Original ArticleA minimal clinically important difference was derived for the Roland-Morris Disability Questionnaire for low back pain
Introduction
Self-reported generic and disease-specific health profile instruments are commonly used in clinical trials and in longitudinal observational studies. However, a statistically significant change in group scores on a health profile tool does not necessarily equate to a change in an individual's score that either the patient or the clinician would identify as being an important change in the patient's health.
Methods for establishing the smallest meaningful or important change for the individual on health profile instruments have recently been reviewed [1]. This minimal important difference (MID) is derived based on one of two approaches. MIDs from anchor-based methods (MID-A) compare changes in scores on the instrument with an anchor. The anchor is often a single question on global health status change—whether the patient believes they are better than at baseline. This anchor has been suggested to be the best for measuring change from the patient's perspective but may be affected by recall bias [1]. The anchor-based approach has several other limitations [1], [2]. Anchors may come from different perspectives (patient, clinician, family) and may, therefore, lead to different MID-As. It relies on the validity and reliability of the anchor chosen. Finally, it does not take into account the variation within the sample nor the measurement precision of the instrument. This last limitation implies that the minimal difference obtained may not lie outside the range of the instrument's random variation, that is, may not exceed the measurement error of the instrument.
Due to the above limitations of anchor-based methods, distribution-based methods (MID-D) are seen as preferable. Distribution-based methods evaluate the minimal difference that is in excess of that expected by either the random variation of the sample or the measurement error of the instrument. Methods based on sample variation include the effect size [based on the baseline standard deviation (SD) of the sample], standardized response mean (SD of change), and the responsiveness statistic (SD of change of stable patients). The derived MID-D will vary across samples if the variability within each sample differs between samples [1], [3]. Therefore, MID-Ds based on the measurement error of the instrument are recommended and these are described later.
The terms MID and minimal clinically important difference (MCID) tend to be used interchangeably. However, it may be difficult for a clinician to interpret what a change of the magnitude of the MID-D means in terms of clinical significance. To qualify as a MCID, the MID-D needs to be shown to relate to associated measures; for example, in back pain sufferers, these might include severity of pain or changes in healthcare use. This implies that both anchor-based and distribution-based methods are needed to establish a MCID for an instrument.
The aim of this study was to derive a rule for determining a MCID for the low back pain-specific Roland-Morris Disability Questionnaire (RMDQ) through a series of related objectives. The objectives were: (1) to compare the MIDs for the RMDQ derived from a range of methods, (2) to establish how the MIDs obtained from the theoretically preferred distribution-based methods relate to other measures of back pain severity, (3) to use the results from objectives (1) and (2) to develop a MCID for the RMDQ by defining a rule for classifying individual patients as clinically improved or not.
Section snippets
Study population and questionnaires
The objectives were addressed using a prospective cohort study of low back pain sufferers in primary care. The study population consisted of 447 consecutive patients (Table 1) consulting primary care for back pain at five general practices in North Staffordshire, and who completed a questionnaire at baseline and at 6 months. Two hundred sixty (58%) subjects were female, with a mean age of 46.8 (SD 8.12, range 30–59). The questionnaire included the RMDQ, a 24-item back pain-specific disability
Stage 1—Comparison of methods to determine a MID for improvement
Stage 1 addresses the first objective of a comparison of MIDs obtained from commonly used anchor-based and distribution-based methods.
Stage 2—Comparison of derived MIDs with other back pain-related measures
Stage 2 shows how the MIDs derived in stage 1 compare to changes in other back pain-related measures (objective 2). Specifically, the sensitivity and specificity of these methods compared to other back pain measures obtained from the study questionnaire were calculated. By comparing these anchors (or reference standards) with the derived MIDs, identification of the most clinically relevant MID obtained from these methods is possible to obtain the minimal clinically important difference (MCID).
