Humans-written versus ChatGPT-generated case reports

Abstract

Aim

Artificial intelligence, especially ChatGPT, has been used in various aspects of medicine; however, whether ChatGPT can be used in case report writing is unknown. This study aimed to provoke discussion and provide a platform for it.

Methods

I wrote a theoretical case report where cyst aspiration cured a twisted ovarian cyst (Manuscript 4). I tasked ChatGPT with generating case reports by inputting information at three different levels: (1) key message and case profile, (2) addition of key introduction information (including known facts and problems to be solved), and (3) further addition of main discussion points. These inputs resulted in the creation of Manuscripts 1–3, which were subjected to analysis. Manuscript 3, generated by ChatGPT with the deepest information input, was compared with Manuscript 4, the human-authored counterpart.

Results

With the least information, Manuscript 1 can stand on its own, but its content is superficial. The more detailed data input, the more readable and reasonable the manuscripts become. A human-written manuscript involves personal experience and viewpoints other than obstetrics-gynecology.

Conclusions

Better input produced more reasonable and readable case reports. Human-written paper, compared with ChatGPT-generated one, can involve “human touch.” Whether such human touch enriches the case report awaits further discussion. Whether ChatGPT can be used in case report writing, and if it can, to what extent, should be worthy of further study. I encourage every doctor to form their own stance towards ChatGPT use in medical writing.