Article Text

Download PDFPDF

Comparison of outcome measures in patients with COPD
  1. Health Services and Respiratory and Environmental Research Units
  2. Institut Municipal d’Investigació Mèdica
  3. C/ Doctor Aiguader 80
  4. 08003 Barcelona, Spain
  1. School of Health and Related Research
  2. University of Sheffield
  3. Department of Medicine and Pharmacology
  4. University of Sheffield
  5. Sheffield S1 4DA, UK

Statistics from

Harper et al 1 found completion rates and success scaling rates of the St George’s Respiratory Questionnaire (SGRQ) lower than those observed in other questionnaires completed by patients with COPD attending an outpatient clinic. Readers may infer from this finding that the SGRQ is a less useful instrument than the Chronic Respiratory Questionnaire (CRQ), but an alternative explanation may lie on some methodological limitations of the study.

Firstly, the CRQ was administered only to an opportunistic subsample and, unfortunately, the authors provide insufficient information to rule out systematic differences with the rest of the patients. Secondly, the claim that SGRQ had lower completion rates may reflect the local condition in which it was applied rather than a general limitation. The CRQ was administered by an interviewer while the SGRQ was self-administered, and it is well established that self-administered questionnaires are more prone to missing items than those interviewer-administered. Moreover, supervision of completion can reduce missing data substantially. In a study of ours in 321 men with COPD,2 75% of patients self-completed the SGRQ and only 22% of individuals with missing items were observed. Missing information was reduced to 0% after the imputation algorithm recommended by the developers of the questionnaire was applied. It is important to note that Harper et al did not apply the algorithm of imputation in the SGRQ but they did impute the missing items of the CRQ. An appropriate design to allow for comparisons requires, at least, that the same instruments be administered to all the individuals controlling for the order and type of administration.3

Finally, the small sample sizes used may have influenced their results in general and, in particular, do not allow the authors to be conclusive about the lower (54%) success scaling rate of the SGRQ—that is, the proportion of items which correlate >0.4 with their hypothesised dimension. In our study substantially higher success scaling rates were found (78%).2

In conclusion, we believe that the comparison of the SGRQ with the CRQ and other instruments by Harper et al is of interest but is inconclusive. A more accurate and specific comparison is therefore still needed.


authors’ reply We thank Dr Ferrer and colleagues for their interest in our paper. We examined our results in relation to completion in two ways; firstly, patients who completed or did not complete all items of the Impact dimension and found small and inconsistent differences between them; and, secondly, patients who completed all four questionnaires or completed three or less and found that the only significantly different patient characteristic was in the distance walked. We feel that such differences are unlikely to explain those observed in general for completion and consistency for the SGRQ.

At the time of our study a method of imputing was under development for the SGRQ. Besides, we have doubts about the validity of substitution for high rates of missing data, nor are we convinced that encouraging patients to complete omitted items produces valid responses. More supervision than was available in our study may have increased the level of completion at some cost, but not necessarily the level of consistency. However, for routine clinical use what we need are questionnaires which do not require any supervision, such as a self-complete version of the CRQ.

The low correlation of items with their hypothesised dimensions for the SGRQ may be indicative of our patient group, which was clearly described as elderly and with longstanding disease, but in our experience this group is typical of those attending outpatient clinics in the UK.

We would like to point out that in our concluding sentence we were careful to express reservations about the instruments of choice. We suggested further development of the two condition-specific questionnaires and possible additions to the generic SF-36, and would certainly like to encourage more specific research into patient perceived quality of life measures for this significant patient group.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.