Using AI for Pandemic Response and Preparedness Strategies

There is a 1 in 4 chance of another COVID-like outbreak by 2033. Can artificial intelligence (AI) help in the next pandemic prevention, preparedness, and response?

 

AI encompasses machine learning (ML), deep learning (DL), natural language processing (NLP), large language models (LLMs), vision and sensory systems, decision aids, robotics, and other tools that have been implemented across healthcare, pharma, communications, and various business domains. 

The risk of another epidemic or pandemic can be reduced through a faster rollout of vaccines and medicines, stronger delivery infrastructures, and other preparedness strategies.

 

As viruses are emerging more frequently, researchers, governments, healthcare providers, pharmaceutical companies, and other players must understand the role of AI before the pandemics strike.

 

In this article, we’ll examine some applications of artificial intelligence techniques for pandemic prediction, mitigation, and management.

 

Let’s start with AI and pandemic preparedness.

 

Case 1. AI-powered Early Outbreak Detection

Early epidemic detection and forecasting are essential because they enable healthcare providers and policymakers to bolster research efforts, preparedness, and supplies. AI has emerged as a crucial ally thanks to its ability to process data and extract invaluable insights from extensive datasets quickly. 

 

Nowadays, organizations can access data from diverse sources, e.g., hospital records, lab results, social media trends, mobility data, and more, and apply various artificial intelligence approaches for pandemic detection. For example:

 

• ML algorithms can detect patterns and anomalies in data that may indicate the onset of an outbreak before traditional reporting systems do.  
• AI models can be pre-trained on data related to particular viruses to identify emerging threats from new strains.  
• Solutions based on convolutional neural networks (CNNs) can help distinguish between cases of different infections that may appear similar to the human eye.
• Graph neural networks (GNNs) can improve the forecasting of infectious disease dynamics.

 

For instance, GNN models accurately predicted COVID-19 cases per region and influenza-like illness rates.

 

In 2019, the Canadian start-up BlueDot was among the first to point to a respiratory illness outbreak in Wuhan, China. BlueDot utilizes NLP and ML algorithms to sift through news reports in 65 languages, airline data, and animal disease networks to detect outbreaks and anticipate the spread of disease.

 

These insights are reviewed and verified by epidemiologists and then shared with public health officials, airlines, and hospitals to help predict and manage risks.

 

Read also: Hospital Management Software: The Complete Guide

 

The University of California, Irvine (UCI) and the University of California, Los Angeles (UCLA) have jointly created an AI-based early warning system (EWS) for pandemic predictions based on social media data. The project builds on a searchable database of 2.3 billion tweets collected since 2015.

 

The tool works by identifying meaningful tweets and training the ML algorithm to categorize significant events that may be indicative of upcoming outbreaks. It can help evaluate the potential outcomes of specific public health policies and even the impact of treatments on the spread of viruses.

 

Read also: Text Classification: Business-Changing Use Cases

 

Case 2. Predictive Modeling of Epidemics

Predictive models can forecast how a disease will spread, the severity of the outbreak, and the resource demands. This information will empower agencies in charge to act proactively with targeted interventions, public health measures, and optimized vaccine distribution and resource allocation.

 

Epidemiological modelling leverages ML methodologies to analyze massive volumes of demographic data, travel behaviours, healthcare records, and environmental variables.

For example, AI-driven models like SIR (Susceptible-Infectious-Recovered) and SIS (Susceptible-Infectious-Susceptible) help forecast the spread of infections. Others can predict disease transmission patterns, pinpoint vulnerable populations, and evaluate the effectiveness of intervention strategies.

 

Read also: Why and How to Achieve Interoperability in Healthcare IT

 

Epidemiologists also utilize sophisticated computational tools to simulate hypothetical scenarios, evaluate the effects of public health policies, and provide informed recommendations to policymakers to curb the spread of infections.

 

Conventional mechanistic and semi-mechanistic disease transmission models offer valuable insights into virus transmission and are utilized to develop counterfactual scenarios.

 

Case 3. AI Improving Healthcare Delivery and Management

The Coalition for Epidemic Preparedness Innovations (CEPI) intends to utilize artificial intelligence in pandemic preparedness and response strategies. Applied to various activities, it can accelerate and enhance preparation for the next pandemic.

 

Before a crisis, artificial intelligence can help optimize hospital workflows, enhance healthcare worker training, forecast patient demand, and support preventive care through predictive analytics. 

 

Read also: Healthcare CRM Development: Key Features & Cost

 

During an epidemic, healthcare systems will face challenges in operating and allocating limited resources in a rapidly changing situation, all while under heavy time pressure.

