Off-the-Shelf Large Language Models for Causality Assessment of Individual Case Safety Reports: A Proof-of-Concept with COVID-19 Vaccines
Drug Safety,
Год журнала:
2025,
Номер
unknown
Опубликована: Март 12, 2025
This
study
evaluated
the
feasibility
of
ChatGPT
and
Gemini,
two
off-the-shelf
large
language
models
(LLMs),
to
automate
causality
assessments,
focusing
on
Adverse
Events
Following
Immunizations
(AEFIs)
myocarditis
pericarditis
related
COVID-19
vaccines.
We
assessed
150
COVID-19-related
cases
reported
Vaccine
Event
Reporting
System
(VAERS)
in
United
States
America
(USA).
Both
LLMs
human
experts
conducted
World
Health
Organization
(WHO)
algorithm
for
vaccine
inter-rater
agreement
was
measured
using
percentage
agreement.
Adherence
WHO
by
comparing
LLM
responses
expected
sequence
algorithm.
Statistical
analyses,
including
descriptive
statistics
Random
Forest
modeling,
explored
case
complexity
(e.g.,
string
length
measurements)
factors
affecting
performance
adherence.
showed
higher
adherence
(34%)
compared
Gemini
(7%)
had
moderate
(71%)
with
experts,
whereas
fair
(53%).
often
failed
recognize
listed
AEFIs,
incorrectly
identifying
6.7%
13.3%
respectively.
inconsistencies
8.0%
46.7%.
For
ChatGPT,
associated
lower
prompt
sections.
The
random
forest
analysis
achieved
an
accuracy
55%
(95%
confidence
interval:
35.7–73.5)
predicting
ChatGPT.
Notable
limitations
have
been
identified
their
use
aiding
assessments
safety.
performed
better,
experts.
In
investigated
scenario,
both
are
better
suited
as
complementary
tools
expertise.
Язык: Английский
Mental Health in the Time of the COVID-19 Pandemic: A Scoping Review of Collateral Effects on Common Mental Disorders (CMDs)
International Journal of Environmental Research and Public Health,
Год журнала:
2025,
Номер
22(4), С. 478 - 478
Опубликована: Март 23, 2025
It
is
unclear
whether
the
COVID-19
pandemic
has
had
consequences
for
common
mental
disorders
(CMDs).
This
scoping
review
aims
to
examine
direct
infection-related
(e.g.,
severe
illness),
psychosocial
social
isolation),
and
indirect
outcomes
changes
in
incidence)
that
have
been
particularly
discussed
so
far.
A
literature
search
clinically
diagnosed
adult
CMDs
was
conducted
using
Pubmed,
Web
of
Science,
PsycInfo
(n
=
5325).
After
completion
screening
process,
26
included
studies
remained
extraction.
None
reported
post-pandemic
data.
The
effects
appeared
be
pronounced
anxiety
obsessive-compulsive
first
year
pandemic.
followed
by
a
period
adjustment,
during
which
rates
disease
its
symptoms
largely
returned
pre-pandemic
levels.
Fluctuating
may
COVID-related
causes.
Preventive
temporary
inpatient
care
could
protective
approach
those
at
risk
or
vulnerable,
as
well
establishing
consultation
building
resilience.
gap
research
lack
comparisons
CMD
data
before,
during,
after
distinguish
transient
from
chronic
requiring
treatment.
Язык: Английский