ChatGPT has become our preferred assistant to carry out a large number of tasks. However, it is still far from being important in certain industries: the medical field, the best example of this.
ChatGPT has had great penetration and notoriety in a large number of sectors: marketing, journalism or design are just a few examples. Their apparent good results invite you to want to use it in a large number of situations, crossing any border. However, there are many fields in which there is still a long way to go, such as the medical field.
A study carried out by Jama Pediatrics, a medical journal specialized in the entire pediatric medical field, has shown how ChatGPT accuracy rate to diagnose pediatric cases only reaches 17%.
Long way to go
The study highlights the importance that clinical experience may have when evaluating medical cases related to children. While it is true that, sooner or later, ChatGPT must integrate into clinical care to help with a large number of tasks that the medical team is currently in charge of, the reality is that there is still a long way to go until this Artificial Intelligence tool can provide real help in this situation.
In July 2023, the same magazine already carried out a similar study. In this case, to evaluate the AI precision for complex diagnoses and, in it, it was shown that its success rate reached 39%. Well above the latest results obtained.
Historically, medicine has always been very proactive in receiving all types of innovations that allow reorienting the available resources to focus medical personnel on those situations that really require their experience and knowledge. Previously, AI has been used to create racial biases or to interpret scans carried out. In addition to the automation of administrative tasks, the main use of ChatGPT today in this field.
Pediatric specialization
The difference in results between general medicine and pediatric specialization has to do with the number of elements that can interfere when carrying out the corresponding work on the little ones. Since it is key to consider the age of the patient and, based on it, it is possible that the diagnosis differs from each other. Even more so if we take into account that many children are not yet at the age to be able to identify or transmit the symptoms they experience when they suffer any alteration in their body.
One of the main complaints from the medical team that verified the results was that ChatGPT failed to fully capture the diagnosis. In many cases, the results were too broad or nonspecific to qualify the diagnosis as correct.
In the final results, ChatGPT only managed to respond correctly in 17 of the 100 cases studied. In 72 of them, the diagnosis was very wrong. While in the remaining 11, it was not as precise as it should be.
