New Delhi: An artificial intelligence tool deployed by the National Centre for Disease Control (NCDC) in 2022 may have helped issue more than 5,000 alerts of infectious outbreaks to health authorities in real-time since installation, according to a study.
Developed by WadhwaniAI, a New Delhi-based healthcare AI solutions provider, the ‘Health Sentinel’ tool could have helped slash 98 per cent of manual workload, enabling a quicker detection of an outbreak and proactive public health response, findings published as a pre-print paper and yet to be peer-reviewed suggest.
Nearly 200 countries are legally bound by the International Health Regulations (IHR) to operate a national disease surveillance system. The IHR and World Health Organization work together in protecting global health security.
News reports in print, electronic and online media are scanned by media scanning and verification tool under India’s ‘Integrated Disease Surveillance Programme’ (IDSP) for unusual health events, which are then shared with authorities for further action, if deemed necessary.
‘Health Sentinel’ scanned media reports and news articles everyday in 13 languages.
The authors wrote, “From April 2022 till date, Health Sentinel has processed over 300 million news articles and identified over 95,000 unique health events across India of which over 3,500 events (four per cent) were shortlisted by the public health experts at NCDC as potential outbreaks.”
Authors from Wadhwani AI told PTI that between April 2022 and April 2025, more than 5,000 real-time alerts have been sent to health authorities across India.
“Traditionally, the process of identifying potential disease events reported in the media involved manual scanning of newspapers, journals, and reports to identify relevant articles,” Parag Govil, national program lead for global health security at Wadhwani AI, told PTI.
“Introducing the ‘Health Sentinel’ solution has replaced this manual process while retaining a human-in-the-loop approach where epidemiologists perform essential verification before the information is disseminated to state and district officials,” he added.
Traditional approaches in disease surveillance rely on ‘passive reporting’ in which reports of infections from physicians and healthcare providers are looked at.
Monitoring informal sources such as online media has also become increasingly popular for disease surveillance, but as volume of articles published everyday has increased, manual workload of screening media has ballooned up and is impractical, the study’s authors said, including those from the National Centre for Disease Control (NCDC).
They proposed the ‘Health Sentinel’ tool that makes use of AI to extract information on unusual health events or outbreaks from news articles.
Health officials and epidemiologists in the government identified that automation of manual screening was among the requirements to strengthen the country’s media-based surveillance, along with faster outbreak detection and multilingual abilities, Govil said.
The research team observed a 150 per cent increase in published events since 2022, compared to previous years of human-based disease surveillance.
Further, 96 per cent of the health events published by the national surveillance system in 2024 were extracted by the AI tool — only four per cent through a manual scanning of media, they said.
Studies have suggested supporting traditional surveillance with looking through online content — news reports or social media posts — for an improved detection of infectious disease outbreaks.
A study, published in February in the ‘Indian Journal of Medical Research: Official Journal of the Indian Council of Medical Research’, piloted an event-based surveillance system that looked at media and rumour registers in six private hospitals of Kerala’s Kasaragod district.
Researchers from ICMR-National Institute of Epidemiology in Chennai and state and district health officials of Kerala developed an algorithm that analysed case records of patients admitted with acute febrile illness (AFI) — a fever that can last up to two weeks.
Symptoms of interest such as AFI with rash or haemorrhage were identified using spatiotemporal clustering of patients or deaths.
During May to December 2023, almost three-fourths of over 4,500 patients with AFI were analysed using the algorithm. Of the 88 clusters identified, 76 per cent were due to severe acute respiratory illness, 10 per cent due to acute encephalitis syndrome and nine per cent due to AFI with rash.
Further, 10 clusters were verified as events, with nine classified as outbreaks, including dengue and COVID-19, the study said.
Authors said, “EBS (Event-based surveillance) pilot in private health facilities complemented the (traditional) system by early detecting outbreaks. This EBS model has the potential for implementation in other districts, especially in districts at higher risk of zoonotic spillover.”
A 2020 review in the Journal of Biomedical Informatics looked at 148 research articles published during 2010-2019 on healthcare surveillance that looked at social media, including posts made on X (then Twitter).
Twenty six of the articles, including those sourced from ‘ScienceDirect’ and ‘PubMed’ databases, were seen to employ machine learning — a form of AI algorithm — in surveillance.
About a fourth of the articles focussed on surveillance of flu or influenza-like illnesses, with Twitter being the most popular source of social media data for performing surveillance research, the study noted.
“The inclusion of online data in surveillance systems has improved the disease prediction ability over traditional syndromic surveillance systems,” authors from the Delhi Technological University wrote.
Another study, published in the American Journal of Tropical Medicine and Hygiene in 2017, pointed to how data from news articles can help with delays in obtaining country-level data on confirmed cases of infections such as dengue fever.
