The Psychometric Properties of the Roland Morris Disability Questionnaire for Patients with Chronic Mechanical Low Back Pain

Purpose: Functional status measures are currently not widely used in South Africa to facilitate clinical decision-making or document treatment outcomes for patients with low back pain (LBP). This study investigated the internal consistency and clinical utility of a back-specific functional status measure, the Roland Morris Disability Questionnaire (RMDQ), and determined its ability to confirm the need for spinal fusion surgery. Method: A retrospective, descriptive design was used with 42 patients with chronic mechanical low back pain who consulted a private Orthopaedic surgeon in Cape Town over a one year period. All patients completed the RMDQ prior to their consultation. On completion of the medical examination, a rating for surgery was determined for each patient. The completed questionnaires were analysed using Statistical Package for the Social Sciences (SPSS). Results: The mean RMDQ score was 8.6 (N=42; median=9.0; range=2-21). Cronbach’s alpha showed a high internal consistency between items (.92). A categorical principal component analysis (CATPCA) identified two distinct dimensions in the RMDQ. Item reduction improved the internal consistency and thus the construct validity of the RMDQ. There was a low correlation between the surgeon’s rating for surgery and RMDQ scores (r=.40; P<.01). Conclusion: The RMDQ shows some good psychometric properties but some adjustments could improve it. The RMDQ cannot be used to predict the need for spinal fusion surgery.


INTRODUCTION
One way of providing evidence of best practice is through the use of appropriate outcome measures (Corr and Siddons, 2005).A variety of functional status questionnaires measuring the impact of low back pain on performance of everyday activities are available to facilitate clinical decision-making and document treatment outcomes (Bombardier, 2000).These are, however, under-utilised due to uncertainty about which instruments to use and how they should be incorporated into practice (Bombardier, 2000, Beurskens et al., 1995).
To ensure that a measurement instrument is the most suitable choice for the intended purpose, it needs to be carefully evaluated.Instruments must have established validity, reliability, sensitivity and clinical utility (Corr andSiddons, 2005, McDowell andNewell, 1996).The Roland Morris Disability Questionnaire (RMDQ) is a back-specific outcome measure that has been used in a wide variety of clinical settings (Roland and Fairbank, 2000).It has shown high internal consistency (Roland andFairbank, 2000, Kopec andEsdaile, 1995) and good correlation with other measures of physical function (Roland and Fairbank, 2000).It has not, however, been tested in South Africa.To evaluate its appropriateness for a group of South African patients, and in the absence of a 'gold standard', construct and content validity are crucial in determining whether the RMDQ measures what it is supposed to measure (Beurskens et al., 1995).
The current study investigated the content and construct validity, internal consistency and clinical utility of the RMDQ in patients with chronic mechanical low back pain (CMLBP).It also set out to determine whether it could be used to confirm the need for spinal fusion surgery.

METHOD
This retrospective, descriptive study was conducted at a private orthopaedic surgery practice in Cape Town.

Sample
The study sample (N=42) included all patients with CMLBP who consulted a single surgeon over a one year period.Patients were included if they had experienced lumbar back pain for more than 3 months that affected their function.Patients were excluded if they had any concurrent medical conditions as these could interfere with their physical function.

Instruments
The RMDQ is a 24-item self-report questionnaire.Items in the scale were chosen for their relevance and focus on physical functions likely to be affected by back pain.The phrase 'because of my back pain' was added to make each item specific to back problems.Patients complete the questionnaire by ticking the items that apply to them 'today' (see Appendix 1).Scores are obtained by adding the number of positive responses and may vary between zero (no pain and normal function) and 24 (severe pain and dysfunction) (Roland andFairbank, 2000, Roland andMorris, 1983).
The RMDQ was selected for the following reasons:  (Roland and Morris, 1983) and is useful for monitoring patients in clinical practice (Roland and Fairbank, 2000).

Procedure
All patients completed a RMDQ using the method described by Roland and Morris (1983)  the Faculty of Health Sciences Research Ethics Committee at the University of Cape Town.

