Objective: Gastrointestinal endoscopy databases are important for surveillance, epidemiology, quality control and research. A good quality of automatically generated databases to enable drawing justified conclusions based on the data is of key importance. The aim of this study is to validate the correctness of coding of a national automatically generated anonymous endoscopy database.
Material and methods: We evaluated a total of 500 colonoscopies performed in five larger hospitals of the TRANS.IT project focusing on endoscopy reporting. Randomly 500 examinations were selected from a total of 5,000 examinations and their generated endoscopic terminology codes as well as complete reports were analysed. Indications for the examination and described findings were scored for correctness and clinical relevance of the coding that would be exported to the anonymous database.
Results: Indications were correctly coded in 92% of all examinations (range 76-100%) per hospital. Correct coding of findings ranged from 42% to 93% per hospital (mean 77%). Different correct coding proportions were seen varying with the diagnosis, with the highest correct coding rates in polyps, carcinoma and diverticular disease. Incorrect coded examinations were scored for clinical relevance. Overall 11% of the investigated examinations were incorrectly coded with clinical relevance.
Conclusions: Accuracy of clinically relevant endoscopy data recorded in the TRANS.IT anonymous central database is high. Further improvement is desirable, which may be achieved by education of individual endoscopists and enhancement of the program.