A new approach to measuring recovery in injured workers with acute low back pain: Resumption of Activities of Daily Living Scale.

Key Words: Back injuries, Recovery, Rehabilitation.

Soft tissue injuries, particularly those resulting in low back pain (LBP), are the largest category of work-related compensation claims.[1] Early referral and aggressive intervention using exercise conditioning regimens have been advocated for treating people with soft tissue injuries.[2,3] In 1989, the Ontario (Canada) Workers' Compensation Board (WCB) implemented a Community Clinic Program "to enable workers to recover faster and more completely." Within a few years, over 100 Community Clinic Programs, operated by physical therapists and chiropractors, were accredited by the WCB across the province of Ontario. To be eligible to participate in a Community Clinic Program, injured workers must have sustained a soft tissue injury, qualify for claim status (although they could be working at the time), and be referred preferably no later than 70 days after the date of injury or recurrence. A similar approach to treating people with soft tissue injuries is being used by other provinces. For instance, Ehrmann-Feldman et al[3] studied Quebec's WCB files and found that exercise and back education were the most frequently selected interventions. They concluded that early referral for physical therapy (within 30 days after injury) was associated with earlier return to work. Battie, however, cautioned that such findings do not answer the question of "whether early referral for physical therapy is beneficial in achieving better patient outcomes"[4(p157)] or which physical therapy interventions are most efficacious.[4]

Recently, the Ontario WCB has questioned its own costly approach to early intervention.[5] In a prospective study conducted by Sinclair et al,[5] injured workers were followed for 1 year and the researchers found no advantages with respect to absence from work or other health-related advantages for clinic attendees over nonattendces. Consequently, the Ontario WCB has changed its policy, and current Community Clinic Program eligibility is restricted to workers who sustained their injury at least 4 weeks prior to program entry.[5] This policy is in line with practice guidelines for treating people with acute low back problems that suggest that, in the absence of "red flags," clinicians should intervene as little as possible in the first few weeks after symptoms begin.[6] A clinical concern with such a policy is that delayed intervention may lead to the development of pain-related behaviors and chronicity.[7] Ethnographic interviews with injured workers revealed similar concerns that back problems may be lifelong.[8] Yet, studies of the natural course of back pain suggest a very optimistic prognosis for most patients.[9,10] Within 2 weeks of an acute episode of back pain, nearly 50% of persons can be expected to spontaneously recover, 70% can be expected to recover within 1 month, and 90% can be expected to recover within 3 months.[10]

Spontaneous recovery from the time of injury to the start of treatment, and concurrent with the course of treatment itself, is a critical consideration both in evaluating the effectiveness of rehabilitation programs and in conducting clinical research on the efficacy of various treatment modalities. As noted, return to work is perhaps the most common indicator of "recovery" used with injured workers. Return to full-time work has been regarded as "proof that the patient had made a full and complete recovery."[2(p515)] Spontaneous recovery is only one of many factors that render return to work a highly problematic outcome measure. Return to work is influenced by several other factors, including local economic conditions, job satisfaction, job demands, employer attitudes, and worker characteristics and resources.[11-15] Deyo and colleagues[14] cautioned against using single, global indicators (eg, return to work, ratings of improvement) as outcome measures for patients with LBP.[14] They argued that, when rating scales that consist of multiple aspects of outcome are combined, important changes in one dimension may be obscured or it may be impossible to tell which dimensions account for improvements or benefits. Furthermore, perceptions by patients with LBP and their therapists may not agree. In one study,[16] 86% of injured workers felt they could not work, whereas their clinicians thought only 49% of these persons could not work. Nevertheless, determinations of return-to-work status for injured workers being discharged from rehabilitation programs continue to be the predominant indicator regarding compensation claims and case closures, as well as the primary dependent variable for research in the field.

The Institute for Work & Health funded our study to examine the use of standardized performance indicators of patient outcomes for Community Clinic Programs. The WCB, much like other health care agencies, was moving from a quality assurance to a continuous quality-improvement approach to accrediting the community clinics.[17] The only requirements for accreditation purposes at the time were documented user rates, aggregate diagnostic and demographic characteristics of clinic users, and a return-to-work recommendation at clinic discharge according to the following categories: (1) return to unrestricted work, (2) return to restricted work, (3) referral for further assessment and counseling if recovery was prolonged, and (4) other. We found that most Community Clinic Program therapists assessed each participant's initial level of physical conditioning (primarily through tests of muscle strength, range of motion, and endurance), although with little consistency across clinics in techniques, recording methods, or repeated testing. In some clinics, therapists also administered various standardized patient questionnaires (eg, Roland-Morris Disability Questionnaire,[19] Oswestry Low Back Pain Disability Questionnaire[20]). As some authors 14,18 have noted, current practice involves limited use of outcome measures, and none of these measures have become the standard for the field.

We set out to explore the issue of "recovery," with injured workers with LBP and their therapists, in order to determine whether existing measures adequately defined this construct. Our ultimate aim was to recommend a standardized performance indicator of "recovery" for use by all Community Clinic Programs in the province.

Method

Scale Development Process

Similar to Tarasuk and Eakin,[8] we started with a qualitative, inductive approach to exploring the experiences of both first-time and previously injured workers with LBP. We also explored clinicians' experiences of the recovery process and their impressions of available tools. Separate focus groups were conducted at a Community Clinic Program with 8 persons with LBP (3 women and 5 men) and with 6 clinicians (3 physical therapists, 2 occupational therapists, and 1 kinesiologist) employed by the clinic.