Stage 3—Incorporation of an anchor
Stage 3 extends stage 2 by assessing the effect of incorporating an anchor in the MID methods applied in stage 2. Several authors have recommended combining an anchor with distribution-based methods to obtain a MCID [1], [3], [17]. The methods used in stage 2 were reevaluated with the additional component that the patients had to have rated their back pain health as better at 6 months to show improvement. The inclusion of the anchor reduced the percentage improved to 13% for the RCI and to 37%
Stage 4—Derivation of a rule for determining clinical improvement for an individual on the RMDQ
Based on stages 2 and 3, a rule can be generated that classifies individuals as definitely clinically improved, possibly improved, and not improved based on the RMDQ (objective 3). This rule takes into account the MID related to measurement error of the RMDQ, allows those starting with less severe disability to show improvement by incorporating an MID dependent on the baseline score of the RMDQ, and includes the patient's own perception of change in their own health to give further evidence of
Stage 5—Application of the rule to a more severe population
The rule derived in stage 4 and shown in Box 1 was applied to a more severe and acute back pain population (n = 319, mean baseline RMDQ 13.5, SD 4.85). Participants in this study had consulted a GP at one of 28 general practices in North Staffordshire, UK for the first or second time in an episode of low back pain that was <12 weeks in duration. Further, they had not received treatment from any other health care professional for this episode of back pain. This population had smaller variation
Discussion
This study has shown that different methods can lead to different levels of MID. The type of change studied, that is, whether it relates to simply that above the measurement error of the tool or to change observed on another related measure will alter the important difference found [18], [19]. The methods used here are often used to assess the responsiveness of health status instruments, but comparisons between instruments can be affected by the methods used, with different ordering of
Conclusions
An original approach, using anchor and distribution-based methods, has derived a rule for determining the MCID on the RMDQ. The derived MCID is clinically relevant, and allows subjects from mild to severe disease to show improvement, while allowing for measurement error in the RMDQ. This rule could be used in clinical trials and longitudinal studies in low back pain as a means of detecting clinical changes in health. Further research is now needed to assess the generalizability of the rule to
Acknowledgments
This work is supported by a research grant from the Wellcome Trust. We are grateful also to the referees for their helpful comments.
References (26)
- et al.
Defining clinically meaningful change in health-related quality of life
J Clin Epidemiol
(2003) - et al.
Methodological problems in the retrospective computation of responsiveness to change: the lesson of Cronbach
J Clin Epidemiol
(1997) - et al.
Methods to explain the clinical significance of health status measures
Mayo Clin Proc
(2002) - et al.
Does questionnaire structure influence response in postal surveys?
J Clin Epidemiol
(2003) - et al.
A method of assessing change in a single subject: an alteration of the RC Index
Behav Ther
(1986) What is a “clinically meaningful” reduction in pain?
Pain
(2001)- et al.
A comparison of different indices of responsiveness
J Clin Epidemiol
(1997) - et al.
Responsiveness of functional status in low back pain: a comparison of different instruments
Pain
(1996) - et al.
Group vs individual approaches to understanding the clinical significance of differences or changes in quality of life
Mayo Clin Proc
(2002) - et al.
A study of the natural history of back pain. Part I: development of a reliable and sensitive measure of disability in low-back pain
Spine
(1983)
Functional status and disability questionnaires: what do they assess? A systematic review of back-specific outcome questionnaires
Spine
The Quebec Back Pain Disability Scale. Measurement properties
Spine
Using the Roland-Morris Questionnaire to make decisions about individual patients
Physiother Can
Cited by (198)
Effects of a mat Pilates-based exercise program for low back pain in helicopter pilots of the Brazilian Air Force: Randomized controlled trial
2024, Journal of Bodywork and Movement TherapiesHow much is enough? Considering minimally important change in youth mental health outcomes
2022, The Lancet PsychiatryMinimal important change was on the lower spectrum of previous estimates and responsiveness was sufficient for core outcomes in chronic low back pain
2022, Journal of Clinical EpidemiologyEffects of ear acupuncture combined with cupping therapy on severity and threshold of chronic back pain and physical disability: A randomized clinical trial
2022, Journal of Traditional and Complementary MedicineAssociation of Functional Disability and Biopsychosocial Factors in Older Adults With Low Back Pain Who Live in the Amazonas State Brazil: A Cross-Sectional Study
2022, Journal of Manipulative and Physiological Therapeutics