[MANAGEMENT OF LABOR ANESTHESIA IN A PATIENT WITH EHLERS-DANLOS SYNDROME WHY DOES CHATGPT ERR IN SOURCE REFERENCING?]

Abstract

Introduction: Ehlers-Danlos Syndrome (EDS) encompasses a spectrum of inherited disorders, characterized by joint hypermobility, skin hyperextensibility, and other features. In some patients, EDS involves the vascular walls, posing a significant clinical challenge due to the resultant propensity for rupture with hemorrhagic complications. Such complications, among others that often occur with EDS, may carry particular importance in the context of pregnancy and labor. This paper presents a patient diagnosed with Ehlers-Danlos Syndrome during her first pregnancy. The patient was planned for elective Cesarean delivery and expressed an interest in regional anesthesia. A literature review was conducted to identify similar cases in which the anesthesia techniques and their potential complications were described, as well as additional risks for EDS patients associated with various anesthesia methods. The review identified only low-quality data, which suggested a higher pain threshold in patients with EDS and an increased risk of bleeding. This case was used to assess the ability of ChatGPT to present a literature review based on reliable sources when the evidence is sparse. The model generated a well-worded response using correct medical terminology, but the report was superficial and provided no data to support clinical decision-making. The model also suggested a different anesthetic approach than the human-generated literature review and supported its findings with links to cited sources. These sources were examined, and concerns regarding their reliability were raised. Lack of reliability remains a major challenge for developers and users of large language models. Fine-tuning (i.e. training the model with examples relevant to specific tasks) which may enhance model output accuracy is discussed in this context.