Dataset Information

Benchmarking large language models' performances for myopia care: a comparative analysis of ChatGPT-3.5, ChatGPT-4.0, and Google Bard.

ABSTRACT:

Background

Large language models (LLMs) are garnering wide interest due to their human-like and contextually relevant responses. However, LLMs' accuracy across specific medical domains has yet been thoroughly evaluated. Myopia is a frequent topic which patients and parents commonly seek information online. Our study evaluated the performance of three LLMs namely ChatGPT-3.5, ChatGPT-4.0, and Google Bard, in delivering accurate responses to common myopia-related queries.

Methods

We curated thirty-one commonly asked myopia care-related questions, which were categorised into six domains-pathogenesis, risk factors, clinical presentation, diagnosis, treatment and prevention, and prognosis. Each question was posed to the LLMs, and their responses were independently graded by three consultant-level paediatric ophthalmologists on a three-point accuracy scale (poor, borderline, good). A majority consensus approach was used to determine the final rating for each response. 'Good' rated responses were further evaluated for comprehensiveness on a five-point scale. Conversely, 'poor' rated responses were further prompted for self-correction and then re-evaluated for accuracy.

Findings

ChatGPT-4.0 demonstrated superior accuracy, with 80.6% of responses rated as 'good', compared to 61.3% in ChatGPT-3.5 and 54.8% in Google Bard (Pearson's chi-squared test, all p ≤ 0.009). All three LLM-Chatbots showed high mean comprehensiveness scores (Google Bard: 4.35; ChatGPT-4.0: 4.23; ChatGPT-3.5: 4.11, out of a maximum score of 5). All LLM-Chatbots also demonstrated substantial self-correction capabilities: 66.7% (2 in 3) of ChatGPT-4.0's, 40% (2 in 5) of ChatGPT-3.5's, and 60% (3 in 5) of Google Bard's responses improved after self-correction. The LLM-Chatbots performed consistently across domains, except for 'treatment and prevention'. However, ChatGPT-4.0 still performed superiorly in this domain, receiving 70% 'good' ratings, compared to 40% in ChatGPT-3.5 and 45% in Google Bard (Pearson's chi-squared test, all p ≤ 0.001).

Interpretation

Our findings underscore the potential of LLMs, particularly ChatGPT-4.0, for delivering accurate and comprehensive responses to myopia-related queries. Continuous strategies and evaluations to improve LLMs' accuracy remain crucial.

Funding

Dr Yih-Chung Tham was supported by the National Medical Research Council of Singapore (NMRC/MOH/HCSAINV21nov-0001).

SUBMITTER: Lim ZW

PROVIDER: S-EPMC10470220 | biostudies-literature | 2023 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Benchmarking large language models' performances for myopia care: a comparative analysis of ChatGPT-3.5, ChatGPT-4.0, and Google Bard.

Lim Zhi Wei ZW Pushpanathan Krithi K Yew Samantha Min Er SME Lai Yien Y Sun Chen-Hsin CH Lam Janice Sing Harn JSH Chen David Ziyou DZ Goh Jocelyn Hui Lin JHL Tan Marcus Chun Jin MCJ Sheng Bin B Cheng Ching-Yu CY Koh Victor Teck Chang VTC Tham Yih-Chung YC

EBioMedicine 20230823

<h4>Background</h4>Large language models (LLMs) are garnering wide interest due to their human-like and contextually relevant responses. However, LLMs' accuracy across specific medical domains has yet been thoroughly evaluated. Myopia is a frequent topic which patients and parents commonly seek information online. Our study evaluated the performance of three LLMs namely ChatGPT-3.5, ChatGPT-4.0, and Google Bard, in delivering accurate responses to common myopia-related queries.<h4>Methods</h4>We ...[more]