Data analysis
Data were analysed using Statistical Package for the Social Sciences (SPSS).Total scores and the mean score for the RMDQ were calculated.Item totals were computed to identify items that were endorsed most frequently.Internal consistency was determined using Cronbach's coefficient alpha 1 .To reveal the inter-relationships between the items in the RMDQ, a categorical principal components analysis (CATPCA) 2 was performed.Inspection of the component loadings for each item identified the items that were more discriminative in measuring functional status.Subsequent correlations and CATPCAs were performed on these items to determine whether item selection improved the reliability of the scale.Spearman's rho (r) was used to examine the relationship between the scores obtained on the RMDQ and the surgery rating of the patients.

RESULTS
Table 1 depicts socio-demographic and medical data for the study sample (N=42).The mean RMDQ score was 8.6 (N=42; median=9.0;range=2-21).Thirty-six patients (85.7%;N=42) were included in the subsequent analyses.Six were excluded as their response patterns differed markedly from the rest.Items with less than 10 responses were excluded for further analysis, as there were insufficient responses to analyse their effects systematically (Table 2).Similarly, items endorsed by more than 33 patients (approximately 80% of the sample) were excluded.This ensured sufficient variation in responses.Altogether, 19 of the 24 RMDQ items were analysed.
Cronbach's alpha was computed with SPSS.Although alpha was high (.92) demonstrating consistency amongst the 19 items, it did not identify which items were better at measuring functional status.A CATPCA was then done.This analysis identified two dimensions in the RMDQ, suggesting that responses referred to two different aspects of back pain.The eigenvalue 3 was 7.6, which explained 40% of the total variance.The first dimension explained 28% of the variance (eigenvalue=5.3) while the second explained 12% (eigenvalue=2.3).Closer inspection of the component loadings for each item identified those that contributed highly to either dimension 1 or dimension 2. Since items with low component loadings seemed unrelated to the remaining items (further analysis had not revealed further clustering of the items), a subsequent CATPCA was performed in which only the items that explicitly measured the phenomenon were included (N=9).Analysis of the selected nine items showed an improved eigenvalue (4.7), and explained 59% of the total variance.Figure 1 illustrates the interrelationship between the selected items which are considered to have made the most contribution to measuring functional status in the study sample.The relationship between items is illustrated by the angle between the vectors: the smaller the angle, the higher the relationship between them, and vice versa.An angle of 90º implies that the items are unrelated.Items formed two clustersthose between A8 and A7, and those between A9 and A6.Examination of these items showed that the former clus-  ter seemed to be related to strategies to manage pain, while the latter appeared to relate to the impact of pain on function.The selection of items improved the consistency between the selected items and thus the construct validity of the adapted scale.
The relationship between the total RMDQ scores and the rating for surgery was low (r=.40;P<.01).Although this is statistically significant, the correlation only explained 16% of the variance.It was therefore not opportune to execute further analyses to examine the predictive power of the RMDQ in discriminating between patients who were thought to need surgery and those who did not.