The people with LBP said that, for them, "recovery" meant "getting back to the way I was before the injury" or "getting back to norm A." When asked what activities had been affected by their injury, they cited a diversity of examples, which included but was not restricted to paid employment. For instance, several people with LBP mentioned disrupted sleeping patterns, some mentioned not being able to look after or play with their children, others talked about jobs around the house, and still others talked about sports they enjoyed or the effect on their sex lives. As Tarasuk and Eakin[8] found, perceived limitations were often phrased in terms of "I can only do half as much ... as I did before" or "I'm not sure whether I'm going to be 100% again."

Clinicians discussed differences among persons entering the program in terms of level of physical conditioning and expectations. They noted that for some individuals, "reemployment may not be a motivating factor; they may hate their jobs, but they love to 10-pin bowl." Again, the themes of preinjury comparisons and individualized consequences emerged.

Subsequent to this general discussion, each group was shown several published measurement tools--the Roland-Morris Disability Questionnaire,[19] the Oswestry Low Back Pain Disability Questionnaire,[20] the Pain Self-Efficacy Questionnaire (PSEQ),[21] and a visual analog scale (VAS).[22] Persons with LBP were asked to complete each measure before their impressions were solicited. Both groups commented that a focus on "pain" is at odds with the philosophy of the Community Clinic Program, and they disliked the implications of permanence or chronicity in the qualifiers "because of my back" (Roland-Morris scale), "despite the pain" (PSEQ), and "pain prevents me from ..." (Oswestry scale). Persons with LBP found many of the Oswestry scale choices confusing and took issue with items such as "My sex life is normal ..." and "My social life is normal frequently commenting "What's normal?"

Although they are called "disability" scales, it has been noted in the literature that both the Roland-Morris scale and the Oswestry scale contain elements of impairment, disability, and handicap according to the World Health Organization's classification system.[23] Apart from the qualifiers "because of my back" and "despite my back/back pain," both the content and the rating formats (unable/able to do, level of difficulty, need for assistance) of the Roland-Morris and Oswestry scales and the more recent Quebec Back Pain Disability Scale[24-26] are very similar to those of general activities-of-daily-living scales. Ratings of difficulty, nonperformance, and reported assistance are not interchangeable and should not be assumed to reflect underlying physical ability.[27,28] Such ratings are influenced by division of labor (eg, concerning household chores), personal preferences, and individual differences in physical ability.[27] Injured workers may not have been able to do, may have had difficulty doing, or may have chosen not to do certain activities before the injury. Resumption of "customary" or "usual" activities seems to us to be a more sensible approach to quantifying recovery.

Accordingly, the Resumption of Activities of Daily Living (RADL) Scale was developed with item content selected in conjunction with both clinicians and patients from the Community Clinic Programs to measure broad areas often affected by back injury and with a standardized, yet individualized, rating format in which the patient served as his or her own benchmark. For instance, a person's "usual" sleeping patterns may be good or poor, a person's chosen recreational pursuits could be minimally demanding or physically challenging, and so on. The extent of resumption of a person's "usual" patterns--whatever they might be--represents the indicator of recovery. Following pilot testing with 2 different clinic samples, the final 12-item RADL (taking about 5 minutes to complete) was subjected to psychometric testing. The final RADL is shown in the Appendix. The total RADL score has a possible range of 0 to 100, summing across the items and dividing by the number of items. We suggest that a minimum of 70% of the items (about 9 out of the 12 items) be completed to calculate a total score for each person.

Sample Recruitment

Eligibility criteria for subject recruitment were the same as for entry into the Community Clinic Programs. At the time of the study, the original criteria (eg, preferably no more than 70 days had elapsed since injury or recurrence of symptoms) were in place. Injured workers need to qualify for claim status and may be working at the time of entry into these programs. We restricted our sample to persons with soft tissue injuries of the back.

After entry into the Community Clinic Programs, the participants received a routine intake physical assessment performed by a physical therapist. After completing this assessment, the therapist asked the participants whether they would be willing to participate in a study examining various assessment tools for injured workers with back pain. Ninety percent of the persons who were approached in both the reliability and validity phases participated. Interested persons were then seen by a researcher to obtain informed consent and complete the various study questionnaires.

Reliability Phase

A group of 20 persons with LBP who attended I of 4 different Community Clinic Programs participated in the reliability testing phase. The RADL was administered immediately after the clinic intake assessment (time 1) and prior to treatment (time 2). The mean interval between assessments was 1.8 days (SD = 1.2, range= 1-5). To examine whether the person's condition had changed from time 1 to time 2, each subject was asked: "Is your back pain/condition the same as it was when you completed these questionnaires the other day?"

Validity Phase

One hundred four workers with LBP who attended 1 of 7 different Community Clinic Programs participated in this phase, which involved administration of a battery of questionnaires (taking about 30 minutes in total) at 2 points: at clinic entry and at discharge from the clinic or 3 weeks after clinic entry (whichever came first). Three weeks was chosen to minimize loss to follow-up and to assess short-term change. At each time point, participants were asked to complete (in random order) the following questionnaires: the RADL, the 24-item Roland-Morris Disability Questionnaire,[19] the 7-item Marlowe-Crowne Scale (MCS),[29] and the 15-item Functional Abilities Confidence Scale (FACS) (see companion article by Williams and Myers in this issue for a detailed description and discussion of the FACS).