 

AI tools may enable them to streamline operations, improve decision-making and planning of data-driven responses, assist with real-time triage, or optimize resource allocation and staff scheduling.

In 2023, the UK’s Accelerated Capability Environment (ACE) explored the use of AI for pandemic response in that area together with the Kettering General Hospital (KGH), the NHS England AI Lab, and Faculty from the ACE’s Vivace supplier community.  

 

They built an ML-supported patient flow tool around a virtual hospital environment modelled on a selection of KGH wards. The tool utilizes historical admission data to plot immediate demand against predicted demand, generates bed suggestions for emergency patients, and even explains its recommendations so that staff can make the final decision.

The proof of concept demonstrated more efficient and effective bed scheduling, improved patient outcomes, and cost reduction due to fewer overall bed moves.

Read also: Understanding the AI Proof of Concept: A Complete Guide

 

AI-powered tools also aid in remote patient monitoring, reducing hospital overload and exposure risks for healthcare workers.

 

Read also: Telemedicine App Development: Types, Features & Costs

 

Case 4. AI-driven Public Health Decisions

During a pandemic, policymakers and authorities will use AI to get real-time, evidence-based insights for better-informed decisions.

 

ML models that analyze infection rates, mobility patterns, genomic sequencing, vaccination coverage, social behavior, and other data across disciplines can forecast epidemic trajectories and evaluate the potential impact of pharmaceutical and non-pharmaceutical interventions, from lockdowns to vaccination campaigns.

 

AI models that utilize geographical and temporal data as inputs can help researchers model problems related to the spread of infectious diseases and their impact. AI can also run simulations to compare policy options, helping decision-makers balance health outcomes with economic and social considerations.

 

Case 5. AI Improving Public Health Communications

Organizations that employ AI for pandemic response communications will ensure the timely delivery of crucial information and tailored messages to diverse audiences.

 

By adopting AI-driven communication strategies and tools, public health authorities will be better equipped to build trust, combat misinformation, and encourage adherence to safety measures.

 

NLP tools can monitor social media, news, and forums to track public sentiment and identify knowledge gaps. For instance, Dr. Philip AbdelMalik of the World Health Organization (WHO) stated that AI can detect online advocacy of potentially dangerous treatments, allowing the WHO to intervene.

 

Then, generative AI tools will help design awareness campaigns and craft clear messages in multiple languages designed for specific communities or demographics.

 

Organizations should also explore the capabilities of large language models (LLMs) like ChatGPT or Google Gemini. During pandemics, AI-powered chatbots and virtual assistants can help disseminate helpful information, answer questions about prevention, vaccination, symptoms, testing, and treatment plans, and provide crisis helplines and mental support.

 

Read also: AI in Mental Health: Use Cases and App Ideas to Watch 

 

Custom chatbots can provide real-time updates and assistance to a particular entity’s customers, partners, and employees, reducing overload on the customer support while maintaining consistent messaging.

Read also: Onix’s Guide to Custom Chatbot Development

 

Case 6. AI-driven Drug and Vaccine Development

Accelerated search for vaccines and simplified clinical trials are truly life-saving applications of artificial intelligence in pandemic responses. 

 

AI‐based approaches can complement traditional procedures to help develop or repurpose drugs and design vaccines. AI technologies and methods are explored to select the candidates for developing vaccines and drugs and to identify potential drug candidates for repurposing against new diseases.

 

Read also: Clinical Trial Management Software: Building a SaaS Solution

 

DL is key to discovering information about the structure of proteins associated with viruses, which is necessary for developing vaccine formulae. Predictive ML techniques can identify compounds that can block the cellular pathway through which a virus enters human cells.

 

AI systems can also successfully identify thousands of molecules for possible medications and use simplified chemical sequences to predict a molecule’s ability to bind to a target protein.

 

One of the prominent methods for drug discovery is reverse vaccinology. It surveys episodes and antigens to identify possible targets without requiring “wet lab” experiments.

 

For example, in 2020, researchers at the University of Michigan, Ann Arbor, developed an ML-based tool they called Vaxign-ML to enhance the accuracy of predictions in their research. They applied it in conjunction with Vaxign (a vaccine target prediction and analysis system) to predict COVID-19 protein candidates for vaccine development.

 

AstraZeneca uses AI to help speed up the discovery of new antibodies that can be utilized in the development of novel vaccines. Its VP of data science and AI R&D said that it enables the company to generate and screen a library of antibodies and bring the highest-quality predictions to the lab, reducing the number of antibodies they need to test and “the time to identify target antibody leads from three months to three days.”

