ChatGPT and Clinical Decision Support: Scope, Application, and Limitations

Abstract

This study examines ChatGPT's role in clinical decision support, by analyzing its scope, application, and limitations. By analyzing patient data and providing evidence-based recommendations, ChatGPT, an AI language model, can help healthcare professionals make well-informed decisions. This study examines ChatGPT's use in clinical decision support, including diagnosis and treatment planning. However, it acknowledges limitations like biases, lack of contextual understanding, and human oversight and also proposes a framework for the future clinical decision support system. Understanding these factors will allow healthcare professionals to utilize ChatGPT effectively and make accurate clinical decisions. Further research is needed to understand the implications of using ChatGPT in healthcare settings and to develop safeguards for responsible use.

Full text

Vol.:(0123456789)
1 3
Annals of Biomedical Engineering
https://doi.org/10.1007/s10439-023-03329-4
LETTER TOTHEEDITOR
ChatGPT andClinical Decision Support: Scope, Application,
andLimitations
JannatulFerdush1· MahbubaBegum2· SakibTanvirHossain3
Received: 15 July 2023 / Accepted: 18 July 2023
© The Author(s) under exclusive licence to Biomedical Engineering Society 2023
Abstract
This study examines ChatGPT’s role in clinical decision support, by analyzing its scope, application, and limitations. By
analyzing patient data and providing evidence-based recommendations, ChatGPT, an AI language model, can help healthcare
professionals make well-informed decisions. This study examines ChatGPT’s use in clinical decision support, including
diagnosis and treatment planning. However, it acknowledges limitations like biases, lack of contextual understanding, and
human oversight and also proposes a framework for the future clinical decision support system. Understanding these factors
will allow healthcare professionals to utilize ChatGPT effectively and make accurate clinical decisions. Further research is
needed to understand the implications of using ChatGPT in healthcare settings and to develop safeguards for responsible use.
Keywords ChatGPT· CDS· Biasness· Ethics
Introduction
In modern healthcare, clinical decision support (CDS) sys-
tems provide valuable information and recommendations to
enhance the clinical decision-making. These systems utilize
various technologies and databases to analyze the patient
data, medical literature, and clinical guidelines. As a result,
healthcare professionals can diagnose diseases, determine
treatment options, and improve patient outcomes [1–3]. In
addition to promoting evidence-based practice, CDS sys-
tems aim to reduce errors. As healthcare evolves, the use of
artificial intelligence (AI) and natural language processing
(NLP) will become increasingly important for clinical deci-
sion support.
The ChatGPT [3] AI language model developed on the
basis of deep learning has gained considerable attention as
a powerful tool for natural language processing. Compared
to traditional AI, ChatGPT has few differences:
AI (Artificial Intelligence)
• AI covers a wide range of technologies and applications,
such as machine learning, natural language processing,
computer vision, and robotics.
• It aims at reproducing human-like intelligence and deci-
sion-making processes.
ChatGPT
• OpenAI’s ChatGPT is an example of a conversational AI
model based on the GPT architecture.
• Its primary focus is generating human-like text responses
in conversational settings.
• The ChatGPT can generate coherent and contextually
relevant responses based on extensive training data.
BIOMEDICAL
ENGINEERING
SOCIETY
Associate Editor Stefan M. Duma oversaw the review of this
article.
Jannatul Ferdush, Mahbuba Begum and Sakib Tanvir Hossain have
contributed equally to this work.
* Jannatul Ferdush
jannatulferdush@just.edu.bd
Mahbuba Begum
mahbuba327@yahoo.com
Sakib Tanvir Hossain
sakib.tanvir.0905059@gmail.com
1 Department ofComputer Science andEngineering, Jashore
University ofScience andTechnology, Jashore7408,
Bangladesh
2 Department ofComputer Science andEngineering, Mawlana
Bhasani Science andTechnology, Tangail1902, Bangladesh
3 Department ofMechanical Engineering, Khulna University
ofEngineering andTechnology, Khulna9203, Bangladesh

J.Ferdush et al.
