Movement diagram and "end-feel" reliability when measuring passive lateral rotation of the shoulder in patients with shoulder pathology.

Key Words: Manual therapy, Movement diagram, Reliability, Shoulder.

Manual therapy, in our opinion, is becoming increasingly popular with clinicians, as evidenced by the inclusion of manual therapy techniques in professional (entry-level) physical therapist curricula.[1] Despite the clinical conviction that the manual evaluation of joint stiffness is a valuable assessment tool, there is a limited amount of literature that examines the reliability of manual estimates of joint stiffness.[2]

Only a few studies have examined the reliability of passive motion assessments, with most studies focusing on therapist agreement when evaluating motion of the spinal intervertebral joint.[3-11] For the most part, reliability findings have been poor, although it appears that the reliability of pain measurements may be acceptable compared with clinical judgments of joint stiffness.[8] Similar studies of manual therapy assessments of peripheral joints such as the shoulder, however, have not been conducted.

Maher and Adams[10] have suggested that the evaluation of joint stiffness involves many dimensions. The perception and documentation of passive stiffness, therefore, are likely to be very complex. We believe a research design that incorporates both graphic and verbal methods of documenting joint findings may help to determine whether reliability is related to how results are recorded. Two such methods available to physical therapists are movement diagrams[12,13] and Cyriax[14] "end-feel" categories.

Movement Diagram

Maitland[12,13] used movement diagrams in peripheral and spinal manual therapy to document resistance, pain, and spasm (involuntary muscle contraction) as they relate to the passive range of motion (ROM) of a joint. In his text Vertebral Manipulation,[13] Maitland illustrated the relationship between normal joint compliance and movement diagrams that he believed exists for the spine. An example of a completed movement diagram for one of the subjects in our sample is illustrated in Figure 1. The dimensions were 39 rum (horizontal) by 27 mm (vertical). The construction of the pain and resistance curves followed the method described by Maitland.[13] The x-axis (line A-B) was used to indicate the onset of pain ([P.sub.1]) reported by the subject and the onset of resistance ([R.sub.1]) and limit of lateral rotation ROM (L) perceived by the therapist. Lines were drawn to indicate the change in pain noted by the patient and the change in resistance perceived by the therapist as the shoulder was moved farther into lateral rotation. Although no comparisons of these lines were made, the therapists were allowed to draw these curves, as it is part of the accepted procedure for the construction of the movement diagram.[13] Maximum pain ([P.sub.2]) reported by the subject or maximum resistance [R.sub.2] perceived by the therapist could limit the movement and was indicated on the top horizontal axis (line C-D) at the top of line L.

[Figure 1 ILLUSTRATION OMITTED]

Because the location of the variable of interest in a movement diagram is marked on a continuous line that is anchored with clear concepts of the beginning and the end of motion, resistance, or pain intensity, this clinical tool may have some conceptual similarities to the visual analog pain scale. Movement diagrams appear to have some of the measurement characteristics deemed desirable by Binkley et al,[9] because the orientation of the scale anchors is consistent from one movement diagram to another and a large number of points are available for rating because of the continuous line.

End-Feel

Cyriax[14] proposed a verbal classification scheme called "end-feel" to describe the resistance to passive movement of a joint. The concept of end-feel defined by Cyriax[14] is related to joint ROM and pain,[15] and it describes a relationship between resistance felt by the examiner (eg, capsular end-feel), pain expressed by the patient (eg, empty end-feel), and joint ROM. End-feel, therefore, may be conceptually similar to movement diagrams because both concepts relate pain and resistance findings to joint ROM. Use of end-feel categories to describe joint function may be preferable to movement diagrams by clinicians because of the simple, categorical measurement scale used in end-feel assessments. Although the reliability and validity of end-feel classifications for patients with osteoarthritis of the knee have been questioned recently,[15] no reliability studies have been done on end-feel classifications for patients with shoulder pathology.

