From: Impact of large language model (ChatGPT) in healthcare: an umbrella review and evidence synthesis
Author, Year | Field of study | Objectives related to ChatGPT | Pros of using ChatGPT | Cons/ Challenges of using ChatGPT | Future directions suggested by author (evidence gap) | Findings of Meta—Analysis |
---|---|---|---|---|---|---|
[10] | Patient care and medical research | To explore the capacity of ChatGPT in enhancing patient care and its contribution to medical research and medical writing | ChatGPT can help with handling patient inquiries, documenting notes, making decisions, enrolling in trials, managing data, providing decision support, assisting with research, and educating patients | The provided solutions often fall short and present conflicting information, leading to concerns about their originality, privacy, accuracy, bias, and legality. Content generated by ChatGPT raises issues related to bias and potential plagiarism | Longitudinal studies on the role of ChatGPT in healthcare and medical education and research as well as comparison of ChatGPT with other AI tools | N/A |
[27] | Support of ChatGPT in conducting a medical systematic review | To evaluate human researchers' performance versus ChatGPT in systematic review tasks (inter- and intrarater reliability, sensitivity, specificity, accuracy, precision, chance hit rate) | A substantial level of agreement between ChatGPT and human researchers in extracting information for systematic review except for reporting study design, clinical task, and clinical implementation | The limitations of ChatGPT render augmented systematic reviews inefficient for experienced researchers. Ethical implications of using ChatGPT in medical scientific writing | Range of application of ChatGPT and other transformer based models in healthcare should be increased | N/A |
[25] | Healthcare education, practice and research | To explore the effectiveness of ChatGPT in healthcare Education, research, and practice, while also emphasizing its potential drawbacks | Research: Review and writing, data analysis, code generation, saving Time for tasks requiring human intelligence Practice: drug discovery and development, workflow optimization, cost reduction, documentation, personalized medicine, health literacy Education: personalized learning, critical thinking and problem-based learning | Ethical concerns, copyright issues, transparency challenges, and legal considerations, along with the potential for bias, Plagiarism, lack of originality, inaccurate content leading to hallucination, limited knowledge, incorrect citations, cybersecurity risks, and the threat of infodemics | Potential areas to be explored include the application of AI in reviewing and editing tasks (for journals) and incorporation of Emotional support in patient care involving the use of ChatGPT. Role of ChatGPT in refining communication skills is also an area of importance | N/A |
[18] | Treatment decisions (for professionals and public) | To explore the usage and pitfalls of ChatGPT in healthcare | The question-and-answer (QA) design of ChatGPT's interface facilitates integration into the current clinical workflow | Unsuitable for direct clinical deployment as it is not designed for clinical applications | Future development of ‘ChatGPT medical professional version’, specific according to medical specialities | N/A |
[36] | Healthcare practice | To investigate the various applications, potential benefits and risks, ethical issues of ChatGPT in healthcare and recommendations for its adoption in medicine and cellular imaging | 1. Disseminates critical information about pandemics and infectious diseases 2. virtual assistant for orthognathic surgery consultations 3. addresses inquiries related to dental practice 4. simplifies complex medical concepts 5. differential diagnoses based on patient history 6. analysis of medical images and cellular imaging 7. expedited research processes through quick literature review | 1.Inability to understand specialized medical terminology 2. Can augment but not replace human judgment in clinical settings 3. In case of inaccurate clinical advice, accountability (legal and ethical regulatory frameworks) have not yet been devised 4. Balancing automation with the human touch | Work should be done for dynamic and real time learning of ChatGPT for obtaining updated information in medicine and health care | N/A |
[31] | Radiology | Evaluating the current applications and future directions of ChatGPT in the field of radiology | Demonstrates substantial potential to augment decision-making and optimizing workflow | Concerns regarding radiologic image processing, ethical and legal implications, potential self-diagnosis and self-management by patients | The image processing potential of ChatGPT 4.0 needs to be validated further | N/A |
[2] | Medical and dental research | To evaluate the reliability and accuracy of ChatGPT in medical and dental research | Meta-analysis shows that the accuracy of ChatGPT in providing correct responses was significantly higher compared to the total responses to queries related to medical examination, systematic reviews, clinical reasoning, diagnostic imaging, liver diseases, and COVID-19 vaccination | Transparency, ethical concern, erroneous content, variation in population, data reproducibility | The studies on ChatGPT have small data points tested (small sample size) which poses threat to extrapolation of findings. Therefore, larger datasets (including concepts, queries, prompts etc.) should be tested using ChatGPT to improve the generalizability of the study results | The meta-analysis showed an odds ratio (OR) of 2.25 and a relative risk (RR) of 1.47 with a 95% confidence interval (CI), indicating that the accuracy of ChatGPT in providing correct responses was significantly higher compared to the total responses for queries |
