Newsletter September 2023: ChatGPT - What does AI know about Patient Safety?

Authors:
Daniel Arnal, Maria Wittmann

“ChatGPT” is an artificial intelligence (AI)-based large language model (LLM) trained on massive text datasets in multiple languages with the ability to generate human-like responses to text input (1).  

Developed since 2018 (2), the release of “ChatGPT” to the broader public in November 2022 (1) and its later dissemination seemed to bring an available artificial intelligence (AI) tool with astonishing potential. Most end-users had been reading about AI, big data, machine learning and neural networks but needed an actionable tool to experience it. It seemed we were talking to a wise human who responded coherently to all questions.  

Would its responses about patient safety (PS) be coherent? We needed to make the try. In February 2023, we worked on the SAFEST EU project (3) for a few months to develop a multidisciplinary consensus for perioperative PS standardised practices. We had performed a systematic review of clinical practice guidelines (4) and were running a modified Delphi survey towards the agreement. What are the ten most recommended safety strategies in the perioperative period? (ChatGPT 3.5, February 2023): 

“There are several recommended safety strategies in the perioperative period to ensure patient safety during surgical procedures. Below, I present 10 of them: 

1. Patient Identity Verification: It is essential to verify the patient’s identity and the accuracy of surgical procedure data before the procedure. 
2. Surgical Site Marking: The surgical site should be clearly and visibly marked to prevent confusion during the procedure. 
3. Infection Control: Strict infection control measures should be followed to reduce the risk of healthcare-associated infections. 
4. Safe Medication Administration: Safe techniques should be used for medication administration, and the correct dose and medication should be verified. 
5. Effective Communication: Maintaining effective communication within the surgical team is crucial to prevent errors and ensure patient safety. 
6. Proper Equipment Preparation: The surgical team should be adequately prepared and have access to all necessary equipment and supplies for the surgical intervention. 
7. Use of Safety Technology: Safety technology, such as surgical information systems and real-time monitoring, can help prevent errors and enhance patient safety. 
8. Patient Evaluation and Follow-up: It is essential to evaluate and follow up with patients after the surgical intervention to detect and address any early complications. 
9. Staff Training and Education: Ensuring staff is well-trained and educated in safety strategies is essential to ensure their effectiveness. 
10. Risk Management: Measures should be implemented to identify, assess, and manage risks in the perioperative period and reduce the likelihood of adverse events. 

It is important to note that these are just some safety strategies that can be implemented in the perioperative period and that patient safety is a shared responsibility among the entire surgical team. “

Reading its response and the work we were taking on the SAFEST project, we had to admit that all these strategies had been identified during the synthesis of the evidence, and all of them were later agreed to be relevant for PS by the Delphi panel.  

Despite our initial successful experience asking about a PS general topic to this tool, using LLM in education, research, and practice is subject to controversy (5). ChatGPT has been criticised for superficial, inaccurate, or incorrect content, lack of transparency of the sources or errors in its citations, and the potential creation of a scientifically plausible hallucination (1). This being acknowledged, we could confirm that the AI response was coherent with our research and our personal experience. 

In any case, LLM is only one of the branches of AI applicable to PS and anaesthesia. Are there other potential benefits for PS?  

ChatGPT had its response (ChatGPT 3.5, August 2023): 

“Artificial intelligence (AI) has significant potential to contribute to patient safety in healthcare. Here are some aspects of patient safety where AI can make a positive impact: 

