Aims Paediatric endoscopy is an invaluable diagnostic tool however there is little evidence base to guide clinicians in appropriate selection of patients for endoscopy. This study aimed to evaluate one unit’s use of endoscopy in newly presenting children with the intention of obtaining data for a prospective study of selection of patients for paediatric endoscopy.

Methods Patient demographics, clinical indications, past medical and family history, macroscopic and histological findings over a period of 6 months were retrospectively reviewed. First diagnostic endoscopic procedures were included in the study. We excluded follow up or therapeutic endoscopies.

Results 218 endoscopies were reviewed in a total of 164 children: 90 having gastroscopy only, 20 having colonoscopy only and 54 patients undergoing both upper and lower gastrointestinal (GI) endoscopies.

49% of all children had both macroscopically and histologically normal endoscopy findings, Macroscopic and histological abnormalities (respectively) for each group were found as follows: 44% and 28% of gastroscopy patients, 25% and 25% of colonoscopy patients, and 53% and 53% of patients undergoing both upper and lower GI endoscopies.

For patients undergoing gastroscopy alone, diarrhoea led to the highest rate of abnormal histological outcome (PPV = 0.31). For patients undergoing colonoscopy only, abdominal pain was the highest predictor of abnormal histology (PPV = 0.38). For patients undergoing both upper and lower GI endoscopies, diarrhoea was the highest predictor for abnormal histology (PPV = 0.61).

In gastroscopy, the positive predictive value (PPV) of having abnormal histology if the endoscopy was macroscopically reported as abnormal was only 0.4 compared to colonoscopy alone and both upper and lower endoscopy patients (PPV 0.8 and 0.93 respectively).

Conclusions This study shows that 51% of patients undergoing diagnostic endoscopies have findings at endoscopy. The rate of histological abnormality in gastroscopy alone was only 28% (including patients with positive coeliac antibodies). There was marked discrepancy between endoscopic abnormalities and histological findings in gastroscopy alone. This study shows that the clinical application of predictive models for endoscopy warrants further research and may contribute to developing guidelines to appropriately select patients for endoscopy.