The Roland-Morris scale and the FACS were used to examine the convergence of the RADL scores. Both scales were developed for the assessment of persons with LBP. The Roland-Morris scale consists of 24 dichotomous items concerning pain, perceived limitations, emotions, and assistance received ("because of my back").[19] The FACS, which we developed concurrently with the RADL, examines self-confidence in performing various movements and postures such as walking, standing, sitting, and climbing up and down stairs. We expected to find moderate correlations between the RADL and both the Roland-Morris scale and the FACS at the initial (baseline) assessment. We expected to find stronger relationships at the follow-up assessment. Improvements on these 3 self-rated measures (ie, increased resumption of customary activities via the RADL, reduced disability via the Roland-Morris scale, and increased self-efficacy via the FACS) were anticipated after 3 weeks of rehabilitation.

The MCS was used to assess social desirability or a tendency for people to pre-sent themselves in a socially desirable manner to achieve the approval of other people. We expected to find a low correlation between the MCS and the RADL at both baseline and follow-up assessments. We did not expect the MCS scores to change over the 3-week program.

A number of single-item predictions and global ratings also were solicited. During the initial assessment, subjects were asked: "How long do you think it will be before you are physically able to return to work on a full-time basis?" (ie, less than 1 week, 1-2 weeks, 2-3 weeks, 3-4 weeks, more than 1 month, more than 3 months, or more than 6 months). During the follow-up assessment, subjects were asked: "How confident are you that you are physically able to return to full-time work?" (rated from 0% = not at all confident to 100% = completely confident) and "Do you think your back condition has changed since the start of the program?" (yes or no).

Clinicians were asked to predict during the initial assessment how long they thought it would be before each subject would be able to return to work on a full-time basis. They also were asked: "How confident are you that this person will be able to improve to his or her preinjury level thorough participation in the program?" (rated from 0% = no confidence at all to 100% = completely confident). During the follow-up assessment, they gave their impressions of whether the subjects' back condition had changed since the start of the program, using the same rating format used by the subjects. The follow-up return-to-work recommendation by the clinician for each subject consisted of the standard 4-category Community Clinic Program format described previously.

In an attempt to standardize clinicians' ratings of the subjects' physical conditioning, a 5-item rating scale was developed as a proxy measure for various assessment tools used at the different clinics. For each subject, the treating clinician was asked to rate the subject's level of physical conditioning with respect to (1) endurance, (2) muscle strength, (3) range of motion, (4) locomotion, and (5) overall ability to perform functional activities on the same 0% to 100% scale during the initial assessment and again during the follow-up assessment. Pilot testing of this scale suggested that clinicians were basing these judgments on both physical tests and clinical observations and experience. Clinicians were not aware of any of the subjects' self-ratings. Because there appears to be a discrepancy between clinicians' and patients' ratings,[16] we did not expect to find a strong relationship between subjects' ratings on the RADL and clinicians' ratings of the subjects' physical conditioning abilities. We expected that clinicians may be more likely to rate individuals as "improved" and able to return to work at discharge compared with the subjects' own ratings.

The discriminant abilities of the RADL were examined by comparing the RADL scores of the total sample as well as the scores between various subgroups (eg, subjects with a previous injury versus subjects without a previous injury, new versus previous program attendees, program completers versus noncompleters, subjects judged by clinicians as able to return to work versus subjects judged as unable to return to work). We expected that subjects who were working while starting the Community Clinic Programs would have higher RADL scores than subjects who were not working. We expected that subjects who completed the program and those who were judged by clinicians as able to return to work would have higher RADL scores compared with their counterparts.

Data Analysis

The Shapiro-Wilks statistic[30] and stem and leaf plots were used to examine the normality of the RADL scores. Intraclass correlation coefficients (ICC[2,1])[31] derived by analysis of variance were used to estimate test-retest reliability. Cronbach's alpha, inter-item and item-correlation coefficients, and factor analysis were used to examine internal consistency, homogeneity, and structure, respectively. If a scale score is internally consistent, each item score should correlate with scores on all other items.[32] Correspondence between subjects' and clinicians' ratings was determined via Kappa and Pearson correlation coefficients for categorical and continuous ratings, respectively. Pearson correlation coefficients also were used to compare total scores on the various scales. We used t tests to compare mean RADL scores for various subject groups (eg, subjects who were working versus subjects who were not working). Paired I tests were used to compare entry and follow-up scores, and effect size was used to estimate the magnitude of change for the sample as a whole. Logistic regression analysis was used to examine whether the baseline RADL scores would predict return to work. As suggested by Lord and colleagues,[33] we also calculated the percentage of improvement for each individual in the sample, as follows: ([follow-up score -- baseline score]/baseline score) X 100.

Results

To assess reliability, the RADL was administered twice (1-5 days apart), before treatment commenced, to a separate sample of 20 subjects (12 women and 8 men; average age = 39 years, range = 19-56; average time since injury = 5 weeks, range = 1-24 weeks). All subjects confirmed that their back condition had not changed during the interim. The ICC was .83 (95% confidence interval = .62-.93).

