Symptom evaluation as an efficient diagnostic tool for upper abdominal diseases

Introduction: GERD and dyspepsia are common conditions, affecting approximately 25-40% of the general population. In the absence of alarm symptoms, the current recommended policy in young dyspeptic patients is a "test and treat" strategy for H. pylori. On the other hand, in GERD patients, a therapeutic trial with proton pump inhibitors (PPI) is the treatment of choice. Aim: This study aimed to create short and simple clinical algorithms, for the diagnosis and treatment of patients with upper gastrointestinal complaints. Methods: Data mining models and algorithms (neural networks, decision trees and logistic regression) were used to create a diagnostic symptom questionnaire to classify the patients into one of two diagnostic groups: GERD vs. non-GERD. The questionnaire was validated against endoscopic and clinical diagnoses of 132 patients in a cross-sectional study, and was designed to yield a "GERD score". The clinical and economical benefits of the new algorithm were evaluated in primary care clinics in Israel. Results: The symptoms chosen for the diagnostic questionnaire were heartburn, acid regurgitation, sour oral taste, aggravation of symptoms after heavy meals, relief of symptoms by antacids, and nocturnal reflux. The use of the algorithm by primary care physicians improved clinical outcomes and reduced health resources consumption. Conclusions: This new short algorithm was found to be useful and easy to apply in clinical practice. The questionnaire can be also implemented in computerized medical systems.