1 3
• It is commonly used in chatbots, virtual assistants, as
well as other conversational applications, supporting cus-
tomers, answering queries, and engaging in dialogue.
Essentially, AI is a field that involves technologies and
applications that aim to simulate human intelligence. On
the other hand, ChatGPT, as a specific implementation of
AI, specializes in generating human-like text responses in
conversational settings, making it valuable for chatbots and
virtual assistants.
By using a transformer-based architecture, ChatGPT is
able to comprehend and generate texts that are human-like.
As a result of its ability to interact with users and respond
to their prompts, it is an ideal candidate for clinical decision
support. Using ChatGPT’s language generation capabili-
ties, it may be able to analyze the patient data, recommend
treatments, and offer insights based on its extensive knowl-
edge. By leveraging its vast knowledge base and language
understanding capabilities, ChatGPT can augment clinical
decision-making processes, potentially improving accuracy,
efficiency, and patient outcomes. However, it is crucial to
thoroughly investigate the scope, application, and limita-
tions of ChatGPT in clinical decision support to ensure its
responsible and effective integration into healthcare work-
flows. This study provides the following contributions to the
field of clinical decision support using ChatGPT:
1. Exploration of ChatGPT’s potential in clinical decision
support, highlighting its ability to analyze the patient
data and generate evidence-based recommendations.
2. Identification of practical applications and use cases for
ChatGPT in diagnosis, treatment planning, and patient
management. Discussion of the limitations and chal-
lenges of ChatGPT, including biases in training data
and the need for human oversight.
3. Implications for healthcare professionals, emphasizing
the importance of critical evaluation and future research
opportunities to optimize ChatGPT’s performance.
Overall, this research enhances our understanding of Chat-
GPT’s role in clinical decision support, offering insights for
improving patient care and outcomes.
Literature Review
We present the following literature review to provide an
overview of existing research and studies relating to Chat-
GPT in clinical decision support. It summarizes the current
knowledge and highlights the key findings, methodologies,
and gaps in the literature.
Artificial intelligence technologies are increasingly being
developed for a range of clinical circumstances [4–7]. In a
recent observational study conducted by Dakuo Wang and
their research group at Pace University, they interviewed 22
clinicians from six clinics in a rural area in China [8]. The
study revealed several challenges in the adoption of an AI-
based Clinical Decision Support System (AI-CDSS) in this
context. One of the major obstacles identified was the mis-
alignment between the design of the AI-CDSS and the local
workflow and context. The authors also found that clinicians
may resist accepting AI-CDSS due to concerns that it would
replace them in their jobs. Another significant barrier high-
lighted was the lack of interpretability and transparency of AI,
as clinicians found it difficult to understand how the algorithm
generated recommendations within the “black box” of the AI
system.
Interestingly, since the invention of ChatGPT in 2022 [9],
researchers have started exploring its potential in medical sci-
ence [10–14], as well as clinical decision support system. In
2023 [15], the authors sought to investigate whether ChatGPT
could provide valuable suggestions to improve the logic of
clinical decision support systems (CDS) and compare its per-
formance to suggestions generated by humans. They discov-
ered that ChatGPT holds promise for leveraging large language
models and reinforcement learning from human feedback to
enhance CDS alert logic and potentially other medical areas
involving complex clinical reasoning. However, it is important
to note that the article did not discuss one of the key features
of ChatGPT–its ability to remember previous conversations,
which could be utilized in this research.
In contrast to traditional AI systems, ChatGPT offers unique
advantages that should be highlighted for future research. Its
automated chat interface, combined with the ability to retain
contextual information from prior interactions, presents excit-
ing opportunities for various applications. By addressing this
challenge, researchers can unlock the full potential of Chat-
GPT and explore its broader scope of applications in the field
of healthcare. Ultimately, this could contribute to the devel-
opment of advanced learning health systems, revolutionizing
clinical decision-making processes.