If end-feel classifications or movement diagrams contribute to the diagnosis, treatment selection, and evaluation provided by physical therapists, an understanding of the reliability of these measurements in patients with shoulder pathology is essential. The purpose of our study was to evaluate intrarater and interrater reliability of a movement diagram and Cyriax[14] classifications of end-feel when assessing and documenting passive lateral rotation of the shoulder in patients with known shoulder pathology. The correlation between movement diagram variables also was evaluated because movement diagrams and end-feel categories both describe constructs that include resistance, pain, and ROM findings.

Method

Subjects

Thirty-six patients with shoulder pathology documented by an orthopedic surgeon were recruited from a large medical facility specializing in the surgical and nonsurgical management of upper-extremity pathological conditions. Patients with a shoulder problem were selected from the surgeon's caseload to provide a spectrum of diagnoses and severities. Thirty-four patients (18 female, 16 male) agreed to participate. The remaining 2 subjects failed to attend for testing. The sample size was sufficient to detect an intraclass correlation coefficient (ICC) greater than .40 at an alpha level of .05 and a beta level of .80.[16]

The mean values ([+ or -] SD) for age, height, and weight were 55.0 [+ or -] 18.5 years (minimum = 15, maximum = 80), 171.5 [+ or -] 9.2 cm (minimum = 154.9, maximum = 187.9), and 78.0 [+ or -] 15.8 kg (minimum = 52.6, maximum = 124.7), respectively. The surgeon's diagnosis was our evidence that there was known shoulder pathology. Ten patients had been managed nonoperatively (humeral fracture, n = 6; scapular fracture, n = 1; frozen shoulder, n = 1; rotator cuff tendinitis secondary to multidirectional instability, n = 1; rotator cuff tear, n = 1) and were tested an average of 5.2 months (minimum=3, maximum = 6) after the onset of their problem. Twenty-four patients received surgery (rotator cuff repair, n = 9; reflex sympathetic dystrophy after surgery for Colles fracture, n = 6; total shoulder joint arthroplasty, n = 4; Bankart repair/ ligament reconstruction, n = 4; acromioplasty, n = 1) and were tested an average of 5.8 months (minimum = 1, maximum = 13) after their operation. Thus, the sample included patients with hypermobile (instability) versus hypomobile (frozen shoulder) joints, bony (fractures) versus soft tissue (tendon) pathologies, surgical versus nonsurgical pathologies, and anatomic versus total arthroplasty joints.

Patients were included in the study if they provided written, informed consent and had a previous or current referral for physical therapy for their current shoulder problems. They were excluded if they had evidence of glenohumeral or acromial dislocation, delayed or nondelayed union of the shoulder girdle skeletal components, or active malignancy in the shoulder and neck region (confirmed by the attending surgeon). Because resistance to passive motion would be affected by abnormal muscle tone, patients with central nervous system lesions were excluded. Patients with a postoperative contraindication to performing lateral rotation of the shoulder joint, as specified by the surgeon, were also excluded.

Orthopedic Physical Therapists

Because level of training and experience may affect examiner error during manual therapy measurements,[17] the therapists recruited for this study were certified as manual therapists at the highest level with their successful completion of the Part A and Part B Manipulative Therapy Certification by the Orthopaedic Division of the Canadian Physiotherapy Association.[18] They had practiced physical therapy for a mean of 16 years, had used manual therapy assessment and treatment techniques for a mean of 14 years, and were actively involved in teaching undergraduate and postgraduate manual therapy courses. Both therapists reported similar patterns of movement diagram use in clinical practice and during teaching activities. Movement diagrams were used more than 75% of the time in teaching activities but less than 50% of the time in clinical practice. The therapists had no a priori knowledge of the study purpose and were asked to participate in a study that would require them to (1) evaluate passive lateral rotation in patients with shoulder pathology and (2) record their findings using a movement diagram and the Cyriax[14] end-feel categories. In this study, therefore, only therapists with extensive, advanced training in manipulative therapy were raters; the same results cannot be assured for persons with lesser levels of training.

