Using ChatGPT for writing hospital inpatient discharge summaries - perspectives from an inpatient infectious diseases service

Abstract

Background: Hospital discharge summaries are important tools for communication between healthcare professionals. They convey events that occurred during hospitalisation, as well as the subsequent follow-up plans. Artificial intelligence models can be used to summarise information succinctly from large amounts of raw data input. We explored ChatGPT's ability to generate effective discharge summaries to assist junior doctors in writing these documents.

Methods: We constructed three hypothetical scenarios of inpatient encounters, with three different outcomes: i) discharge home with follow-up with a general practitioner, ii) discharge to a stepdown facility for further physical rehabilitation, iii) transfer to a tertiary centre for more advanced care. ChatGPT was used to generate discharge summaries for these three scenarios. The quality of the responses provided were evaluated.

Results: ChatGPT was able to provide an effective framework for discharge summaries. It processed large volumes of text, summarising pertinent issues and communicating follow-up plans clearly. It is a potentially useful tool for documentation for clinicians. However, pitfalls remain, where close reading is still required to ensure the veracity of the output provided.

Conclusions: ChatGPT was able to synthesize patient information from a long prosaic format to provide a structured discharge summary. Future prospective study could evaluate if this framework provided by ChatGPT is helpful to aid junior doctors in learning about and writing discharge summaries more efficiently.

Keywords: Artificial intelligence; GPT; Hospital inpatient discharge summary; Infectious diseases.

Fig. 1

Contextualisation given to GPT-3.5 and process for generating hospital inpatient discharge summaries