In conclusion, the literature indicates that ChatGPT holds
significant potential in clinical decision support, with promis-
ing applications in diagnosis, treatment planning, and deci-
sion-making. While acknowledging its limitations, researchers
and healthcare professionals are actively working on address-
ing biases, improving transparency, and integrating human
expertise to ensure responsible and effective use of ChatGPT
in healthcare decision-making processes.

ChatGPT andClinical Decision Support: Scope, Application, andLimitations
1 3
Scope ofChatGPT inClinical Decision
Support
ChatGPT holds significant potential in the realm of clini-
cal decision support. This section examines the scope
of ChatGPT’s application in this domain, highlighting
its ability to assist healthcare professionals in making
informed decisions.
Clinical decision support involves utilizing technology
and data analysis to provide healthcare professionals with
relevant information, recommendations, and alerts to aid
in clinical decision-making. It encompasses various tasks
such as diagnosis, treatment planning, and patient man-
agement. The scope of clinical decision support is broad,
spanning multiple medical specialties and healthcare
settings.
• Exploration of How ChatGPT Can Contribute to Clini-
cal Decision-Making ChatGPT’s language generation
capabilities make it a valuable tool for clinical deci-
sion-making. It can analyze the patient data, medical
literature, and clinical guidelines to generate evidence-
based recommendations and insights. By engaging in
interactive conversations, ChatGPT can assist health-
care professionals in exploring different treatment
options, understanding complex medical information,
and obtaining relevant insights to support decision-
making.
• Discussion of the Potential Benefits of Using ChatGPT
in Healthcare Settings with Examples The utilization of
ChatGPT in healthcare settings offers several benefits.
Firstly, it can provide rapid access to up-to-date medi-
cal literature and clinical guidelines, allowing healthcare
professionals to stay abreast of the latest research and
best practices. Secondly, ChatGPT’s ability to analyze
and interpret patient data can aid in accurate diagnosis
and personalized treatment planning. It can identify pat-
terns, predict outcomes, and provide treatment recom-
mendations based on existing medical knowledge.
ChatGPT, for instance, can provide potential diag-
noses and treatment options that may have been over-
looked in rare or complex diseases by leveraging its
vast knowledge base. Additionally, ChatGPT can help
healthcare professionals determine the likelihood of
certain outcomes or complications based on patient
characteristics and medical history.
By employing ChatGPT in clinical decision support,
healthcare professionals can benefit from its ability to
provide timely and contextually relevant information,
augmenting their decision-making process and potentially
improving patient outcomes.
Application ofChatGPT inClinical Decision
Support
ChatGPT provides healthcare professionals with advanced
tools for improving clinical decision-making based on evi-
dence and patient data, enabling them to make evidence-
based decisions. ChatGPT can be effectively employed in
several areas of clinical decision support, including:
• Diagnosis Assistance: By analyzing patient symptoms,
medical history, and diagnostic test results, ChatGPT
can help diagnose. It can generate potential diagnoses
based on patterns and correlations found in medical
literature and databases. This aids healthcare profes-
sionals in considering a wide range of possibilities and
refining their diagnostic assessments.
For example, in a case where a patient presents with
ambiguous symptoms, ChatGPT can provide insights
and propose differential diagnoses, facilitating a more
comprehensive diagnostic evaluation.
• Treatment Planning and Recommendations: ChatGPT
can offer valuable support in developing personalized
treatment plans by leveraging evidence-based guide-
lines, clinical expertise, and medical literature. It can
provide healthcare professionals with treatment recom-
mendations, considering factors such as patient demo-
graphics, medical history, and existing comorbidities.
For instance, ChatGPT can help explore various ther-
apeutic options for a specific condition. This includes
medications, dosage regimens, and potential side
effects
• Clinical Guidelines and Best Practices: Using Chat-
GPT, healthcare professionals can access clinical
guidelines and best practices in real time, which can be
a valuable resource. With the latest information avail-
able regarding diagnostic criteria, treatment protocols,
and follow-up strategies, healthcare professionals can
align their practices with current and evidence-based
practices.