 

The Harvard Medical School and the University of Oxford developed EVEScape, an AI tool that makes predictions about new variants of coronavirus, HIV, influenza, and other viruses.

 

Used early in a pandemic, it can benefit vaccine manufacturers and researchers looking for therapeutics, “particularly antibodies to get some insight early on into which mutations might arise even a year in the future,” said one of the researchers involved in the project.

 

Read also: Generative AI in Pharma: Here's Why It’s a Must

 

Insilico Medicine created an entire AI-driven drug discovery pipeline. By utilizing aging as a means to identify disease, the company has trained its AI to identify new targets and molecules.

 

ML and other AI methods also promise the development of personalized immunizations and vaccines tailored to individual genetic compositions.

It’s also wise to explore the capabilities of artificial intelligence in pandemic responses unrelated to healthcare. 

 

Case 7. AI Supporting Supply Chains and Logistics during Outbreaks

For example, in 2020, researchers working with a large retail chain detected anomalies associated with panic-buying behaviors in their customers’ online orders, promotions, external online searches, and social media posts.

This approach enabled the company to adjust the distribution of essential products amid the COVID-19 pandemic. Data from March 2020 showed that the tool can help retailers increase access to essential products by 56.74%.

 

AI can also be used in predictive models to anticipate changes in demand and adjust distribution patterns accordingly. 

During pandemics, when people are displaced, logistics systems will face many challenges. Decision-making models can be trained on historical data from past epidemics or other emergencies to optimize the distribution of essential goods and vehicle routing.

 

Case 8. AI-assisted Public Health Measures

AI can help organizations and individuals implement recommended public health measures and monitor adherence. For example, intelligent systems can automatically identify individuals who are not wearing masks in public spaces, such as supermarkets, shopping malls, train stations, and airports. 

 

One such image classification system, based on a convolutional neural network VGG-16, was created in 2020.

 

Image source

 

The project results, including the correct classification of 90.2% of the validation images, are suitable for a wider practical application of the trained model.

 

Onix’s experts also develop cutting-edge video analytics tools for crowd behavior analysis, facial recognition, and object tracking. For example, a crowd video analysis system built for one of our clients can enhance public space and event security by detecting threats and risks and issuing alerts in real time. 

 

During future pandemics, LLM-powered virtual assistants and service robots, which are evolving rapidly, may play a vital role in maintaining safety and continuity.

 

These intelligent systems would enable people to access reliable, up-to-date information, receive personalized guidance, and even conduct routine health checks from the safety of their homes. Robots may deliver medicines, groceries, and other essentials, reducing the need for physical contact and lowering exposure risks for healthcare workers.

 

When it comes to AI and pandemic preparedness and responses, the applications are limited only by our imagination. However, the implementation of such solutions comes with many limitations and challenges.

 

Some Challenges of Leveraging Artificial Intelligence for Pandemic Preparedness and Response Strategies

 

Data-related restrictions and problems

AI that relies heavily on historical data sets for pattern recognition may not be the most suitable tool for detecting novel pandemics. 

Currently, there are fewer applications of AI in infectious disease epidemiology than in areas such as patient diagnosis, disease risk prediction, and decision support for healthcare professionals. It is difficult to obtain large-scale, standardized, and representative data essential for training and evaluating AI or ML models with variable parameters. 

 

Read also: How to Build Machine Learning Teams for AI Projects 

 

Epidemiology may utilize data from trial registries or cohorts, as well as medico-administrative, administrative, and environmental data. However, these data sources have limitations, including accessibility, particularly at the regulatory level, the need to construct them for population health studies, and questionable validity and reliability.

 

Read also: A Guide to HIPAA Compliance for Software Developers

 

During an epidemic, data can be collected from healthcare facilities, public health agencies, academic institutions, community organizations, and other sources that use different data formats, structures, and terminology, making it difficult to merge them seamlessly.

 

Standardizing data formats, definitions, and coding systems, as well as ensuring interoperability between disparate datasets, is essential to overcoming this challenge.

 

Information from social media also has limitations. For example, the UCI/UCLA’s AI-based EWS was focused on the US and reliant on X, which is not accessible in some countries. Before expanding the coverage to other regions, the team must overcome the data scarcity and potential bias.

 

Such datasets may also contain false information. Moreover, if the input data is representative only of specific subpopulations, the output will be representative of limited subpopulations or contain significant misinformation.