Of the 104 subjects who participated in the validity testing phase, 94 subjects (69 men and 25 women) completed both initial and follow-up assessments. Most of the dropouts either did not attend the program past the initial assessment (n = 4) or did not attend the program beyond one session (n = 3). The average age of the subjects in the validation sample was 37 years (SD = 11, range = 19-64). About half of the validation sample (57%, n = 54) reported having a previous back injury (6 subjects had undergone back surgery); and 29% (n = 27) had attended a similar program previously. The time from onset of injury to program entry ([bar]X = 6 weeks, SD = 11, range = 1-52) was less than 2 weeks for the majority of our sample (61%). Only 8 subjects had been injured more than 10 weeks prior to clinic entry. Twelve subjects were working at the time they were initially seen in the clinic. The average number of sessions attended over the 3-week period was 14 (SD = 3, range = 4-21). The aver-age length of time the participants were in a Community Clinic Program was 22 days (SD = 3, range = 8-30), with 32 subjects completing the program over the study period. Data for subjects who had missing values on the RADL items were dropped from the analysis. Items that were marked "N/A" (not applicable) also were dropped from the analysis, and a total score was computed based on the number of items answered.

The Shapiro-Wilks test showed that RADL scores were not normally distributed (P [is less than] .001). Because the Shapiro-Wilks test is very conservative, we chose to treat the scores as normally distributed based on the mean and median values being similar and a full range of scores for the RADL. The stem and leaf plots indicated that scores were not severely skewed. Given that the sample mean (44.6) and the median (40) were similar and good score distribution emerged, we believe the use of parametric statistics was justified.[34,35]

Internal Consistency

Cronbach's alpha (.89) indicated that the RADL has high internal consistency. Both inter-item correlations (.20-.74) and item-total correlations (.43-.82) were fairly high, further suggesting scale homogeneity and that all items should he retained (as does the finding of minimal change in alpha values with each item sequentially deleted).[34]

Principal-components factor analysis using varimax rotation revealed three factors for the RADL, accounting for 68% of the total variance (28%, 23%, and 17% for factors 1, 2, and 3, respectively). Factors should be considered "pure" when items loaded highly only onto one factor and close to zero on the other factors.[34] As shown in Table 1, several of the RADL items loaded moderately (.26-.52) on more than one factor, suggesting they may be measuring different aspects of the same construct. The resulting factor structure of the RADL indicates that the less physically taxing activities loaded more highly onto the first factor, whereas more demanding activities loaded onto either the second or third factor. Although subscales could be created, the mixed factor loadings, together with the advantage of representing a continuum of physical demands, argue in favor of retaining a single, total scale score.

Table 1.
Factor Analysis of the Resumption of Activities of Daily Living
Scale(a)

                                       Factor   Factor    Factor
Activity                                  1        2         3

Sleeping                                .69      .26       -.03
Sexual activities                       .53      .43        .09
Self-care                               .78      .08       -.04
Light household chores                  .52      .61        .16
Heavy household chores                  .06      .83        .28
Shopping                                .39      .70        .18
Socializing inside home                 .78      .09        .32
Socializing outside home                .67      .40        .42
Traveling for less than 30 minutes(b)   .65     -.03        .62
Traveling for more than 1 hour          .20      .30        .82
Recreational activities                 .12      .77        .20
Engaging in paid employment            -.07      .39        .71

(a) Items with loadings above .5 are boldfaced.

(b) Underlined loadings indicate item was factorially complex (ie, showed very similar loadings on factors 1 and 3).

Discriminative Abilities

Baseline and follow-up RADL scores are shown in Table 2 for the sample as a whole as well as for various subgroups. Between-group comparisons revealed that neither entry nor follow-up RADL scores differed for subjects with and without previous back injuries or for new and previous clinic attendees, suggesting that the RADL may not be influenced by patients' expectations of rehabilitation. There was, however, a difference in baseline RADL scores (t = 2.45, df = 56, P [is less than] .05) for 12 subjects who were working at the time of clinic entry (X = 61, SD = 20, range = 24-80) as compared with 80 subjects who were not working at the time of clinic entry (X = 41, SD = 18, range = 3-87). Information on working status was missing for 2 subjects. Similarly, the 10 dropouts (persons who did not attend beyond the first session) had higher RADL scores (X = 60, SD = 17, range = 33-80) than did the 94 subjects who continued in the program (t = 2.15, df = 94, P [is less than] .05). These 10 individuals also were younger (mean age = 27 years, SD = 5; t = 4.90, df = 94, P [is less than] .0001) and more likely to be working at the time they were first seen in the clinic (40% compared with 13% for the program continuers, Fisher's Exact Test, P [is less than] .05).

Subjects who completed the program (left or were discharged) by 3 weeks (n = 32) had entry and follow-up RADL scores similar to those of subjects who did not complete the program in 3 weeks (n = 61). Data were missing for one subject. Subjects judged by the clinicians as able to return to work at 3 weeks had both higher baseline (t = 2.07, df = 83, P [is less than] .05) and follow-up (t=3.23, df=80, P [is less than] .001) RADL scores than persons judged as not able to return to work (Tab. 2).

