[Acute abdominal pain--standardized findings as diagnostic support. Results of a prospective multicenter intervention...

Despite powerful diagnostic tools (e.g. ultrasound, special laboratory investigations), the diagnosis of acute abdominal pain is still a considerable problem. Several studies in the UK have shown that the diagnostic accuracy can be improved by structured and standardized history taking and clinical examination and by computer-aided diagnosis. In the framework of a concerted action of the European Community we have conducted a prospective multicenter interventional trial comparing two consecutive phases: a) a baseline phase in clinical routine without additional intervention, b) a test phase with structured and standardized history and clinical examination (questionnaire, documentation programme). In addition, a computer-aided diagnostic system developed in the UK was applied to the cases in the test phase. Outcome criteria were the diagnostic accuracy of the initial and the final examiner, the perforated appendix rate, the negative appendectomy rate, the negative laparotomy rate and the rates of diagnostic errors with missing indication to operation and of delayed urgent operations. No differences could be found between the phases with respect to the outcome criteria. In the baseline phase (test phase) diagnostic accuracy was 59% (59%), diagnostic accuracy after investigation (senior examiner) was 77% (78%), perforated appendix rate was 11% (16%), negative appendectomy rate was 13% (15%), negative laparotomy rate was 7% (8%), the rate of missed urgent indications to operation was 1.1% (1.9%) and the rate of delayed urgent operations was 3.4% (2.4%). Major differences between the centers were recorded. Computer-aided diagnosis resulted in a diagnostic accuracy of 51%. The introduction of structured and standardized history taking and clinical examination has not brought any improvement of the good results in clinical routine. It is doubtful, whether existing systems of computer-aided diagnosis are able to significantly decrease the still remaining error rate of 20%.