1. Diagnosis and Decision Support: AI can analyse large amounts of patient data, including medical records, laboratory results, and imaging scans, to assist healthcare professionals in making accurate and timely diagnoses. AI algorithms can help identify patterns, flag potential errors, and suggest appropriate treatment options, reducing diagnostic errors and improving patient safety. 
2. Medication Safety: AI can enhance medication safety by helping healthcare providers identify potential adverse drug interactions, allergies, or contraindications based on patient data. AI systems can also assist in medication reconciliation, ensuring that patients receive the correct medications and dosages during care transitions. 
3. Predictive Analytics: AI can analyse patient data and predict the likelihood of adverse events, such as hospital-acquired infections or falls, allowing healthcare providers to take proactive measures to prevent harm. AI can also identify patients at high risk for readmission, enabling targeted interventions to improve patient outcomes. 
4. Surgical Safety: AI can be used to optimise surgical planning, improve intraoperative decision-making, and enhance surgical safety. For example, AI algorithms can analyse pre-operative imaging to assist surgeons in identifying the best surgical approach and avoiding critical structures. In the operating room, AI can provide real-time feedback to surgeons, helping them avoid errors and complications. 
5. Monitoring and Early Warning Systems: AI-powered monitoring systems can continuously analyse patient vital signs, detect subtle changes, and issue alerts when a patient’s condition deteriorates. Early warning systems can enable timely interventions, reducing the risk of adverse events and improving patient safety, particularly in critical care settings. 
6. Clinical Workflow Optimisation: AI can streamline clinical workflows by automating repetitive tasks, facilitating information exchange, and reducing the burden on healthcare providers. By minimising administrative duties, AI allows clinicians to focus more on patient care, enhancing safety and improving overall healthcare delivery. 

However, it’s important to note that while AI has great potential, its implementation should be carefully managed and continuously validated. Ethical considerations, data privacy, transparency, and ongoing human oversight are critical to ensuring AI’s safe and effective use in healthcare. “

To support this response, a few publications are synthesising the potential benefits of the different applications of AI for PS or the perioperative period. A systematic review identified 53 heterogeneous studies of AI interventions and their quantitative impact on PS, most of them with positive results in terms of safety (6). In general and paediatric anaesthesia, AI has been applied in pre-operative anaesthesia, intra-operative anaesthesia, post-operative anaesthesia, and operating room management (7,8), in line with the chatGPT response. Patient risk factor prediction, anaesthetic depth estimation, anaesthetic medication/technique decision guidance, intubation assistance, airway device selection, physiological variable monitoring, and operating room scheduling are the main current AI applications. 

The potential of the different AI applications to improve PS is undoubtedly exciting. These lines highlight what will be there to research and validate. Exploring AI applications to enhance perioperative PS may be one strategy to be included in the top ten in the near future. However, it was outside the ChatGPT’s response about the topic. 

References 

1. Sallam M. ChatGPT Utility in Healthcare Education, Research, and Practice: Systematic Review on the Promising Perspectives and Valid Concerns. Healthc Basel Switz. 2023 Mar 19;11(6):887.
2. GPT-3. In: Wikipedia [Internet]. 2023 [cited 2023 Aug 20]. Available from: https://en.wikipedia.org/w/index.php?title=GPT-3&oldid=1170092033
3. Arnal D, Wittman M. Newsletter September 2022: ESAIC embarks on a new EU Project – SAFEST | ESAIC [Internet]. ESAIC; 2022 [cited 2023 Aug 20]. Available from: https://www.esaic.org/esa-news/newsletter-september-2022-esaic-embarks-on-a-new-eu-project-safest/
4. PROSPERO [Internet]. [cited 2023 Aug 20]. Synthesis of Evidence-Based Standardised Practices to Improve Perioperative Safety Period in Europe: Protocol of Systematic Review of Clinical Practice Guidelines- SAFEST project. Available from: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=347449
5. Grigio TR, Timmerman H, Wolff AP. ChatGPT in anaesthesia research: risk of fabrication in literature searches. Br J Anaesth. 2023 Jul;131(1):e29–30.
6. Choudhury A, Asan O. Role of Artificial Intelligence in Patient Safety Outcomes: Systematic Literature Review. JMIR Med Inform. 2020 Jul 24;8(7):e18599.
7. Bellini V, Rafano Carnà E, Russo M, Di Vincenzo F, Berghenti M, Baciarello M, et al. Artificial intelligence and anesthesia: a narrative review. Ann Transl Med. 2022 May;10(9):528.
8. Antel R, Sahlas E, Gore G, Ingelmo P. Use of artificial intelligence in paediatric anaesthesia: a systematic review. BJA Open. 2023 Mar;5:100125.

Acknowledgement 

AI tool ChatGPT 3.5 was used to respond to specific questions reported in this manuscript. Its responses were transcribed unaltered.