Background and objective: Predictive symptomatic scoring models have been proposed to select patients with dyspepsia, who may be candidates to endoscopy. In a previous study performed by gastroentrologists from our group, we obtained three scales of symptoms to predict organic dyspepsia, peptic ulcer and esophagitis, respectively. Here we analyze the reproducibility of those scoring models of symptoms when used either by other gastroenterologists or general practitioners.
Patients and method: It was a clinical prospective study of 230 patients from the Viladecans hospital area (120 from primary healthcare and 110 from the gastroenterologist's consultation). The three validation scales were performed in each patient. Then, we performed a diagnostic gastroscopy which allowed to classify patients into those with organic dyspepsia (ulcer, esophagitis, cancer) or those with functional dyspepsia. We calculated the overall predictive accuracy for the gastroenterologist and the general practitioner and for the three diagnoses (organic dyspepsia, ulcer or esophagitis) using the C statistic.
Results: Discriminative capacities were 0.75 and 0.82 for organic dyspepsia, 0.78 and 0.86 for ulcer disease, and 0.78 and 0.82 for esophagitis, for the general practitioner and the gastroenterologist, respectively. In this validation study, the best cut-off value, namely the one combining good sensitivity and specificity, was found to be 7.
Conclusions: A correct predictive capacity of the symptomatic score models when used by other gastroenterologists or by primary healthcare practioners confirms its reproducibility and transferability. The use of predictive symptomatic score models in everyday clinical practice can allow to rationalize the referral for endoscopy in our local setting.