Enhancing AI Chatbot Responses in Health Care: The SMART Prompt Structure in Head and Neck Surgery
OTO Open,
Год журнала:
2025,
Номер
9(1)
Опубликована: Янв. 1, 2025
Abstract
Objective
This
study
aims
to
evaluate
the
impact
of
prompt
construction
on
quality
artificial
intelligence
(AI)
chatbot
responses
in
context
head
and
neck
surgery.
Study
Design
Observational
evaluative
study.
Setting
An
international
collaboration
involving
16
researchers
from
11
European
centers
specializing
Methods
A
total
24
questions,
divided
into
clinical
scenarios,
theoretical
patient
inquiries,
were
developed.
These
questions
entered
ChatGPT‐4o
both
with
without
use
a
structured
format,
known
as
SMART
(Seeker,
Mission,
AI
Role,
Register,
Targeted
Question).
The
AI‐generated
evaluated
by
experienced
surgeons
using
Quality
Analysis
Medical
Artificial
Intelligence
instrument
(QAMAI),
which
assesses
accuracy,
clarity,
relevance,
completeness,
source
quality,
usefulness.
Results
generated
scored
significantly
higher
across
all
QAMAI
dimensions
compared
those
contextualized
prompts.
Median
scores
for
prompts
27.5
(interquartile
range
[IQR]
25‐29)
versus
(IQR
21.8‐25)
unstructured
(
P
<
.001).
Clinical
scenarios
inquiries
showed
most
significant
improvements,
while
also
benefited,
but
lesser
extent.
AI's
improved
notably
prompt,
particularly
questions.
Conclusion
suggests
that
format
enhances
approach
improves
completeness
information,
underscoring
importance
well‐constructed
applications.
Further
research
is
warranted
explore
applicability
different
medical
specialties
platforms.
Язык: Английский
Enhancing AI Chatbot Responses in Healthcare: The SMART Prompt Structure in Head and Neck Surgery
Research Square (Research Square),
Год журнала:
2024,
Номер
unknown
Опубликована: Авг. 23, 2024
Abstract
Objective.
To
evaluate
the
impact
of
prompt
construction
on
quality
AI
chatbot
responses
in
context
head
and
neck
surgery.
Study
design.
Observational
evaluative
study.
Setting.
International
collaboration
involving
16
researchers
from
11
European
centers
specializing
Methods.
A
total
24
questions,
divided
into
clinical
scenarios,
theoretical
patient
inquiries,
were
developed.
These
questions
inputted
ChatGPT-4o
both
with
without
use
a
structured
format,
known
as
SMART
(Seeker,
Mission,
Role,
Register,
Targeted
Question).
The
AI-generated
evaluated
by
experienced
surgeons
using
QAMAI
instrument,
which
assesses
accuracy,
clarity,
relevance,
completeness,
source
quality,
usefulness.
Results.
generated
scored
significantly
higher
across
all
dimensions
compared
to
those
contextualized
prompts.
Median
scores
for
prompts
27.5
(IQR
25–29)
versus
21.8–25)
unstructured
(p
<
0.001).
Clinical
scenarios
inquiries
showed
most
significant
improvements,
while
also
benefited
but
lesser
extent.
AI's
improved
notably
prompt,
particularly
questions.
Conclusions.
study
suggests
that
format
enhances
This
approach
improves
completeness
information,
underscoring
importance
well-constructed
applications.
Further
research
is
warranted
explore
applicability
different
medical
specialties
platforms.
Язык: Английский
Generative artificial intelligence (GenAI) and decision-making: Legal & ethical hurdles for implementation in mental health
International Journal of Law and Psychiatry,
Год журнала:
2024,
Номер
97, С. 102028 - 102028
Опубликована: Окт. 19, 2024
Язык: Английский
Expanding the Reach of Mindfulness: A Mechanistic Approach and AI Applications
Mindfulness,
Год журнала:
2024,
Номер
unknown
Опубликована: Дек. 13, 2024
Язык: Английский
Efficacy of Different Beta Blockers in Reducing Mortality in Heart-Failure Patients
Cureus,
Год журнала:
2024,
Номер
unknown
Опубликована: Ноя. 21, 2024
This
study
evaluated
the
comparative
efficacy
of
different
beta
blockers
bisoprolol,
carvedilol,
and
metoprolol
in
reducing
mortality
hospitalizations
among
120
heart-failure
(HF)
patients.
The
sample
had
an
equal
gender
distribution
(50%
male,
50%
female)
with
a
mean
age
69.28
years.
Baseline
characteristics,
such
as
systolic
blood
pressure
(mean:
134.36
mmHg)
left
ventricular
ejection
fraction
(LVEF)
40.24%),
were
comparable
across
treatment
groups.
Patients
treated
either
bisoprolol
(30%),
carvedilol
or
(40%)
for
average
27.54
weeks.
utilized
Poisson
negative
binomial
regression
models
to
assess
hospitalization
rates,
chi-square
tests
compare
outcomes.
Results
revealed
that
was
44.2%
entire
cohort,
no
significant
differences
between
three
beta-blocker
groups
(p
=
0.301).
Similarly,
observed
0.276)
ICU
admissions
0.797).
However,
patients
on
exhibited
slight
improvement
New
York
Heart
Association
(NYHA)
class
LVEF,
though
this
not
statistically
0.145
p
0.477,
respectively).
Side
effects,
including
bradycardia,
fatigue,
hypotension,
noted
32.5%,
21.7%,
23.3%
patients,
respectively.
These
findings
suggest
all
are
similarly
effective
mortality,
may
offer
better
control
HF
symptoms.
Язык: Английский