Using Artificial Intelligence ChatGPT to Access Medical Information About Chemical Eye Injuries: Comparative Study
Bibliographic record
Abstract
BACKGROUND: Background: Chemical ocular injuries are a major public health issue. They cause eye damage from harmful chemicals and can lead to severe vision loss or blindness if not treated promptly and effectively. Although medical knowledge has advanced, accessing reliable and understandable information on these injuries remains a challenge. This is due to unverified online content and complex terminology. Artificial Intelligence (AI) tools like ChatGPT provide a promising solution by simplifying medical information and making it more accessible to the general public. OBJECTIVE: Objective: This study aims to assess the use of ChatGPT in providing reliable, accurate, and accessible medical information on chemical ocular injuries. It evaluates the correctness, thematic accuracy, and coherence of ChatGPT's responses compared to established medical guidelines and explores its potential for patient education. METHODS: Methods: Nine questions were entered to ChatGPT regarding various aspects of chemical ocular injuries. These included the definition, prevalence, etiology, prevention, symptoms, diagnosis, treatment, follow-up, and complications. The responses provided by ChatGPT were compared to the ICD-9 and ICD-10 guidelines for chemical (alkali and acid) injuries of the conjunctiva and cornea. The evaluation focused on criteria such as correctness, thematic accuracy, coherence to assess the accuracy of ChatGPT's responses. The inputs were categorized into three distinct groups, and statistical analyses, including Flesch-Kincaid readability tests, ANOVA, and trend analysis, were conducted to assess their readability, complexity and trends. RESULTS: Results: The results showed that ChatGPT provided accurate and coherent responses for most questions about chemical ocular injuries, demonstrating thematic relevance. However, the responses sometimes overlooked critical clinical details or guideline-specific elements, such as emphasizing the urgency of care, using precise classification systems, and addressing detailed diagnostic or management protocols. While the answers were generally valid, they occasionally included less relevant or overly generalized information. This reduced their consistency with established medical guidelines. The average FRES was 33.84 ± 2.97, indicating a fairly challenging reading level, while the FKGL averaged 14.21 ± 0.97, suitable for readers with college-level proficiency. Passive voice was used in 7.22% ± 5.60% of sentences, indicating moderate reliance. Statistical analysis showed no significant differences in FRES (p = .385), FKGL (p = .555), or passive sentence usage (p = .601) across categories, as determined by one-way ANOVA. Readability remained relatively constant across the three categories, as determined by trend analysis. CONCLUSIONS: Conclusions: ChatGPT shows strong potential in providing accurate and relevant information about chemical ocular injuries. However, its language complexity may prevent accessibility for individuals with lower health literacy and sometimes miss critical aspects. Future improvements should focus on enhancing readability, increasing context-specific accuracy, and tailoring responses to person needs and literacy levels. CLINICALTRIAL: This is not RCT.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.011 | 0.056 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".