• Rare or Complex Cases: It may be possible to get valu-
able insights and suggestions from ChatGPT in rare or
complex cases where expertise is limited. In order to
propose potential diagnostic or treatment approaches
that may have been overlooked, ChatGPT analyzes
similar cases from medical literature.
• Clinical Research and Data Analysis: Healthcare pro-
fessionals can use ChatGPT to analyze the clinical
research data, extract relevant information, and iden-
tify the key findings. It can aid in data interpretation,
providing insights into treatment outcomes, prognostic
factors, and potential areas for further investigation.
For example, ChatGPT can analyze clinical trial data

J.Ferdush et al.
1 3
and generate summaries of the findings. In this way,
healthcare professionals can make informed decisions
about treatment efficacy and safety.
• Patient Education and Communication; ChatGPT can
help patient education and communication. It serves as
a conversational agent to answer patients’ questions and
provide information about their condition, treatment
options, and potential side effects. It can improve patient
understanding, empower them to make informed deci-
sions, and enhance patient–provider communication.
ChatGPT, for instance, provides patients with custom-
ized educational materials based on their preferences and
needs.
• Remote Healthcare and Telemedicine; Healthcare pro-
fessionals can use ChatGPT as a virtual assistant during
remote consultations in the context of remote healthcare
and telemedicine. Patients’ symptoms can be triaged,
initial recommendations provided, and remote health-
care experiences can be facilitated more smoothly. As
an example, ChatGPT can help determine a patient’s
urgency and provide initial guidance on next steps before
a healthcare professional conducts a more thorough
examination.
• Healthcare Resource Allocation: Healthcare resource
allocation decisions can be improved through ChatGPT
during the periods of resource constraint, such as pan-
demics and natural disasters. As a result of analyzing
patient data and available resources, ChatGPT can pro-
vide insight into optimal resource utilization, care prior-
itization, and staffing allocation. The ChatGPT platform,
for instance, can be used to determine the best way to
distribute vaccine doses to maximize the population ben-
efits.
These applications illustrate the versatility of ChatGPT in
clinical decision support, extending its potential to vari-
ous aspects of healthcare delivery and decision-making
processes.
Limitations ofChatGPT inClinical Decision
Support
While ChatGPT presents significant potential, it is essential
to consider its limitations and challenges when used in clini-
cal decision support [16].
• Discussion of Potential Biases in Training Data and
Model Outputs: ChatGPT’s responses are influenced by
the data it was trained on, which can introduce biases.
Biases in training data, such as the under representation
of certain demographics or medical conditions, may
affect ChatGPT’s recommendations accuracy and reli-
ability. Careful attention should be given to addressing
these biases to ensure equitable and unbiased decision
support.
• Consideration of the Lack of Contextual Understanding
in ChatGPT’s Responses: Despite its ability to under-
stand medical situations’ context and nuances, ChatGPT
may generate responses that seem plausible, but lack the
depth or specificity necessary to make accurate decisions.
In order to ensure that ChatGPT’s responses are appro-
priate and relevant within the clinical context, human
oversight and critical evaluation are imperative.
• Importance of Human Oversight and Critical Evaluation
of ChatGPT’s Recommendations: Despite ChatGPT’s
ability to provide valuable insights into clinical decision
support, healthcare professionals should exercise caution
and judgment when interpreting and applying the recom-
mendations. Human oversight is essential to verify the
accuracy, relevance, and appropriateness of ChatGPT’s
suggestions prior to making clinical decisions.
Mitigating Limitations andResponsible Use
ofChatGPT
It is essential to address ChatGPT’s limitations and develop
strategies for its responsible deployment in order to ensure
its responsible and effective use.
• Strategies for Addressing Biases and Improving Trans-
parency in ChatGPT’s Training: Identifying and address-
ing biases in ChatGPT’s training data involve ensuring
diverse and representative datasets that cover a wide
range of demographics and medical conditions. Fur-
thermore, transparency in the training process, includ-
ing documentation of data sources and model training
methodologies, can improve ChatGPT’s understanding
and scrutiny.
