al., 2023). The surge in video conferencing due to homeschooling and telework has led to increased use of face detection and portrait segmentation, enhancing convenience in these activities. Additionally, robotic technologies incorporating face detection, voice recognition, and sensors have been widely utilized for tasks such as examination, healthcare, monitoring, disinfection, cleaning, delivery, and logistics, reducing human contact and enhancing daily life support. Hence, the development of intelligent robots is a crucial direction for the future (Firouzi et al., 2021; Gao et al., 2021; Jiao et al., 2023; Sarker et al., 2021).

Contact tracing

IoT devices are widely used in epidemic intelligence as a contact tracing tool. It is used to track the places visited by infected individuals and identify close contacts based on that. Close contact is being guided by health professionals to protect themselves and to prevent them from further spreading the virus. The utilization of contact tracing systems has also reached rural areas to track individual movements accurately. Contact tracing systems in urban areas utilize technologies like GPS, Bluetooth, and Wi-Fi networks for authorities to quickly identify potential exposures and implement necessary preventive measures (Ahmad et al., 2023; Alo et al., 2022; Firouzi et al., 2021; Menni et al., 2020; Pillai & Kumar, 2021). Other than that, NFC, a wireless communication protocol operates over a short range of up to 4 cm at speeds around 424 kbps, was also considered as an emerging technology in contact tracing (Alo et al., 2022). Although contact tracing can be challenging due to high mobility of individuals, it is important to identify infected individuals who should be quarantined and to determine containment zones for managing disease transmission (Pillai & Kumar, 2021; Sudre et al., 2021).

Disease mapping

GIS is important in visually displaying the factors that correlate with the disease occurrence (Sarwar et al., 2020). It analyzes spatial autocorrelation, a concept that helps understand how proximity impacts the likelihood of interaction and similarity among neighboring factors to understand the spatial relationship in the spatial epidemiology. GIS observes patterns between time and space to identify how infectious diseases are linked between the socioeconomic factors of different groups and the effectiveness of various prevention and control strategies. The advancements in geospatial analysis together with GIS tools enables better handling and analyzing vast amount of data (Bag et al., 2020; Sarfo & Karuppannan, 2020; Sarwar et al., 2020; Scarpone et al., 2020).

During COVID-19, geospatial technologies gained attention through web-mapping applications and dashboards as they provide visual graphics on the outbreak’s information which are important to inform the public, raise social awareness, and manage uncertainties in pandemic (Gangwar & Ray, 2021; Xu et al., 2022). Notable dashboards like HealthMap utilized online data for real-time surveillance to enhanced spatial resolution in specific areas while WHO offered near-real-time official data, updated approximately every 15 min. Real-time mapping was essential for risk assessment and to support decision-making. GIS systems not only used for disease mapping, but also as visual platforms that display real time data and track individuals at risk (Carballada & Balsa-Barreiro, 2021).

Web apps and platforms for contact tracing help analyze the spatial patterns and spread of the virus over time. Most of the countries developed their own system, such as China’s Close Contact Detector (Carballada & Balsa-Barreiro, 2021), Singapore’s Trace Together (Carballada & Balsa-Barreiro, 2021; Pillai & Kumar, 2021), and Switzerland’s SwissCoVID (Alo et al., 2022; Carballada & Balsa-Barreiro, 2021), highlighting the significance of geospatial analysis during the pandemic (Carballada & Balsa-Barreiro, 2021). This period urged advancements in geospatial analysis, with improved computational capabilities of GIS tools enabling better handling of large datasets and enhanced analytical precision (Carballada & Balsa-Barreiro, 2021).

GIS systems are more than just mapping tools; they serve as centralized platforms for near-real-time data management, aiding in tracking both reported COVID-19 cases and individuals at risk (Carballada & Balsa-Barreiro, 2021; Franch-Pardo et al., 2020; Gangwar & Ray, 2021; Malone et al., 2019).

Potential of geospatial artificial intelligence in health geographics

Geospatial artificial intelligence (GeoAI) is a combination of AI and GIS which is being used widely in spatial health analysis. It helps understand the disease spread by using data analysis techniques to track possible viruses that may cause infectious disease using location analytics and patient travel data to find effective measures to mitigate outbreak (Mir et al., 2021). GeoAI is also being utilized to monitor and forecast disease hotspots as it understands how disease spread geographically (Kamel Boulos, Peng & Vopham, 2019; Peng et al., 2021). It analyzes location data to identify factors at specific places that might affect health outcomes across different social groups. GeoAI also combined with IoT devices to integrate personal analytics by collecting data from devices such as mobile phones and wearables sensors. It integrates metadata and satellite imagery to enable geo-database in analysis ready format (Mir et al., 2021).

Data mining and machine learning algorithms like Python and Anaconda, together with geographic knowledge, facilitate accurate prediction of health and environmental threats (Sahana et al., 2023). Geostatistical and algorithmic methods are applied to identify risks from environmental and spatial factors, assessing their impact on infectious disease routes (Mir et al., 2021).