Table 2. Resumption of Activities of Daily Living Scale Scores at Baseline and Follow-up for Entire Sample and Subgroups

                                  Baseline Score

                              [bar] X    SD    Range
All subjects(a)                 44.6    21.1    3-88
Previous injury(a)              46.4    22.4    3-88
No previous injury(a)           42.4    19.6    9-86
Previous attendees(b)           43.6    19.0    3-80
New attendees(a)                45.0    22.1    9-88
Completers(c)                   42.0    21.0   13-84
Noncompleters(c)                46.1    21.5    3-88
Able to return to work(c)       51.4    22.4   21-88
Unable to return to work(d)     41.7    19.7    3-87

                                  Follow-up Score

                              [bar] X    SD     Range   df
All subjects(a)                 60.8    23.3    8-100   83
Previous injury(a)              58.9    23.0    8-100   45
No previous injury(a)           64.0    23.9   17-96    33
Previous attendees(b)           55.4    16.2   23-100   23
New attendees(a)                62-9    25.4    8-100   55
Completers(c)                   62.0    24.0   17-98    27
Noncompleters(c)                60.0    24.0    8-100   54
Able to return to work(c)       72.0    24.2   17-100   28
Unable to return to work(d)     55.5    20.0   23-97    52

(a) Significance of difference from baseline to follow-up, P <.0001, paired t tests.

(b) Significance of difference from baseline to follow-up, P <.05, paired t tests.

(c) Significance of difference from baseline to follow-up, P <.001, paired t tests.

(d) Significance of difference from baseline to follow-up, P <.01, paired t tests.

Responsiveness to Change

Good sample variability was evident for the 12 items of the RADL. As we expected, at the time the subjects were first seen in the clinic, they were more likely to have resumed the more basic, less taxing activities. For instance, average baseline ratings were highest for the self-care item ([bar] X=75, SD=28) and lowest for resumption of paid employment ([bar] X=12, SD=29), engaging in usual recreational activities ([bar] X=23, SD=27), and heavy household chores ([bar] X=23, SD=27). Average baseline ratings on the other scale items ranged from 31 (traveling for longer than 1 hour) to 65 (socializing in the home).

Table 2 shows that there were changes (paired t tests) in RADL scores from the baseline assessment to the follow-up assessment for the sample as a whole as well as for the various subgroups. Previous clinic attendees showed less improvement than new clinic attendees. The subjects who the clinicians judged as unable to return to work also showed less improvement on RADL scores over the 3 weeks than did the subjects who the clinicians judged as able to return to work.

Effect size for the RADL (.77) was positive in direction and moderate to large in magnitude.[36] A change of 16 units on the 100-point RADL represents a clinically important difference (calculated by multiplying the effect size by the standard deviation of the sample's baseline scores).[37] Scores also improved on 2 of the other standardized scales: the FACS (P [is less than] .0001) and the Roland-Morris Disability Questionnaire (P [is less than] .0001). The effect size for the Roland-Morris scale was .85, and a change of 4 units on the 24-point scale depicted clinically significant change.

Although effect size is a measure of the magnitude of change for the sample as a whole, we also calculated the percentage of improvement for each individual. Calculation of individual change[33] showed that 24% of the subjects had no improvement (0% or a negative change) from the baseline assessment to the follow-up assessment, 7% showed slight improvement ([is less than] 10% improvement), and 69% improved by at least 10% (about half of these subjects improved by more than 50%). One subject improved by more than 300% (baseline score=17, discharge score=76). Another way of interpreting the data is to examine the proportion of individuals who scored 80 or above on the RADL (indicating almost complete or complete resumption of customary activities). Only 6% of the subjects had RADL scores of 80 or above at the initial assessment; this proportion increased to 25% and the follow-up assessment. Fourteen percent of the subjects who the clinicians judged as unable to return to work had RADL scores of 80 or better. Forty-five percent of the subjects who the clinicians judged as able to return to work had RADL scores of 80 or better.

Convergent and Discriminant Validity

Correlations between ratings on the RADL and ratings on the other measures are shown in Table 3. Correlations were moderate to high (r=.44-.82) between subjects' ratings on the RADL and their ratings on both the Roland-Morris scale and the FACS; the relationships were stronger at the follow-up assessment. There was a slight inverse correlation between subjects' ratings on the RADL and their ratings on the MCS, but only at the initial assessment.

Table 3. Correlation Coefficients of Resumption of Activities of Daily Living Scale Scores With Subjects' Scores on Other Measures and Clinicians' Ratings

                                          Baseline        Critical
                                          Score      df   Value

Subjects' ratings
  Roland-Morris Disability
   Questionnaire[27]                      -.62(a)    86   .217
  Marlowe-Crowne Scale[25]                -.21(b)    86   .217
  Functional Abilities Confidence Scale    .44(a)    86   .217

Clinicians' ratings
  Endurance                                .50(a)    85   .217
  Muscle strength                          .34(c)    85   .217
  Range of motion                          .43(a)    85   .217
  Locomotion                               .51(a)    85   .217
  Overall ability                          .46(a)    85   .217

                                          Follow-up        Critical
                                          Score       df   Value

Subjects' ratings
  Roland-Morris Disability
   Questionnaire[27]                      -.82(a)     83   .217
  Marlowe-Crowne Scale[25]                -.13        83   .217
  Functional Abilities Confidence Scale    .76(a)     83   .217

Clinicians' ratings
  Endurance                                .44(a)     82   .217
  Muscle strength                          .34(a)     82   .217
  Range of motion                          .39(a)     82   .217
  Locomotion                               .45(a)     82   .217
  Overall ability                          .37(c)     82   .217

(a) Significance of Pearson correlation coefficient, P <.0001.

(b) Significance of Pearson correlation coefficient, P <.05.

(c) Significance of Pearson correlation coefficient, P <.001.

