Original articleA critical look at transition ratings
Introduction
Clinicians often face decisions whether to recommend treatments for symptomatic benefit. Knowing the typically wide heterogeneity in treatment response, clinicians often conduct “trials of therapy” [1]. Patients try a medication, and report the extent to which treatment ameliorated their symptoms. If improvement was absent or marginal, medication is discontinued. If patients report an important improvement, be it small, moderate, or large, they will generally agree to long-term medication.
Thus, the question: “Are you feeling better or worse, and if so, what is the extent of the change?” is integral to clinical practice. Questions of this sort, which require patients to remember a prior health state and compare it to how they are feeling currently, are labeled “transition” questions. Although most self-report health-related quality of life (HRQL) questionnaires ask patients to describe a current or recent health state, some include transition items.
Although transition questions have intuitive appeal, patients have considerable difficulty recalling prior health states [2]. Under these circumstances, patients appear to focus on their current health state. If they are feeling well, they rate themselves as improved; if feeling unwell, they rate themselves as having deteriorated.
In our prior work in HRQL, we have used transition questions to enhance target questionnaires' interpretability, or meaningfulness [3]. Global transition ratings provide an independent standard that are themselves easily interpretable. When they show a moderate to large correlation with the target HRQL instrument, they appear to meet all the requirements of an anchor that can enhance the interpretability of an HRQL measure [4]. Furthermore, a moderate to high correlation (we have used a threshold of 0.5) suggests the validity of the transition rating. For instance, if the correlation between a global transition rating of dyspnea (overall, how much better or worse is your dyspnea in daily activities) and the change in score on a dyspnea questionnaire is over 0.5, it supports the validity of both the transition rating and the target questionnaire.
This correlation between the change in the target measure and the transition rating may, however, be misleading. Assume that patients who are doing well have tended to improve and those who are doing badly have tended to deteriorate. Were this the case, one might see a substantial correlation between change in the target measure and the transition rating even if patient scores were reflecting not the degree to which they have changed, but rather their current health state.
This reasoning suggests that, to be confident of the validity of transition ratings, one would require evidence beyond confirmation of their expected degree of correlation with independent measures. How might one generate such evidence? As it turns out, if the variability of the pre and post test scores are equal, were transition measures working in the way they should, one would anticipate an equal and opposite correlation of the transition measure with the pre test score and the post test score. For instance, were variances in pre and post test scores more or less equal, one might expect a correlation of the transition score with the post test of 0.25, with the pre test of −0.25, and with the change score of over 0.5. The Appendix presents a formal proof of the relationship.
We have found, using global transition ratings as an anchor, a striking consistency in estimates of the minimum important difference (MID) of approximately 0.5 on a seven-point scale 5, 6, 7, 8. If the transition measures, despite correlations of >0.5 with change scores on the target questionnaires, are not working in the way they should, our estimates of the MID may be biased. For this reason, and in response to our curiosity about the extent to which patients' initial state bears on their global transition ratings, we explored the performance of transition ratings in three of our data sets. The three questionnaires used in this study share a similar structure, with response options framed as seven-point scales. Each has strong evidence of validity, and none use transition questions as part of the questionnaire. Thus, in this investigation, we assume the adequate and consistent performance of the questionnaires, and focus on the performance of the transition ratings.
Section snippets
The questionnaires
The Asthma Quality of Life Questionnaire (AQLQ) is a disease-specific 32-item instrument including four domains: symptoms, emotions, exposure to environmental stimuli, and activity limitations [9]. Patients rate the impairments they have experienced during the previous 14 days and respond to each item on seven-point scales. The instrument, in which higher numbers represent better function, has proved responsive and valid 10, 11, 12 and seen wide use.
The Chronic Respiratory Questionnaire (CRQ)
Results
We present the data regarding the correlations between the HRQL questionnaire scores and the transition ratings in Table 1, Table 2, Table 3, Table 4, Table 5. In each table, the first row presents the correlations between the prescore and the transition rating. The second row presents the correlations between the post score and the transition ratings. The third row presents the correlation between the change score (post–pre) and the transition rating. The fourth row presents the regression
Discussion
Transition ratings have a number of possible uses. First, clinicians and investigators can rely on transition ratings as primary measures of outcome. Use of transition ratings as primary outcomes may be particularly important if traditional questionnaires are likely to be too coarse to detect small but important changes. Second, transition ratings can provide an independent measure for construct validation of questionnaires designed to measure change over time. Third, transition ratings provide
Acknowledgments
We would also like to thank Holger Shünemann for a very helpful review of an earlier draft of the paper. This work was supported in part by a grant from the Medical Research Council of Canada.
References (29)
- et al.
Measurement of health statusascertaining the meaning of a change in quality-of-life questionnaire score
Control Clin Trials.
(1989) - et al.
Determining a minimal important change in a disease-specific quality of life questionnaire
J Clin Epidemiol
(1994) - et al.
Interpretation of rhinoconjunctivitis quality of life questionnaire data
J Allergy Clin Immunol
(1996) - et al.
Measuring functional status in chronic lung diseaseconclusions from a randomized control trial
Respir Med
(1989) - et al.
Interpretation of rhinoconjunctivitis quality of life questionnaire data
J Allergy Clin Immunol
(1996) - et al.
Randomized controlled trial of respiratory rehabilitation
Lancet
(1994) - et al.
Methodological problems in retrospective computation of responsiveness to changethe lessons of Cronbach
J Clin Epidemiol
(1997) - et al.
The neasurement of dyspnea. Contents, interobserver agreement, and physiologic correlates of two new clinical indexes
Chest
(1984) - et al.
Determining optimal therapy—randomized trials in individual patients
N Engl J Med
(1986) Relation of implicit theories to the construction of personal histories
Psychol Rev
(1989)