 

Learn more: 30 Major Machine Learning Limitations, Challenges & Risks

 

Privacy and security concerns

The last pandemic highlighted the need for a thorough investigation into the security and privacy of personal data.

 

Artificial intelligence techniques for pandemic prediction, detection, simulation, or response planning and monitoring may require the use of personal information, such as GPS location, CT scans, and daily activity records. However, people often resist sharing their personal information.

 

Read also: How AI Is Transforming CRM – And What It Means for You

Model development and validation challenges

AI models designed for epidemiology must accurately capture the complex dynamics of infectious diseases while accommodating the variability inherent in real-world data. Models that must provide reliable insights promptly require a delicate balance between speed and accuracy.

Moreover, it may be challenging to ensure that AI models trained on data from one region or demographic can effectively generalize to other diverse populations.

 

Experimental modelling is frequently labour- and resource-intensive and susceptible to ethical and practical limitations. For example, disease transmission models are associated with considerable computational costs, partly due to the extensive complexities involved in numerical methods and inference in a high-dimensional parameter space.

 

Technology limitations

Dr. In-Kyu Yoon, then Global Head of CEPI’s programmes and innovative technology, said that AI needs to develop and mature. Moreover, it is still up to people to determine where to apply it and to make decisions.

 

Ethical considerations

It may be challenging to implement robust ethics and risk assessment protocols during a crisis or when new technology is implemented at an unprecedented pace and scope, e.g., when a team uses AI to help speed up the discovery of new antibodies or drugs to save lives.

 

Rather than overlooking ethical concerns, technologists, scientists, medical professionals, and others must learn to put ethics into practice immediately. 

 

Detecting and mitigating biases related to gender, ethnicity, race, and other factors requires special attention to ensure that the developed AI models are fair, trustworthy, and beneficial for all.

Summing up the Role of AI before the Pandemics and during Outbreaks

The integration of artificial intelligence in pandemic management significantly influenced the number of survivors over time, paving the way for leveraging AI for pandemic preparedness and responses in the future. As BlueDot put it, “Don't fight tomorrow’s outbreaks with yesterday’s tools.”

 

Between epidemics, artificial intelligence contributes to disease research, early warning systems, the development of prevention methods and crisis strategies, and the optimization of healthcare systems’ operations. AI has significantly improved the predictability of pathogenic emergence and transmissibility.

 

The applications of AI for pandemic response purposes range from epidemiological modeling to accelerating vaccine development. Data-driven decision-making enables forecasting, mitigation, and response to an epidemic.

 

Overall, we are in a better position for the next pandemic, partly thanks to progress in AI technology. However, limitations in ML technology, data-related restrictions, ethical concerns, and other problems may hinder or slow down the use of AI in infectious disease research, outbreak prediction, prevention, and response. A Nature study emphasizes that the continued success of AI depends on data transparency and reduced training costs.

 

We would add “...and competent developers!” Onix can be your trusted partner in adopting AI and other technologies for your unique goals and needs. Contact us today to discuss your project or ideas, request a consultation, or learn more about our machine learning development services!

 

FAQ

Can artificial intelligence predict the onset of a pandemic in advance?

Yes, AI-based systems can identify patterns in data that signal the onset of epidemics and pandemics.

 

What AI techniques are used for detecting and forecasting infectious diseases?

Here are just several examples:

 

• NLP is used to detect and forecast outbreaks by analyzing electronic health records (EHR), social media, and other data sources.
• The unsupervised ML techniques of clustering and dimensionality can also be used to identify clusters of cases and patients with similar traits in EHR.
• ML and other AI tools have demonstrated accuracy, ranging from 84.3% to 100%, in detecting COVID-19 through the analysis of chest X-rays and CT scans.
• Computer vision can help identify early symptoms, such as high fever or coughing, in real-time surveillance video.

How is AI being applied in infectious disease research?

The use of AI in infectious disease research encompasses, but is not limited to

 

• applications of ML, computational statistics, information retrieval, and data science to infectious disease surveillance data and AI-driven infectious disease modelling to answer key epidemiological questions
• AI models applied to genomic data to elucidate virus lineages, viral origin, pathogenicity, transmissibility, and the pathogen’s potential to evade immune responses, improving the accuracy of phylogenetic inference
• AI algorithms providing insights into infection diagnosis, treatment prognosis, predicting antimicrobial resistance, generating novel antimicrobial compounds, and tailoring personalized antimicrobial therapies

 

Can AI accelerate the discovery of new antibodies and medicines?

Yes. Computational model-based design of drugs and vaccines facilitates faster and more cost- and labor-efficient discovery.