PMID: 37625267

Similar Datasets

Project description:BackgroundsThe use of large language models (LLMs) in medicine can help physicians improve the quality and effectiveness of health care by increasing the efficiency of medical information management, patient care, medical research, and clinical decision-making.MethodsWe collected 34 frequently asked questions about glucocorticoid-induced osteoporosis (GIOP), covering topics related to the disease's clinical manifestations, pathogenesis, diagnosis, treatment, prevention, and risk factors. We also generated 25 questions based on the 2022 American College of Rheumatology Guideline for the Prevention and Treatment of Glucocorticoid-Induced Osteoporosis (2022 ACR-GIOP Guideline). Each question was posed to the LLM (ChatGPT-3.5, ChatGPT-4, and Google Gemini), and three senior orthopedic surgeons independently rated the responses generated by the LLMs. Three senior orthopedic surgeons independently rated the answers based on responses ranging between 1 and 4 points. A total score (TS) > 9 indicated 'good' responses, 6 ≤ TS ≤ 9 indicated 'moderate' responses, and TS < 6 indicated 'poor' responses.ResultsIn response to the general questions related to GIOP and the 2022 ACR-GIOP Guidelines, Google Gemini provided more concise answers than the other LLMs. In terms of pathogenesis, ChatGPT-4 had significantly higher total scores (TSs) than ChatGPT-3.5. The TSs for answering questions related to the 2022 ACR-GIOP Guideline by ChatGPT-4 were significantly higher than those for Google Gemini. ChatGPT-3.5 and ChatGPT-4 had significantly higher self-corrected TSs than pre-corrected TSs, while Google Gemini self-corrected for responses that were not significantly different than before.ConclusionsOur study showed that Google Gemini provides more concise and intuitive responses than ChatGPT-3.5 and ChatGPT-4. ChatGPT-4 performed significantly better than ChatGPT3.5 and Google Gemini in terms of answering general questions about GIOP and the 2022 ACR-GIOP Guidelines. ChatGPT3.5 and ChatGPT-4 self-corrected better than Google Gemini.

Project description:Background: This study evaluates the ability of six popular chatbots; ChatGPT-3.5, ChatGPT-4.0, Gemini, Copilot, Chatsonic, and Perplexity, to provide reliable answers to questions concerning keratoconus. Methods: Chatbots responses were assessed using mDISCERN (range: 15-75) and Global Quality Score (GQS) (range: 1-5) metrics. Readability was evaluated using nine validated readability assessments. We also addressed the quality and accountability of websites from which the questions originated. Results: We analyzed 20 websites, 65% "Private practice or independent user" and 35% "Official patient education materials". The mean JAMA benchmark score was 1.40 ± 0.91 (0-4 points), indicating low accountability. Reliability, measured using mDISCERN, ranged from 42.91 ± 3.15 (ChatGPT-3.5) to 46.95 ± 3.53 (Copilot). The most frequent question was "What is keratoconus?" with 70% of websites providing relevant information. This received the highest mDISCERN score (49.30 ± 4.91) and a relatively high GQS score (3.40 ± 0.56) with an Automated Readability Level Calculator score of 13.17 ± 2.13. Moderate positive correlations were determined between the website numbers and both mDISCERN (r = 0.265, p = 0.25) and GQS (r = 0.453, p = 0.05) scores. The quality of information, assessed using the GQS, ranged from 3.02 ± 0.55 (ChatGPT-3.5) to 3.31 ± 0.64 (Gemini) (p = 0.34). The differences between the texts were statistically significant. Gemini emerged as the easiest to read, while ChatGPT-3.5 and Perplexity were the most difficult. Based on mDISCERN scores, Gemini and Copilot exhibited the highest percentage of responses in the "good" range (51-62 points). For the GQS, the Gemini model exhibited the highest percentage of responses in the "good" quality range with 40% of its responses scoring 4-5. Conclusions: While all chatbots performed well, Gemini and Copilot showed better reliability and quality. However, their readability often exceeded recommended levels. Continuous improvements are essential to match information with patients' health literacy for effective use in ophthalmology.