DISCUSSION
The mean RMDQ score in this study was lower than that reported by Roland and Morris with patients seen in primary care (1983), and Lee et al (2001) with CMLBP patients at an out-patients clinic.This finding was not anticipated considering that the patients in the current study had CMLBP and were expected to have high RMDQ scores.The mean score did, however, fall within the range for chronic LBP defined by Stratford and Binkley (1999).
As the study was explorative in nature, the overall pattern of responses, rather than individual item responses, was of interest.In dichotomous scales such as the RMDQ, items where one alternative has a very high (or very low) endorsement rate are usually eliminated, as they do not improve the psychometric properties of the scale (Streiner and Norman, 1995).This prevents the occurrence of floor 4 and ceiling 5 effects which are problematic if the scale is used to evaluate change over time.
Some floor and ceiling effects have been described for the RMDQ (Atlas et al., 2003, Beurskens et al., 1995, Bombardier et al., 2001).As a result, several researchers have suggested that the RMDQ may be more sensitive than the Oswestry Disability Questionnaire (Fairbank et al., 1980) (Kopec et al., 1995).This may be due to the item selection of the RMDQ being superior to competing measures (Stratford and Binkley, 1997).If this is true, then introducing a multilevel scoring scheme may enhance the properties of the RMDQ.
Previous studies identified RMDQ items that were less discriminative in measuring functional status.For example, Stratford and Binkley (1997) discovered six items that could be deleted from the scale, while Atlas et al (2003) deleted five items, replacing them with four new ones.Four of the items (15, 19, 20 and 24) identified by Stratford and Binkley (1997) and three (15, 19 and 20) by Atlas et al (2003) were also eliminated in the current study.Removing the less discriminative RMDQ items reduced the level of statistical 'noise' as evidenced by the improved construct validity of the scale.Removal of items can also, however, compromise the validity and reliability of an instrument (Kopec and Esdaile, 1995).Atlas et al (2003) found that reducing the number of RMDQ items increased the floor and ceiling effect rendering it less effective in distinguishing between patients with differing pain severities at a point in time and in assessing change over time.
Problems related to the design of the RMDQ, which could have affected the scoring and thus the study results included: • Items are checked if they apply 'today', but there is no mechanism for identifying whether an item has been mistakenly or purposely omitted.The scoring system also does not permit a non-applicable response.• Some patients wrote words (such as 'sometimes') rather than checking an item, suggesting that they may not have completed the scale based on their back pain 'today'.• Statements such as 'I only walk short distances' may not have been completed accurately due to the absence of a quantitative description of what is meant by 'short'.These deficiencies may result in inconsistent responses between patients and have been encountered previously (Lee et al., 2001, Stratford and Binkley, 1997, Turner et al., 2003).A further weakness of the RMDQ is its limited range (Lee et al., 2001, Roland andFairbank, 2000).It should thus be used with other measures where necessary to obtain a comprehensive view of the impact of LBP on the person's life.The study supported Roland and Fairbank's (2000) assertion that the RMDQ is practical to administer and score which is a strength for its content validity.
The high internal consistency among the RMDQ items compares favourably with other studies (Kopec and Esdaile, 1995, Kopec et al., 1995, Jarvikoski et al., 1995, Stratford et al., 2000) and suggests that all items analysed were related and contributed to measuring functional status.According to Streiner and Norman (1995) Cronbach's alpha should not be higher than .90.The high score could therefore, indicate a high level of item redundancy.
The identification of two dimensions in the RMDQ is a unique finding.No other studies were located in which a CATPCA has been conducted on the RMDQ although factor analysis has been used to develop other functional status measures for LBP (Delitto, 1994, Stratford andBinkley, 1997).The two dimensions identified, measure different aspects of the construct of 'function'.This finding could thus be used to address criticisms about the conceptual basis of the RMDQ (Delitto, 1994, Kopec and Esdaile, 1995, Kopec et al., 1995).
The low correlation between the RMDQ and the rating for surgery suggests that the RMDQ cannot be used to determine the need for spinal fusion surgery in patients with CMLBP.This indicates that the severity of functional impairment does not necessarily predict the need for surgery.4 Floor effect: the patient scores at the low end of the scale making it difficult to detect deterioration in functional status.5 Ceiling effect: if a patient scores at the top end of the scale it becomes difficult to measure an improvement in functional status (Streiner and Norman, 1995).

Yes No
9. I get dressed more slowly than usual because of my back.
10.I only stand for short periods of time because of my back.
11.Because of my back, I try not to bend or kneel down.
12. I find it difficult to get out of a chair because of my back.
13. My back is painful almost all the time.
14.I find it difficult to turn over in bed because of my back.
15. My appetite is not very good because of my back pain.
16.I have trouble putting on my socks (or stockings) because of the pain in my back.
17.I only walk short distances because of my back.
18.I sleep less well on my back.
19.Because of my back pain, I get dressed with help from someone else.
20.I sit down for most of the day because of my back.
21.I avoid heavy jobs around the house because of my back.
22.Because of my back pain, I am more irritable and bad tempered with people than usual.
23.Because of my back, I go upstairs more slowly than usual.
24.I stay in bed most of the time because of my back.

