Given the constraints of time that are imposed on medical staff, tools to provide quick and accurate information in an easily accessible form could see more prove useful. However, computerised aids are not always readily accepted by medical staff ,  and . We have shown that NLG technology can indeed be employed successfully in a medical setting to produce compact, targetted textual summaries of a patient’s history. In particular, we show that such summaries of large medical datasets can significantly improve the efficiency
of clinicians in certain critical settings. Moreover, the clinicians in our study were overwhelmingly enthusiastic about the automatically generated summaries, a finding that is particularly encouraging given the novelty
of the documents and the natural reluctance of clinicians towards computer-generated reports. The familiarity of the textual medium no doubt played an important role in the success of our system. Combined with graphical facilities, we suspect that it may be possible to Z-VAD-FMK increase even further the efficiency of clinicians in the specific context of making an initial assessment of a patient based solely on their medical history, and we are now investigating this. Although the study reported here focuses on cancer treatment, the techniques that underpin the Report Generator can be applied to almost any medical context. Nevertheless, the Report Generator is to-date a proof-of-concept research system; transformation to a full-deployable clinical tool would require further
software development and testing. Additionally, as with any data-presentation system, the accuracy of the generated summary is fully dependent on the accuracy of its input, in this case: Data quality : the accuracy of the data contained in the Acyl CoA dehydrogenase patient record; In the language of AI, this is termed “garbage-in, garbage-out”. This study demonstrates that AI technology can be successfully employed to write textual summaries of a patient’s medical history. Such summaries are not only accurate (to the extent that the recorded patient data is accurate), but can provide clinicians with key information about a patient’s history in about half the time that it would take if the clinician were instead having to search through the patient’s textual record. A significant portion of a clinician’s time is taken up with non-clinical tasks such as reading the medical records of patients that they are about to see, or having seen the patient, writing letters or reports about the patient. Automatically generated summary overviews of a patient’s medical history can potentially enhance doctor–patient interactions by significantly reducing the time required for doctors to carry out some of these tasks. The authors have no conflict of interest to declare.