Moderate correlations (r=.34-.51) emerged between subjects' RADL scores and clinicians' ratings of various aspects of the subjects' level of physical conditioning to perform functional activities at both the initial and follow-up assessment (Tab. 3). At the follow-up assessment, clinicians' ratings of each subject's level of physical conditioning were moderately correlated with both the subjects' own ratings of improvement (r=.36-.42, P [is less than] .001) and the subjects' ratings of their confidence in being physically able to return to work (r=.30-.46, P [is less than] .0001). Subjects' ratings of their confidence in being physically able to return to work was more strongly correlated (r=.78, P [is less than] .0001) with their own follow-up RADL scores.

When the subjects were first seen at the clinic, each subject and his or her clinician independently predicted how long it would take for the individual to be able to return to work on a full-time basis (less than 1 week to more than 6 months). Moderate agreement (weighted Kappa =.44) was found between subjects' and clinicians' predictions. Although 35% of the subjects thought they would be able to return to full-time work within 3 weeks, clinicians' estimations were much higher for this group of subjects (49%). The subjects' RADL scores at the initial assessment were predictive of clinicians' ratings of the subjects' ability to return to work obtained at the follow-up assessment. Using logistic regression, the odds of making a return-to-work recommendation at the follow-up assessment was 1.33 (95% confidence interval=1.05-1.69) times higher for a subject who scored 100 on the RADL than for a subject who scored 0 on the RADL at the initial assessment.

As we expected, poor concordance (Kappa=.23) was found between subjects' and clinicians' global ratings of improvement. At the follow-up assessment, clinicians rated 87 of the 94 individuals as improved, whereas only 70 subjects rated themselves as improved (not necessarily the same people who the clinicians rated as improved). The RADL scores were higher at the follow-up assessment for these 70 subjects compared with the 24 subjects who rated themselves as not improved ([bar] X=65.7 versus 43.0, t=3.85, df=75, P [is less than] .001). Baseline RADL scores for these two groups were similar ([bar] X=44 versus 42), indicating that the 24 subjects who did not improve were not necessarily more severely injured at the time they were first seen in the clinic. Paired t tests showed that the 70 subjects who thought they had improved had a change in RADL scores (P [is less than] .0001), whereas the 24 subjects who thought they had not improved did not show a change.

Discussion

Deyo et al contended, "Clinical research related to the treatment of back pain would be enormously facilitated if a small number of patient-oriented questionnaires became highly used."[14(p2035S)] Return to work continues to be the predominant indicator regarding case closures and patient outcomes. Our findings substantiate Cross-man and colleagues' contention that patients' perceptions of readiness to return to work may not agree with clinicians' judgments.[16] We also found poor agreement between subjects' and clinicians' global ratings of improvement. In contrast, we found good concordance among subjects' self-ratings on the different scales and between subjects' RADL scores and their own perceptions of improvement and confidence in ability to return to work.

We found good psychometric support for the use of the new RADL scale. The scale showed high temporal stability over a short period prior to intervention for a group of 20 subjects who confirmed that their back condition had not changed. A wide range of scale scores is evidence of suitability for people with LBP. The scale appears to be internally consistent and responsive to change over a 3-week period in a rehabilitation clinic. Regarding discriminant abilities, subjects who were working at the time they were first seen in the clinic as well as those who did not continue in the program (beyond the first session) reportedly resumed more of their customary activities, which would be expected. Subjects' RADL scores obtained during the initial assessment also were able to predict clinicians' recommendations of being able to return to work after 3 weeks of rehabilitation. The RADL scores did not discriminate between persons who completed or left the program at 3 weeks and persons who did not. This finding may be due to the fact that only 32 subjects completed the program at 3 weeks. Because the average time in the program is 4 weeks,[38] the 3-week time frame (arbitrarily chosen to minimize loss to follow-up) may have been too short to assess recovery.

Only 25% of this sample achieved scores of 80 or better (with a score of 100 representing complete resumption of usual activities) on the RADL at the follow-up assessment. Subjects were first seen at the clinic from 1 to 24 weeks postinjury (average of 5 weeks). At the initial assessment, only 6% of the subjects scored 80 or above on the RADL. Although there was change for the sample as a whole, with a fairly large effect size, calculation of individual change showed that some subjects improved substantially, whereas other subjects showed no improvement (and the status of a few subjects even deteriorated).

Roland-Morris Disability Questionnaire[19] scores also were responsive to change over the 3-week period. Given that scores on the Roland-Morris scale and the RADL were highly correlated, we believe the important issue is whether the RADL has any advantages over the Roland-Morris scale or other related scales (eg, Oswestry Low Back Pain Disability Questionnaire,[20] Quebec Back Pain Disability Scale[24-26]). Investigators have been admonished for developing new measures when suitable measures already exist, often without conducting head-to-head comparisons.[14,27,39] Available disability and handicap measures for people with LBP include various items related to activities of daily living. Even their rating formats (level of difficulty, able/unable to do, amount of assistance received) are similar to those of more general activities of daily living scales. All these scales fail to consider the preinjury functioning of persons with LBP. That is, these individuals may have had difficulty doing particular activities or may have chosen not to do particular activities prior to their back injury. Level of physical conditioning, health status, and lifestyle choices all have an important bearing on a person's customary or "usual" activities (whether in the occupational or familial domains, a person's sex life, or a person's social or recreational pursuits). Using sleeping as an example, "normal" or "usual" patterns can be expected to vary tremendously across individuals.