Project description:Digital teaching diversifies the ways of knowledge assessment, as natural language processing offers the possibility of answering questions posed by students and teachers.ObjectiveThis study evaluated ChatGPT's, Bard's and Gemini's performances on second year of medical studies' (DFGSM2) Pathology exams from the Health Sciences Center of Dijon (France) in 2018-2022.MethodsFrom 2018 to 2022, exam scores, discriminating powers and discordance rates were retrieved. Seventy questions (25 first-order single response questions and 45 second-order multiple response questions) were submitted on May 2023 to ChatGPT 3.5 and Bard 2.0, and on September 2024 to Gemini 1.5 and ChatGPT-4. Chatbot's and student's average scores were compared, as well as discriminating powers of questions answered by chatbots. The percentage of student-chatbot identical answers was retrieved, and linear regression analysis correlated the scores of chatbots with student's discordance rates. Chatbot's reliability was assessed by submitting the questions in four successive rounds and comparing score variability using a Fleiss' Kappa and a Cohen's Kappa.ResultsNewer chatbots outperformed both students and older chatbots as for the overall scores and multiple-response questions. All chatbots outperformed students on less discriminating questions. Oppositely, all chatbots were outperformed by students to questions with a high discriminating power. Chatbot's scores were correlated to student discordance rates. ChatGPT 4 and Gemini 1.5 provided variable answers, due to effects linked to prompt engineering.ConclusionOur study in line with the literature confirms chatbot's moderate performance for questions requiring complex reasoning, with ChatGPT outperforming Google chatbots. The use of NLP software based on distributional semantics remains a challenge for the generation of questions in French. Drawbacks to the use of NLP software in generating questions include the generation of hallucinations and erroneous medical knowledge which have to be taken into count when using NLP software in medical education.

Project description:BackgroundThe increasing application of generative artificial intelligence large language models (LLMs) in various fields, including dentistry, raises questions about their accuracy.ObjectiveThis study aims to comparatively evaluate the answers provided by 4 LLMs, namely Bard (Google LLC), ChatGPT-3.5 and ChatGPT-4 (OpenAI), and Bing Chat (Microsoft Corp), to clinically relevant questions from the field of dentistry.MethodsThe LLMs were queried with 20 open-type, clinical dentistry-related questions from different disciplines, developed by the respective faculty of the School of Dentistry, European University Cyprus. The LLMs' answers were graded 0 (minimum) to 10 (maximum) points against strong, traditionally collected scientific evidence, such as guidelines and consensus statements, using a rubric, as if they were examination questions posed to students, by 2 experienced faculty members. The scores were statistically compared to identify the best-performing model using the Friedman and Wilcoxon tests. Moreover, the evaluators were asked to provide a qualitative evaluation of the comprehensiveness, scientific accuracy, clarity, and relevance of the LLMs' answers.ResultsOverall, no statistically significant difference was detected between the scores given by the 2 evaluators; therefore, an average score was computed for every LLM. Although ChatGPT-4 statistically outperformed ChatGPT-3.5 (P=.008), Bing Chat (P=.049), and Bard (P=.045), all models occasionally exhibited inaccuracies, generality, outdated content, and a lack of source references. The evaluators noted instances where the LLMs delivered irrelevant information, vague answers, or information that was not fully accurate.ConclusionsThis study demonstrates that although LLMs hold promising potential as an aid in the implementation of evidence-based dentistry, their current limitations can lead to potentially harmful health care decisions if not used judiciously. Therefore, these tools should not replace the dentist's critical thinking and in-depth understanding of the subject matter. Further research, clinical validation, and model improvements are necessary for these tools to be fully integrated into dental practice. Dental practitioners must be aware of the limitations of LLMs, as their imprudent use could potentially impact patient care. Regulatory measures should be established to oversee the use of these evolving technologies.