• Importance of Integrating Human Expertise and Over-
sight in Decision-Making: In order to effectively use
ChatGPT, healthcare professionals should actively par-
ticipate in the decision-making process and critically
evaluate its recommendations. By combining the knowl-
edge and experience of healthcare professionals with the
insights provided by ChatGPT, healthcare professionals
can make more informed and accurate decisions.
• Ethical Considerations and Responsible Implementa-
tion of ChatGPT in Healthcare Settings: As part of the
implementation of ChatGPT in healthcare settings, ethi-
cal considerations play a significant role. When utilizing
ChatGPT for clinical decision support, it is essential to
maintain patient privacy, confidentiality, and informed
consent. Additionally, healthcare organizations should
formulate clear guidelines and protocols for the use of

ChatGPT andClinical Decision Support: Scope, Application, andLimitations
1 3
artificial intelligence technology, which will ensure that
ChatGPT is implemented responsibly and aligns with
ethical standards.
Proposed Framework
The provided figure (Fig.1) illustrates a general framework
for a Clinical Decision Support System (CDSS) based on
ChatGPT. The framework demonstrates various potential
applications of ChatGPT within a CDSS context.
In this framework, the CDSS utilizes a primary scru-
tinizing system to perform the initial assessment of tests
and provide recommendations from doctors. Additionally,
the CDSS includes a treatment and recommendation block,
which automatically designs treatment plans based on doc-
tors’ suggestions. To address privacy and ethical concerns
when accessing patient data, all data are encrypted before
being shared with external researchers. The CDSS-research
and data collection component handles this encryption pro-
cess. Patients and doctors can request access to their data
at any time, and an automatic report is generated for them.
One of the key features of this system is its explainabil-
ity. Patients can always inquire about their data, test results,
and the reasons behind certain outcomes. The CDSS, being
a chat-based agent powered by ChatGPT, aims to keep
interactions engaging and informative. Moreover, doctors
can use natural language processing capabilities of Chat-
GPT to communicate in their own language without needing
knowledge of computer programming languages. This aspect
makes it an excellent educational tool for both doctors and
patients, fostering learning and knowledge exchange.
Overall, the framework presents a comprehensive CDSS
powered by ChatGPT, combining automated decision-mak-
ing, personalized data access, and educational opportunities
for healthcare professionals and patients alike.
Conclusion
In conclusion, ChatGPT has the potential to significantly
impact clinical decision support in healthcare settings. This
article has explored the scope of ChatGPT’s application,
including its potential benefits in diagnosis and treatment
planning. There have also been a number of limitations iden-
tified, such as biases and a lack of contextual understand-
ing, that need to be addressed in order to ensure respon-
sible implementation. By mitigating these limitations and
integrating human expertise and oversight, ChatGPT can be
harnessed as a valuable tool in healthcare decision-making
processes. It is crucial to continue research, collaboration,
and the responsible adoption of ChatGPT in healthcare in
Fig. 1 An example of general
framework for ChatGPT-based
Clinical Decision Support Sys-
tem (CDSSS)

J.Ferdush et al.
1 3
the future in order to refine its capabilities, address limi-
tations, and identify the best practices for integration. To
maximize ChatGPT’s potential and ensure its responsible
and ethical use in healthcare, researchers, healthcare profes-
sionals, and AI developers need to collaborate.
Acknowledgements The authors acknowledge that this article was par-
tially generated by ChatGPT (powered by OpenAI’s language model,
GPT; http:// openai. com). The editing was performed by the authors.
Declarations
Conflict of interest The authors declare no conflict of interest
References
1. Ingraham, N. E., E. K. Jones, S. King, J. Dries, M. Phillips, T.
Loftus, H. L. Evans, G. B. Melton, and C. J. Tignanelli. Re-aiming
equity evaluation in clinical decision support: a scoping review
of equity assessments in surgical decision support systems. Ann.
Surg. 277(3):359–364, 2023.
2. Meunier, P.-Y., C. Raynaud, E. Guimaraes, F. Gueyffier, and L.
Letrilliart. Barriers and facilitators to the use of clinical decision
support systems in primary care: A mixed-methods systematic
review. Ann. Fam. Med. 21(1):57–69, 2023.
