Validation of a clinical scoring system to detect gastro-oesophageal reflux disease (GERD) in epidemiological surveys

Abstract

BACKGROUND: The prevalence of GERD is increasing worldwide. Community prevalence in Sri Lanka is unknown. OBJECTIVES: To develop a clinical score to screen for GERD in the community and assess whether a score using symptom frequency and severity correlates better to an objective measure of GERD than one using only symptom frequency. METHODS: A cross-sectional validity study carried out on 72 patients (endoscopy positive) and 75 controls (comparable in age and gender). All faced a GERD-specific interviewer-administered questionnaire assessing seven upper gastro-intestinal symptoms, each graded for frequency (4- items) and severity (5-items). Two scores were generated. Score 1: sum of frequency of symptoms and score 2: sum of products of frequency and severity of each symptom. All patients underwent 24 h pH-metry. Validity established by correlating symptom scores with 24 h pH-metry parameters. Cut-off values determined by receiver-operating characteristic curves. RESULTS: Mean scores of cases were significantly higher than controls (p < 0.001). Cut-off score for score 1 was ≥11.50 (sensitivity 91.7%, speci- ficity 82.7%, positive and negative predictive values 70.0% and 95.9%). Cut-off score for score 2 was ≥14.50 (sensitivity 94.4%, specificity 78.7%, positive and negative predictive values 66.0% and 97.0%). Both scores showed high reproducibility (intra-class correlation coefficient of score 1 = 0.95 and score 2 = 0.89). There was good correlation between symptom scores and 24-h pH parameters (Spearman rank correlation, p = 0.01), but score 2 showed a significantly better correlation. CONCLUSIONS: Our GERD questionnaire is valid, reproducible, with better correlation with an objective test when both severity and frequency of symptoms were scored