TOTALS SCORE _______ (total number of ticks in the 'yes' column)
The score may range from a minimum of 0 to a maximum of 24.

APPENDIX 3: RMDQ ITEMS THAT MADE THE MOST CONTRIBUTION TO THE MEASUREMENT OF FUNCTIONAL STATUS
When your back hurts, you may find it difficult to do some things you normally do.This list contains sentences that people have used to describe themselves when they have back pain.When you read them you may find that some stand out because they describe you today.Compressed air massage can be varied by using different applicator heads or changing the air pressure.The heads differ in the size, number and configuration of the outlet holes through which the air leaves the applicator head.These range from a single 5 mm diameter hole to multiple pinholes in a linear array or a single narrow slit.The applicator head with the single 5 mm hole has been shown to transmit the greatest pressure to subcutaneous tissue and is therefore more likely to cause tissue damage (Mars 2003).
A single, 10 minute treatment, at 1 Bar pressure, using the applicator head with the single 5 mm diameter hole has been shown to cause ultrastructural changes to skeletal myofibres.These include juxta-nuclear and intermyofibrillar oedema, electron-lucent spaces filled with swollen mitochondria and elements of the sarcoplasmic reticulum (SR) and occasional aggregates of glycogen and other non-contractile organelles in oedematous, sub-sarcolemmal regions, immediately after treatment.Myofibre oedema was significant, with a 17.2 % increase in mean fibre diameter after treatment.24 hr after treatment, intermyofibrillar oedema was reduced, but SR swelling remained and many fibres were characterised by focal and large areas of myofibrillar disorganisation.Myofibre oedema, while present, was reduced to 5.5 %.With the exception of occasional swollen elements of the SR and a single internalised nucleus, myofibres morphology had returned to normal 6 days after treatment (Gregory and Mars 2004).These changes were less frequent and of less severity than those noted after a 10 minute treatment using deep transverse friction (Gregory et al 2003, Gregory and Mars 2004).

RMDQ item 1 .
I stay at home most of the time because of my back.15.My appetite is not very good because of my back pain.19.Because of my back pain, I get dressed with help from someone else.20.I sit down for most of the day because of my back.24.I stay in bed most of the time because of my back.

Figure 1 :
Figure 1: Component loadings for RMDQ items after removal of items not contributing to the measurement of functional status.

Variable Categories No. of subjects % of total sample Male Female Total
* Percentages total more than 100% due to rounding off of decimal places

Table 2 : Items excluded from the analysis of the RMDQ.
1 Cronbach's alpha is 'used to express the internal consistency reliability of a test' by examining the correlation between items in a test (McDowell and Newell, 1996, p. 499). 2 CATPCA performs a non-linear Principal Component analysis.The analysis forms the items into clusters of variables (or factors) that are related to each other but measure a distinct aspect of the phenomenon (McDowell & Newell, 1996).3 Eigenvalue: the proportion of variance explained by each factor

Remember, only tick the 'yes' column if you are sure the item describes you today.
As you read the list think of yourself today.When you read a sentence that describes you today, tick the 'yes' column.If the sentence does not describe you, then tick the 'no' column.Four vervet monkeys underwent one, 15 min, treatment of compressed air massage at 1 Bar, to the tibialis anterior muscle and four animals received similar treatment to the whole lower leg on three consecutive days.The tibialis anterior of the treated and untreated limbs was biopsied immediately after the final treatment.Muscle fibre diameters were measured from 1µm thick toluidine blue stained resin embedded sections using light microscopy and computerized image analysis software.
Conclusions: Repeated treatment causes skeletal muscle oedema, and this appears to be dose related.Skeletal muscle oedema after three treatments is less than after a single treatment.Further studies on the use of compressed air massage on injured muscle are warranted.KEY WORDS: COMPRESSED AIR MASSAGE, SKELETAL MYOFIBRES, LIGHT MICROSCOPY, MORPHOMETRY.