Professional Documents
Culture Documents
need to be addressed. Thus the aims of this study were were modified accordingly. A batch of 28 patients with
(1) to develop a questionnaire for Chinese patients with acid reflux symptoms and erosive oesophagitis on
GERD (Chinese GERDQ) and (2) to validate the ques- endoscopy were interviewed to discuss these reflux
tionnaire by examining its psychometric properties. symptoms and to volunteer further symptoms that they
encountered. The response and wording were discussed,
and the questionnaire was modified accordingly.
METHODS The questionnaire consisted of questions on the severity
or frequency of 20 gastrointestinal symptoms related to
Patients and controls
gastro-oesophageal reflux disease. The method of scoring
The gold standard of GERD in this study was either was suggested, discussed and agreed by our panel of
a positive 24-h ambulatory oesophageal pH study6 or gastroenterologists, family physicians and statisticians.
the presence of endoscopic oesophagitis.7 One hundred The severity and frequency of the symptoms were
GERD patients (63 patients with erosive oesophagitis, graded on a five-point Likert scale as follows: 1 (none:
37 patients with positive 24-h ambulatory oesophageal no symptoms/none in the past year); 2 (mild: symptoms
pH study) were recruited. These 100 patients were can be easily ignored/less than once per month); 3
randomly identified from the endoscopy database7 and (moderate: awareness of symptoms but easily tolerated/
the out-patient clinic records of the Gastroenterology ‡ once per month); 4 (severe: symptoms sufficient to
Clinic of the Department of Medicine, Queen Mary cause an interference with normal activities/‡ once
Hospital. Those with symptoms of weight loss, gastro- weekly); 5 (incapacitating: incapacitating symptoms
intestinal bleeding or abnormal physical or laboratory with an inability to perform daily activities or requiring
findings from prior out-patient visits were excluded from a day off work/‡ once daily).9 All patients with GERD
study. Additional exclusion criteria were inability to were also subjected to a validated Chinese version of a
understand Chinese, cognitive impairment affecting quality-of-life score (SF-36) to measure construct valid-
understanding of the questionnaire, previous gastric ity.10 In addition, all patients were asked whether the
surgery, active non-steriodal anti-inflammatory drug symptoms assessed were relevant to their illness and
intake, history of malignancy or significant systemic whether the wording of the individual items was easily
disease. Controls were recruited from healthy subjects understood. The first part of the questionnaire was
aged 18 or above accompanying patients to out-patient administered again 7–14 days after completion of the
clinics. Those under active medical care for gastro- first questionnaire. Comparison between the two
intestinal complaints or taking regular medication for responses was used to determine test–retest reproduci-
other medical illnesses were excluded. However, we did bility of individual responses. Since there is no direct
not exclude subjects with gastrointestinal symptoms translation of the word ‘heartburn’ in the Chinese
not requiring treatment. This study was approved by language, a burning pain or discomfort behind the
the Ethics Committee of the University of Hong Kong. breastbone rising up towards the neck was used as the
definition of heartburn in the current study. This
definition has been found to identify more patients with
Development of the questionnaire
heartburn according to the Genval workshop report.11
The questionnaire was developed from the reviews of
published papers and interviews with patients. The Data analysis and statistics. Statistical analysis was per-
major symptoms were identified from a review of the formed using SPSS (SPSS inc. USA) and JMP (SAS, USA)
literature. The main framework was based on a statistical softwares. Discriminatory power and aspects of
validated questionnaire published previously.8 A range questionnaire reliability and validity were assessed.
of symptoms were identified, inclusive of questions that
are encountered in our practice, such as ‘acid feeling in Discriminatory power of the questionnaire. In performing
the stomach.’ These were discussed among the members cross-section comparisons, an instrument will need to
of the research team comprising gastroenterologists and have good discriminative power to detect differences
family physicians in order to form the initial question- between individuals and groups at a given point in
naire. Those questions were translated, back-translated time.12 The ability of the questionnaire to distinguish
and compared with the original questions. Differences between normal controls and patients with GERD was
tested using a multiple logistic regression model with GERD score calculated from the items was correlated with
group as the dependent variable and the 20 items as the SF-36 quality-of-life domains by Kendall’s tau coefficient.
explanatory variables. To determine the best model, the
explanatory variable with the largest P-value based on Discriminant validity. Discriminant validity is a measure
the likelihood ratio test was eliminated at each step of the tool’s ability to distinguish clinically significant
until all the variables had P-values less than 0.15. A differences in therapeutic responses between patients
composite score, derived from the seven items that and within patients over time.16 In this study
accounted for most of the differences between the we assessed discriminant validity in GERD patients by
controls and GERD patients, was generated by principal- administering the questionnaire before and after
component analysis. A cut-off score that best discrimi- 4 weeks of proton pump inhibitor treatment. Patients
nated between GERD and healthy controls was reporting an improvement in GERD symptoms were
determined by plotting a receiver operating characteristic expected to have a lower symptom severity score. In
curve.13 Sensitivity and specificity were determined. addition, after the completion of 4 weeks proton pump
inhibitor treatment, patients with resolution of acid
Reliability and internal consistency. Test–retest reliability reflux symptoms were stepped down to antacid treat-
of a measurement tool refers to the stability of ment for 4 weeks. The symptom score of patients after
measurement under different conditions. In the present cessation of proton pump inhibitor treatment was
study, test–retest reliability was assessed between base- compared with their own symptom scores while on
line and retesting after an interval of 10–14 days by the proton pump inhibitor treatment.