We believe the important issue, from a rehabilitation perspective, is the degree to which a person's customary patterns have been disrupted by a back injury and the extent to which a person gradually resumes his or her customary patterns as a result of natural recovery or rehabilitation, or both. The RADL does not differ substantially from previous questionnaires developed to assess the functional status of people with LBP in terms of content. The primary difference lies in the rating format, using a person's customary level of functioning as the individualized benchmark for "recovery." A continuous item rating format from 0% to 100% is also advantageous (as compared with the dichotomous format [ie, yes or no] of the Roland-Morris scale 19) to chart incremental progress in various items. For example, on the RADL, improvement in sleeping patterns can be from 0% to 100%, whereas on the Roland-Morris scale, the subject can either turn over in bed or not. The RADL was developed using an inductive, qualitative approach involving therapists and, most importantly, the intended test takers. Such a process is more likely to produce what Feinstein et al[39] referred to as a "clinically sensible" tool.

There is still a great deal of work to be done in determining whether early referral leads to better patient outcomes and which interventions are the most efficacious and cost-effective in treating people with low back injuries.[4-14] The natural course of LBP is an important consideration, as is the injured person's level of physical conditioning, the severity of the injury, and the person's level of motivation concerning rehabilitation. It is essential to measure the extent of recovery from time of injury to start of treatment, and concurrent with the course of treatment itself, in conducting clinical trials on the efficacy of various treatment modalities and in evaluating the cost-effectiveness of a rehabilitation program. Although clinics focus on improving a person's physical condition and preventing further injuries, the ultimate goal for patients is "getting back to normal." The RADL is unique in that the individual's customary level of functioning serves as the benchmark of recovery.

Conclusions

The RADL is a promising new tool with good administrative properties, including test-retest reliability, internal consistency, responsiveness to change, and discriminative abilities. It is easy to administer (it takes only 5 minutes to complete), score, and interpret. The RADL is a standardized, yet individualized, measure of a person's perceptions of "recovery" that can be used for epidemiological studies, clinical trials, and evaluation of rehabilitation programs.

References

[1] Andersson GBJ, Fine LJ, Silverstein BA. Musculoskeletal disorders. In: Levy BS, Wegman DH, eds, Occupational Health: Recognizing and Preventing Work-Related Disease. Boston, Mass: Little, Brown & Co Inc; 1995:455-470.

[2] Mitchell RI, Carmen GM. Results of a multicenter trial using an intensive exercise program for the treatment of acute soft tissue and back injuries. Spine. 1990;15:514-521.

[3] Ehrmann-Feldman D, Rossignol M, Abenhaim L, Gobeille D. Physician referral to physical therapy in a cohort of workers compensated for low back pain. Phys Ther. 1996;76:150-156.

[4] Battie MC. Invited commentary on "Physician referral to physical therapy in a cohort of workers compensated for low back pain." Phys Ther. 1996;76:156-157.

[5] Sinclair SJ, Hogg-Johnson S, Mondloch MV, Shields SA. The effectiveness of an early active intervention program for workers with soft tissue injuries: the Early Claimant Cohort Study. Spine. 1997;22:2919-2931.

[6] Agency for Health Care Policy and Research. Acute Low Back Problems in Adults: Clinical Practice Guideline. Rockville, Md: US Department of Health and Human Services; 1994.

[7] Fordyce WE, ed. Back Pain in the Workplace: Management of Disability in Non-Specific Conditions. Seattle, Wash: International Association for the Study of Pain Press; 1995.

[8] Tarasuk V, Eakin JM. Back problems are for life: perceived vulnerability and its implications for chronic disability. Journal of Occupational Rehabilitation. 1994;4:55-64.

[9] Nachemson AL. Newest knowledge of low back pain: a critical look. Clin Orthop. 1992;279:8-20.

[10] Mayer TG, Gatchel RJ. Functional Restoration for Spinal Disorders: The Sports Medicine Approach. Philadelphia, Pa: Lea & Febiger; 1988.

[11] Bigos SJ, Battie MC, Spengler DM, et al. A prospective study of work perceptions and psychosocial factors affecting the report of back injury. Spine. 1991;16:1-6.

[12] Riihimaki H. Low back pain, its origin, and risk indicators. Scand J Work Environ Health. 1991;17:81-90.

[13] Bongers PM, de Winter CR, Kompier MAJ, Hildebrandt VH. Psychosocial factors at work and musculoskeletal disease. Scand J Way* Environ Health. 1993;19:297-312.

[14] Deyo RA, Anderson G, Bombardier C, et al. Outcome measures for studying patients with low back pain. Spine. 1994;19:2032S-2036S.

[15] Ekberg K. Workplace changes in successful rehabilitation. Journal of Occupational Rehabilitation. 1995;5:253-296.

[16] Crossman J, Zuliani A, Preston J, Gluck L. Factors influencing motivation to return to work and effort in rehabilitation among injured workers. Physiotherapy Canada. 1996;48:263-265

[17] Myers AM. Coming to grips with changing Canadian health care organizations: challenges for evaluation. Canadian Journal of Program Evaluation. 1996;11:127-147.

[18] Cole B, Finch E, Gowland C, Mayo N. Physical Rehabilitation Outcome Measures. Toronto, Ontario, Canada: Canadian Physiotherapy Association in cooperation with Health and Welfare Canada Communication Group-Publishing, Supply and Services Canada, 1994.

