Blog

It is not a very distant memory when thousands were contracting the COVID-19 virus and succumbing to it—people scrambling to find ICU beds and oxygen cylinders on Twitter, Facebook, and WhatsApp.

The pandemic also resulted in many people losing their jobs, with the lockdowns forcing them to migrate from cities to their hometowns in search of livelihood and sustenance. 

This picture was not limited to India; the entire world faced a similar predicament. Although the world economy is gradually recovering, the loss of lives is irreversible.

Looking back to a year ago, it is evident that the world cannot afford another pandemic of this magnitude. With the emergence of artificial intelligence (AI), we are better equipped to handle such disasters or even prevent them from occurring in the first place. 

Whenever such an outbreak happens, the government or concerned authorities work to contain its spread and prevent the loss of lives. The management of an outbreak requires quick screening, forecasting the disease burden, and allocating healthcare personnel and infrastructure, including ICU beds, oxygen, drugs, and vaccines, effectively. 

A large-scale and quick screening is required to identify the infected individuals and isolate them, thereby containing the spread of the outbreak. AI tools can play a crucial role in this screening process. For instance, AI-based tools can analyse cough sounds to detect diseases, such as tuberculosis or COVID-19, and help with the treatment of infected individuals. 

AI can also provide diagnostic recommendations to expert doctors, reducing the turnaround time for making a diagnosis, for example, detecting different clinical features in a chest radiograph that could indicate certain diseases. 

The Central TB Division of India is leveraging AI across the tuberculosis cascade of care. One of these interventions is the automated interpretation of the Line Probe Assay test used for TB diagnosis. Similar tools can be used for epidemic-prone diseases to quicken the process of reaching a diagnosis and help contain an outbreak. 

Another key use of AI predictive models in pandemic management is to forecast the spread of disease—helping the public health system be better prepared in advance. AI models can be used to predict future infection cases, hospitalisations, requirements for ICU beds, oxygen, and deaths that are likely to occur in a region.

These models take into account a variety of factors, including population density, age distribution, and the effectiveness of social distancing measures. Using these forecasts, the authorities can allocate personnel and healthcare infrastructure such as hospital beds, ventilators, and personal protective equipment more efficiently to save lives. 

Since identifying and isolating infected cases is one of the most important parts of managing a pandemic, a large country such as India needs large-scale contact tracing. AI models can help automate this process by analysing data from mobile phones and other sources and identifying new cases quickly and efficiently. 

In India, where resources are limited, resource allocation and distribution are always a challenge. AI models can be used in supply chain optimisation, reduce lead times, significantly lower costs, and enable faster distribution of resources by predicting alternate routes.

During the COVID-19 pandemic, government agencies globally utilised these predictive models to predict the requirement of medical facilities.

AI predictive models can also be used to inform policy decisions, enforce social distancing measures, and optimise distribution. For example, models can be trained to evaluate the effectiveness of social distancing measures such as school and office closures and travel restrictions. By simulating various scenarios, policymakers can determine which interventions would be most effective in controlling the spread of disease. 

AI models can accelerate the treatment identification process and development of vaccines in response to emerging outbreaks by analysing data from a large number of clinical trials and medical records. 

The Indian government has been using an AI-powered media surveillance system since April 2022 under its Integrated Disease Surveillance Programme (IDSP). The programme has a Media Scanning and Verification Cell established in July 2008 at the National Centre for Disease Control, Delhi. It is used to manually scan English and Hindi newspapers to detect potential outbreaks and issue timely alerts to local authorities.

Using AI, the manual task of scanning articles is automated, enabling coverage of millions of internet articles. The AI tool scans nationwide newspapers in eleven Indian languages, including English, to detect adverse healthcare events and potential outbreak events. 

The Integrated Health Information Platform (IHIP), developed by the WHO to support the IDSP programme, records that the AI tool has scanned 80,000 health-related articles on average per day since April 2022. Of these, it shows 20-200 adverse healthcare events depending on seasonal variation, endemic situations, outbreaks, etc., to the Media Surveillance Assistant (MSA) who handles this tool.

The MSA then shortlists the events from this list by verifying the sources. After these shortlisted articles are approved by an epidemiologist, they are published, and an alert is generated for the concerned local authorities.

In conclusion, while there could be some challenges to the use of AI in managing pandemics, the potential benefits are substantial and could help save lives and mitigate the economic and social impact of the pandemic.

Mukul Kumar is an Associate ML Scientist at Wadhwani AI.