[3] | Neurosurgery | To examine the potential benefits and limitations of ChatGPT in neurosurgical practice and education | Personalized treatment plans, supporting surgical planning and navigation, and enhancing large data processing efficiency and accuracy | Question format limitations, validation challenges, and algorithmic bias along with ethical issues related to its usage | Incorporating longitudinal patient data into predictive models can enhance outcome prediction | N/A |
[17] | Medical examinations | To assess the performance in medical examinations with multiple-choice questions | ChatGPT correctly answered the majority of multiple-choice questions in medical examinations with a passing grade | Preparations for medical examinations using ChatGPT should be done with caution | Development of training data set specific for medical education Exploring the role of Future AI chatbots in medical examination preparation | Overall performance of ChatGPT ranged from 40% in the biomedical admission test to 100% in a diabetes knowledge questionnaire. The mean performance of ChatGPT was 61.1% (95% CI 56.1%–66.0% |
[20] | Medicine and healthcare | To explore the potential of ChatGPT in medicine | Streamline and simplify complex tasks, improve patient care, enhance clinical decision making, and facilitate communication among healthcare professionals | Privacy, ethical considerations, tokenization, sensitivity of wording of prompt, lack of capacity to handle image based questions | Involvement of medical practitioners in training ChatGPT | N/A |
[15] | Gastroenterology | To assess the applications, benefits and limitations of ChatGPT in the field of gastroenterology | It can provide recommendations, enhance communication between patients and caregivers, and prompt valuable research inquiries | Obstacles in decoding intricate medical questions, yielded inconsistent responses at times, and exhibited limitations in generating novel content | The model of ChatGPT used in research influences its outcomes, so a comparable version should be used Research on prompt selection for ChatGPT is lacking which can drastically change the outcome/ response | N/A |
[32] | Public health dentistry | To find out applications and drawbacks of ChatGPT in public dental health schooling, writing for academic use, research and clinical practice in public dental health | Helps scholars with the authoring of scientific research and dental studies Scientists can focus and allocate more time on experimentation by delegating some tasks to ChatGPT | Prejudice in the training data, undervaluing human skills, possible fraud, legal and reproducibility concerns | Inclusion of more homogenous studies in terms of quality in order to improve the generalizability of systematic review findings | N/A |
[22] | Neurosurgery | To explore the potential benefits and limitations of ChatGPT in the field of neurosurgery | Accuracy and efficiency of neurosurgical procedures, as well as diagnosis, treatment, and patient outcomes | 1. Need for large datasets 2. Potential for errors in the output | To provide extensive database necessary to train ChatGPT without breaching patient confidentiality | N/A |
[4] | Radiology | To conduct a qualitative and quantitative analysis of ChatGPT literature in radiology, assessing its scope and impact | Enhances patient education, protocol selection, and differential diagnosis generation Improves radiology report structuring and examination preparation | 1. Inconsistency in performance and information accuracy 2. Challenges in fully integrating AI into clinical radiology workflows | Research on factually incorrect information/ hallucinations generated from ChatGPT is needed | N/A |
[5] | Disease risk detection, personalized healthcare procedures, and enhancing patient care safety and quality | To explore the efficacy of ChatGPT and generative AI tools in medical diagnostics, treatment recommendations, and improving healthcare practices | 1. Enhances diagnostic accuracy and surgical planning 2. Reduces administrative tasks, improving physician efficiency 3. Supports evidence-based decision-making and clinical education | 1. Potential inaccuracies in clinical letter generation 2. Requires regulation and careful integration into clinical workflows 3. Risk of misinterpreting treatment guidelines affecting patient care | Predictive analytics to observe how well can ChatGPT assess the real-time data streams | N/A |
[35] | Ophthalmology | To evaluate the effectiveness and potential of ChatGPT in improving ophthalmological care, specifically in diagnosis, patient interaction, and educational roles | 1. Offers rapid, accessible information and support for clinical decisions 2. Enhances patient education through simplified explanations of conditions and treatments | 1. Potential inaccuracies in medical advice or diagnostic information 2. Ethical concerns around patient data privacy and the reliability of AI-generated advice | Explore the effectiveness of ChatGPT in a diverse linguistic landscape | N/A |
[24] | Medical research | To evaluate ChatGPT's application and effectiveness in medical research, including treatment, diagnosis, medication provision, and more | Offers potential benefits in drug development, medical report improvement, providing treatment and medical information, literature review writing, research conduction, data analysis, and personalized medicine | Concerns about ChatGPT's accuracy, originality, academic integrity, and ethical issues like privacy and security in medical research | Overcome issues pertaining to academic integrity, privacy, and ethics | N/A |