[19] Roland M, Morris R. A study of the natural history of back pain, part 1: the development of a reliable and sensitive measure of disability in low back pain. Spine. 1983;8:141-145.

[20] Fairbanks JTC, Couper J, Davies JB, O'Brien JP. The Oswestry Low Back Pain Disability Questionnaire. Physiotherapy. 1980;66:271-273.

[21] Nicholas MK. An Evaluation of Cognitive, Behavioural, and Relaxation Treatments for Chronic Lao Back Pain. Sydney, New South Wales, Australia: I University of Sydney; 1988. Doctoral thesis.

[22] Scott J, Huskisson EC. Graphic representation of pain. Pain. 1976; 2:175-184.

[23] Delitto A. Are measures of function and disability important in low back care? Phys Ther. 1994;74:452-462.

[24] Kopec JA, Esdaile JM, Abrahamowicz M, et al. The Quebec Back Pain Disability Scale: measurement properties. Spine. 1995;20-341-352.

[25] Kopec JA, Esdaile JM, Abrahamowicz M, et al. The Quebec Back Pain Disability Scale: conceptualization and development. J Clin Epidemiol. 1996:49-151-161.

[26] Schoppink LEM, van Tulder MW, Koes BW, et al. Reliability and validity of the Dutch adaptation of the Quebec Back Pain Disability Scale. Phys Ther. 1996;76:268-275.

[27] Myers AM. The clinical Swiss army knife: empirical evidence on the validity of IADL functional status measures. Med Care. 1992;30: MS96-MS111.

[28] Myers A.M. Holliday PJ, Harvey KA, Hutchinson KS. Functional performance measures: Are they superior to self-assessments? J Gerontol. 1993;48:M196-M206.

[29] Fischer DG, Fick C. Measuring social desirability: short forms of the Marlowe-Crowne Social Desirability Scale. Educational and Psychological Measurement. 1993;53:417-425.

[30] Cody RP, Smith JK. Applied Statistics and the SAS Programming Language. New York, NY. Elsevier Science Publishing Co; 1991.

[31] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-428.

[32] Streiner DL, Norman GR. Health Measurement Scales:. A Practical Guide to Their Development and Use. New York, NY: Oxford University Press Inc; 1995.

[33] Lord SR, Llyod DG, Nirui M, et al. The effect of exercise on gait pattern of older women: a randomized controlled trial. J Gerontol. 1996;51:M64-M70.

[34] DeVellis RF. Scale Development Theory and Application. Newbury Park, Calif: Sage Publications; 1991.

[35] Portney GL, Watkins MP. Foundations of Clinical Research: Application to Practice. East Norwalk, Conn: Appleton & Lange; 1993.

[36] Cohen J. Statistical Pouter Analysis for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum Associates Publishing Co; 1988.

[37] Kazis L, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989:27(suppl):S178-S189.

[38] Report to the Workers' Compensation Board on the Evaluation of the Community Clinic Program in the Rehabilitation of Workers With Soft Tissue Injury. Toronto, Ontario, Canada, Institute for Work & Health; 1995.

[39] Feinstein AR, Josephy BR, Wells CK. Scientific and clinical problems in indexes of functional disability. Ann Intern Med. 1986;105:413-420.

Appendix.

Resumption of Activities of Daily Living Scale: Instructions, Rating Format, and Content

Since your injury, to what extent hove you resumed your usual activities in each of the following areas? If you do not do an activity, put N/A (nonapplicable) beside the scale. As you rote each activity, think of how you are today. Circle the number on the scale for each question.

0%       10%   20%   30%   40%
Not at
All

50%           60%   70%   80%   90%   100%
Moderate                              Complete
Resumption                            Resumption

1. Sleeping patterns
2. Sexual activity
3. Self-care (eg, washing, dressing)
4. Light household chores (eg, doing dishes, making beds,
   preparing meals)
5. Heavy household chores (eg, yardwork, cleaning windows,
   doing laundry)
6. Shopping
7. Socializing with family and friends inside your
8. Socializing with family and friends outside your home a
9. Traveling (in cars, buses, sic) for less than 30 minutes
10. Traveling (in cars, buses, sic) for longer than 1 hour
11. Engaging in your usual recreational activities
12. Engaging in your usual paid employment

RM Williams, PhD, PT, is Assistant Professor, School of Rehabilitation Science, McMaster University, Bldg T-16, Room 128G, 1280 Main St W, Hamilton, Ontario, Canada L8S 4K1 ([e-mail]). Address all correspondence to Dr Williams.

AM Myers, PhD, is Associate Professor, Department of Health Studies and Gerontology. University of Waterloo, Waterloo, Ontario, Canada.

This study was completed in fulfillment of the thesis requirements for Dr Williams' Doctor of Philosophy in Health Studies degree at the University of Waterloo.

This study was approved by the Office of Human Research at the University of Waterloo and by the ethics committees of the Kitchener-Waterloo Hospital, Kitchener, Ontario, Canada, and the Hamilton Civic Hospitals, Hamilton, Ontario, Canada.

This research was supported by a grant from the Institute for Work & Health, Toronto. Ontario, Canada.

The results of this study, in part, were presented at the Canadian Evaluation Society Conference, May 5-7, 1997, Ottawa, Ontario, Canada, and the Canadian Physiotherapy Association Conference, June 19-23, 1997, Winnipeg, Manitoba, Canada.

This article was submitted July 11, 1997, and was accepted February, 9, 1998.