Accuracy of provisional diagnoses of dyspepsia in patients undergoing first endoscopy.

Background: This study assessed agreement between provisional and endoscopic diagnoses for patients with dyspepsia undergoing initial endoscopy, and examined variation between clinicians at 2 hospitals.

Methods: This was a retrospective review of 423 consecutive patients.

Results: Crude percentage agreement ranged from 55% to 97%. Kappa scores revealed poor agreement: peptic ulcers (0.11: 95% CI [0.05, 0.17]); gastroesophageal reflux disease (0.29: 95% CI [0.20, 0.38]); benign esophageal stricture (0.33: 95% CI [0.08, 0.58]); and cancer (0.12: 95% CI [-0.12, 0.36]). Positive and negative predictive values for cancer and benign esophageal stricture showed that agreement for a negative diagnosis was almost perfect, whereas agreement for a positive diagnosis was low. Only 17% of patients with cancer were given an accurate provisional diagnosis. Accuracy of diagnosis did not vary substantially between hospitals.

Conclusions: Crude percentage agreement is misleading. Emphasis should be placed on better prediction of cancer, benign esophageal stricture, and peptic ulcer disease. Accuracy of provisional diagnosis in everyday practice is no worse than that found in prospective studies in which clinicians knew a priori that diagnoses would be scrutinized. The difficulty of predicting diagnoses supports increased reliance on endoscopy.