intra-class correlation coefficient between the first and
second responses to questionnaire administration. An
RESULTS
intra-class correlation of 0.7 or above is generally
considered to be sufficient to demonstrate reliability. A total of 201 subjects were recruited: 100 patients (38
Internal consistency of the questionnaire items was females, age 52 ± 12.2) and 101 controls (55 females,
assessed by Cronbach’s alpha coefficient. A Cronbach’s age 53 ± 13.1). All subjects completed the study.
alpha value of 0.7 is generally considered to be sufficient
to demonstrate internal consistency.14
Discriminatory power
Content validity. Content validity relates to the choice, The ability of the questionnaire to distinguish between
appropriateness, importance and representativeness of normal controls and GERD patients was tested using a
the instrument’s content.12 The individual components multiple logistic regression model with group as the
in the questionnaire must be seen as being compre- dependent variable and the 20 item scores as the
hensible and relevant to the multiple gastrointestinal explanatory variables. The variable with the largest P-
symptoms suffered by GERD patients. In this study, value based on the likelihood ratio test was eliminated
subjects were asked if the individual items assessed in at each step until all variables had P-values less than
the questionnaire were easily understood. Relevance of 0.15. The final model was tested for misclassification of
symptoms to disease was assessed in the GERD patients. cases and controls. Seven question items that best
We arbitrarily defined a question to have poor face accounted for the difference among patients with GERD
validity if less than 80% of subjects rated the item as and controls were selected into the final questionnaire
being comprehensible and relevant. (Table 1). Using this model, the probability of misclas-
sification as a control was 18% and the probability of
Construct validity. A construct is a theoretical idea misclassification as GERD is 16%, giving a sensitivity of
developed to explain and to organize some aspects of 82% and specificity of 84% based on the data from the
existing knowledge.15 Construct validity, in this aspect, 201 patients and controls. Using principal component
refers to evaluation of a test by looking at the analysis, relative weightings of each of the seven
relationship between the test and the various pheno- symptoms of the first principal component were assessed
mena which the theory predicts. In this study it was (Table 2). Since the weightings were similar to each
assumed that patients with more severe GERD other, the items were not individually weighted. The
symptoms would have a lower quality of life. The total sum of the seven item scores was used as the
Table 1. Questionnaire items, percentage of GERD patients and controls who reported positive symptoms and percentage of subjects who
considered the symptom assessed as easily comprehensible
Weighting by principal-
Symptom assessed component method
12 Hutchinson A, Bentzen N, Konig-Zahn C. Cross cultural 18 Shaw MJ, Talley NJ, Beebe TJ, et al. Initial validation of a
health outcome assessment: a user’s guide 1994. diagnostic questionnaire for gastroesophageal reflux disease.
13 Erdreich LS, Lee ET. Use of relative operating characteristic Am J Gastroenterol 2001; 96: 52–7.
analysis in epidemiology. A method for dealing with subject- 19 Ofman JJ, Shaw M, Sadik K, et al. Identifying patients with
ive judgment. Am J Epidemiol 1981; 114: 649–62. gastroesophageal reflux disease: validation of a practical
14 Nunnally JC. Psychometric Theory, 3rd edn. New York: screening tool. Dig Dis Sci 2002; 47: 1863–9.
Mc Graw Hill, 1994. 20 Fass R, Fennerty MB, Ofman JJ, et al. The clinical and eco-
15 Silva F. Psychometric Foundations and Behavioural nomic value of a short course of omeprazole in patients with
Assessment. Newbury Park, California: Sage Publications, noncardiac chest pain. Gastroenterology 1998; 115: 42–9.
1993. 21 Juul-Hansen P, Rydning A, Jacobsen CD, Hansen T. High-dose
16 Bombardier C, Tugwell P. A methodological framework to proton-pump inhibitors as a diagnostic test of gastro-
develop and select indices for clinical trials: statistical and oesophageal reflux disease in endoscopic-negative patients.
judgmental approaches. J Rheumatol 1982; 9: 753–7. Scand J Gastroenterol 2001; 36: 806–10.
17 Carlsson R, Dent J, Bolling-Sternevald E, et al. The usefulness 22 Spechler SJ, Jain SK, Tendler DA, Parker RA. Racial differ-
of a structured questionnaire in the assessment of sympto- ences in the frequency of symptoms and complications of
matic gastroesophageal reflux disease. Scand J Gastroenterol gastro-oesophageal reflux disease. Aliment Pharmacol Ther
1998; 33: 1023–9. 2002; 16: 1795–800.