Project description:BackgroundLarge language models (LLMs) are emerging artificial intelligence (AI) technologies refining research and healthcare. However, the impact of these models on presurgical planning and education remains under-explored.ObjectivesThis study aims to assess 3 prominent LLMs-Google's AI BARD (Mountain View, CA), Bing AI (Microsoft, Redmond, WA), and ChatGPT-3.5 (Open AI, San Francisco, CA) in providing safe medical information for rhinoplasty.MethodsSix questions regarding rhinoplasty were prompted to ChatGPT, BARD, and Bing AI. A Likert scale was used to evaluate these responses by a panel of Specialist Plastic and Reconstructive Surgeons with extensive experience in rhinoplasty. To measure reliability, the Flesch Reading Ease Score, the Flesch-Kincaid Grade Level, and the Coleman-Liau Index were used. The modified DISCERN score was chosen as the criterion for assessing suitability and reliability. A t test was performed to calculate the difference between the LLMs, and a double-sided P-value <.05 was considered statistically significant.ResultsIn terms of reliability, BARD and ChatGPT demonstrated a significantly (P < .05) greater Flesch Reading Ease Score of 47.47 (±15.32) and 37.68 (±12.96), Flesch-Kincaid Grade Level of 9.7 (±3.12) and 10.15 (±1.84), and a Coleman-Liau Index of 10.83 (±2.14) and 12.17 (±1.17) than Bing AI. In terms of suitability, BARD (46.3 ± 2.8) demonstrated a significantly greater DISCERN score than ChatGPT and Bing AI. In terms of Likert score, ChatGPT and BARD demonstrated similar scores and were greater than Bing AI.ConclusionsBARD delivered the most succinct and comprehensible information, followed by ChatGPT and Bing AI. Although these models demonstrate potential, challenges regarding their depth and specificity remain. Therefore, future research should aim to augment LLM performance through the integration of specialized databases and expert knowledge, while also refining their algorithms.Level of evidence 5

Project description:BackgroundRecent advancements in artificial intelligence (AI) have sparked interest in its integration into clinical medicine and education. This study evaluates the performance of three AI tools compared to human clinicians in addressing complex orthopaedic decisions in real-world clinical cases.Questions/purposesTo evaluate the ability of commonly used AI tools to formulate orthopaedic clinical decisions in comparison to human clinicians.Patients and methodsThe study used OrthoBullets Cases, a publicly available clinical cases collaboration platform where surgeons from around the world choose treatment options based on peer-reviewed standardised treatment polls. The clinical cases cover various orthopaedic categories. Three AI tools, (ChatGPT 3.5, ChatGPT 4, and Bard), were evaluated. Uniform prompts were used to input case information including questions relating to the case, and the AI tools' responses were analysed for alignment with the most popular response, within 10%, and within 20% of the most popular human responses.ResultsIn total, 8 clinical categories comprising of 97 questions were analysed. ChatGPT 4 demonstrated the highest proportion of most popular responses (proportion of most popular response: ChatGPT 4 68.0%, ChatGPT 3.5 40.2%, Bard 45.4%, P value < 0.001), outperforming other AI tools. AI tools performed poorer in questions that were considered controversial (where disagreement occurred in human responses). Inter-tool agreement, as evaluated using Cohen's kappa coefficient, ranged from 0.201 (ChatGPT 4 vs. Bard) to 0.634 (ChatGPT 3.5 vs. Bard). However, AI tool responses varied widely, reflecting a need for consistency in real-world clinical applications.ConclusionsWhile AI tools demonstrated potential use in educational contexts, their integration into clinical decision-making requires caution due to inconsistent responses and deviations from peer consensus. Future research should focus on specialised clinical AI tool development to maximise utility in clinical decision-making.Level of evidenceIV.

Dataset Information

Benchmarking large language models' performances for myopia care: a comparative analysis of ChatGPT-3.5, ChatGPT-4.0, and Google Bard.

Background

Methods

Findings

Interpretation

Funding

Publications

Benchmarking large language models' performances for myopia care: a comparative analysis of ChatGPT-3.5, ChatGPT-4.0, and Google Bard.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets