Integrating artificial intelligence (AI) into pediatrics can
revolutionize healthcare through innovative diagnostic methods, treatment
planning, and customized clinical decision support. In pediatric specialty
cases, ChatGPT displayed differing levels of accuracy. For example, the model
received a grade of C for misdiagnosing congenital heart disease, cardiogenic
shock, urinary tract infections, and bullous impetigo as persistent pulmonary
hypertension of the newborn, septic shock, acute pyelonephritis, and
Epidermolysis bullosa, respectively. These results emphasize both the model's
strengths and areas needing improvement.
When achieving an A grade, ChatGPT excelled in delivering
comprehensive information, including – diagnosis, differential diagnosis, causes,
complications, investigations, and treatment strategies. However, there was
evident room for improvement in scenarios where responses were merely
acceptable. The inaccuracies likely stem from the complexities and nuances of
these conditions, as well as overlapping symptoms that make the diagnoses
challenging. Further training on specific medical scenarios is necessary to
enhance the model's accuracy.
ChatGPT's potential as a Clinical Decision Support System (CDSS)
in pediatrics lies in its ability to provide detailed information across
various scenarios, making it a valuable tool for clinicians seeking additional
insights. Nevertheless, the identified inaccuracies in critical situations
highlight the need for caution when using ChatGPT for clinical decision
support. While it can serve as a valuable supplementary resource, it should not
replace the expertise and judgment of healthcare professionals.
ChatGPT is yet to be a substitute for standard textbooks and guidelines and should be used cautiously. Evaluating ChatGPT's performance in pediatric scenarios revealed varying proficiency levels, highlighting the necessity for ongoing refinement and collaboration with experienced pediatricians to ensure reliable decision support. While ChatGPT offers valuable insights and assistance, its reliance on preexisting data and lack of real-time updates warrant cautious use in the medico-legal context.
Source: Andykarayalar R, Surapaneni KM. Indian
Pediatrics. 2024 Mar 7:S097475591600610-.
Please login to comment on this article