3. Xu, Q., W. Xie, B. Liao, C. Hu, L. Qin, Z. Yang, H. Xiong, Y.
Lyu, Y. Zhou, A. Luo, etal. Interpretability of clinical decision
support systems based on artificial intelligence from technological
and medical perspective: A systematic review. J. Healthcare Eng.
2023. https:// doi. org/ 10. 1155/ 2023/ 99192 69.
4. Pierce, R. L., W. Van Biesen, D. Van Cauwenberge, J. Decruyen-
aere, and S. Sterckx. Explainability in medicine in an era of ai-
based clinical decision support systems. Front. Genet. 13:903600,
2022.
5. Panigutti, C., Beretta, A., Giannotti, F., Pedreschi, D.: Understand-
ing the impact of explanations on advice-taking: a user study for
ai-based clinical decision support systems. In: Proceedings of the
2022 CHI Conference on Human Factors in Computing Systems,
pp. 1–9 (2022)
6. Sloane, E.B., Silva, R.J.: Artificial intelligence in medical devices
and clinical decision support systems. In: Clinical Engineering
Handbook, pp. 556–568 (2020)
7. Wang, L., X. Chen, L. Zhang, L. Li, Y. Huang, Y. Sun, and X.
Yuan. Artificial intelligence in clinical decision support systems
for oncology. Int. J. Med. Sci. 20(1):79, 2023.
8. Wang, D., Wang, L., Zhang, Z., Wang, D., Zhu, H., Gao, Y., Fan,
X., Tian, F.: “brilliant ai doctor” in rural clinics: Challenges in
ai-powered clinical decision support system deployment. In: Pro-
ceedings of the 2021 CHI Conference on Human Factors in Com-
puting Systems. CHI ’21. Association for Computing Machinery,
New York, NY, USA (2021). https:// doi. org/ 10. 1145/ 34117 64.
34454 32
9. Introducing ChatGPT — openai.com. https:// openai. com/ blog/
chatg pt. [Accessed 08-Jul-2023]
10. Biswas, S. S. Role of chat gpt in public health. Ann. Biomed. Eng.
51(5):868–869, 2023.
11. Sallam, M.: The utility of chatgpt as an example of large language
models in healthcare education, research and practice: System-
atic review on the future perspectives and potential limitations.
medRxiv, 2023–02 (2023)
12. Garg, R.K., Urs, V.L., Agrawal, A.A., Chaudhary, S.K., Paliwal,
V., Kar, S.K.: Exploring the role of chat gpt in patient care (diag-
nosis and treatment) and medical research: A systematic review.
medRxiv, 2023–06 (2023)
13. Liu, J., C. Wang, and S. Liu. Utility of chatgpt in clinical practice.
Journal of Medical Internet Research. 25:48568, 2023.
14. Temsah, M.-H., Aljamaan, F., Malki, K.H., Alhasan, K.,
Altamimi, I., Aljarbou, R., Bazuhair, F., Alsubaihin, A., Abdulma-
jeed, N., Alshahrani, F.S., etal.: Chatgpt and the future of digital
health: a study on healthcare workers’ perceptions and expecta-
tions. In: Healthcare, vol. 11, p. 1812 (2023). MDPI
15. Liu, S., Wright, A.P., Patterson, B.L., Wanderer, J.P., Turer, R.W.,
Nelson, S.D., McCoy, A.B., Sittig, D.F., Wright, A.: Using AI-
generated suggestions from ChatGPT to optimize clinical decision
support. Journal of the American Medical Informatics Association
30(7), 1237–1245 (2023) https:// doi. org/ 10. 1093/ jamia/ ocad0 72
16. Dave, T., S. A. Athaluri, and S. Singh. Chatgpt in medicine: an
overview of its applications, advantages, limitations, future pros-
pects, and ethical considerations. Front. Artif. Intell. 6:1169595,
2023.
Publisher's Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.