UKHSA is exploring the role that artificial intelligence
(AI) could play in
helping scientists to detect and investigate foodborne illness
outbreaks.
In a new study, UKHSA
experts have assessed different types of AI for their ability to detect
and classify text in online restaurant reviews, which could one
day be used to identify and potentially target investigations
into foodborne illness outbreaks.
Foodborne gastrointestinal (GI) illness – which usually presents as
vomiting and diarrhoea – is a major burden on society's health in
the UK, causing millions of people to become unwell every year.
However, it is estimated that most cases of GI illness are not formally diagnosed.
UKHSA tech experts and scientists looked at a range of large
language models and rated their ability to trawl thousands of
online reviews for information about symptoms which might relate
to GI illness – such as
diarrhoea, vomiting and abdominal pain – as well as different
food types people report eating.
UKHSA scientists believe that gathering information in this way
could one day become routine, providing more information on rates
of GI illness which are not
captured by current systems as well as vital clues around
possible sources and causes in outbreaks.
However, the study has highlighted key challenges around the
approach that would need to first be overcome, particularly
around access to real-time data.
While it is possible to use the approach to gather general
information on the type of food people have eaten and which may
be linked to illness, determining which specific ingredients or
other factors that may be linked is difficult. Variations in
spelling and the use of slang were also identified as potential
challenges, as well as people misattributing their illness to a
given meal.
Professor Steven Riley, Chief Data Officer at UKHSA said:
We are constantly looking for new and effective ways to enhance
our disease surveillance.
Using AI in this way
could soon help us identify the likely source of more foodborne
illness outbreaks, in combination with traditional
epidemiological methods, to prevent more people becoming sick.
Further work is needed before we adopt these methods into our
routine approach to tackling foodborne illness outbreaks.
While work has previously been carried out to consider how
AI could help in
reviewing restaurant reviews in this way, UKHSA's latest study
goes a step further by looking at a much more detailed list of
terms and language that could potentially help to identify
illness outbreaks.
This works forms part of UKHSA's evaluation
of AI to perform
different tasks within public health.
Over three thousand reviews were manually annotated by
epidemiologists after being collected and filtered.
Reviews were filtered for those containing a comprehensive list
of possible GIrelated
keywords, which were then further examined for relevant symptoms.
Symptoms such as headache, fever, and respiratory symptoms were
not annotated, because they are not sufficiently specific to
GI illness.