Movement Diagram

A movement diagram constructed according to the description given by Maitland[13] was used to document the passive movement findings. Seven variables were derived from the movement diagram. Two variables quantified the location in range of the onset of resistance and the onset of pain: horizontal distance from the y-axis (line A-C) to [R.sub.1] and to [P.sub.1] (in millimeters), respectively. Two variables quantified the location in range of the maximum resistance and the maximum pain: horizontal distance from the y-axis (line A-C) to [R.sub.2] and to [P.sub.2] (in millimeters), respectively. Two variables quantified the maximum magnitude of the passive resistance and the maximum pain intensity: vertical distance from the x-axis (line A-B) to [R.sub.2] and to [P.sub.2] (in millimeters), respectively. The seventh variable defined the total passive lateral rotation ROM: horizontal distance from the y-axis (point A) to L. To obtain these variables from the therapists' movement diagrams, measurements were made by an independent observer who was blinded to the purpose of the study, the raters' identity, and the order of testing. Although this procedure may have reduced error in the measurements, it would not occur in practice. Prior to obtaining values with a ruler, the independent observer was trained by the investigators in a consistent approach to measurement.

Each subject was positioned comfortably in the supine position, with the shoulders exposed and the arms resting at the side of the body in a small amount of abduction (approximately 20 [degrees]). All assessments were commenced from this position with the elbow in 90 degrees of flexion. The therapist stood at the side of the treatment table. The subject was asked to (1) report the onset of and the change in shoulder pain as the humerus was passively rotated and, (2) inform the therapist if he or she did not want the am to be moved any farther. The evaluation procedure was always performed in a slow, controlled manner and limited to a maximum of 5 passive motions. Otherwise, the therapists were allowed to perform the assessment to. a manner they were accustomed to performing in their teaching and clinical duties. The uninvolved shoulder was always examined prior to the evaluation of the involved shoulder. The therapists were required, to hand in each movement diagram immediately on completion, and they were not allowed access to the first movement diagram during their-second assessment.

End-Feel

After drawing the movement diagram, the therapists recorded the end-feel of the involved shoulder joint according to the categories described by Cyriax.[14] They were not allowed to see their first decision regarding end-feel during their second assessment.

End-feel has been defined as the sensations that are felt by the examiners' hands at the extreme of the possible range during passive movement testing of a joint.[14] Table 1 lists the "end-feel" categories used in this study. Following discussions with the 2 therapists, an additional category was added to differentiate between a normal capsular end-feel and an abnormal capsular end-feel. A capsular end-feet was originally defined by Cyriax[14] as a "hardish arrest of movement, with some give in it ... the way the normal shoulder or hip stops at the extreme of each rotation. This feeling, appearing before normal full range is reached, suggests non-acute ... arthritis." This definition suggests that the important difference between a normal capsular end-feel and an abnormal capsular end-feel is the location in range of the "hardish arrest of movement." For the purpose of our study, therefore, we defined an abnormal capsular end-feel as a hardish arrest of movement with some give in it, appearing before full range is reached. We defined a normal capsular end-feel as a hardish arrest of movement with some give, but occurring at the extreme of full range.

Table 1.
Categories of Joint "End-Feel"[14] Used in the Study

End-Feel Category      Description of End-feel

Bone-to-bone           Abrupt halt to movement when two hard
                       surfaces meet
Spasm                  Hard vibrant twang
Capsular
  Normal               Hardish arrest to movement with some give,
                       full ROM(a)
  Abnormal(b)          Hardish arrest to movement with some give,
                       less than full ROM
Springy block          Rebound is seen and felt at extreme of
                       possible range
Tissue approximation   Limb engages limb, but could move farther
Empty                  A lot of pain before end ROM, no
                       resistance, patient says "stop"

(a) ROM=range of motion.

(b) Abnormal end-feel category and description added for this study.

Test Routine

Prior to testing, the 2 physical therapists were briefed regarding their tasks. This briefing consisted of a review of the approach to composing the movement diagram and definitions for the Cyriax[14] end-feel categories. The subjects were then directed to separate treatment rooms. Both therapists were given a randomized order of patients for testing. For each group of patients, therapists were given a randomly generated order for testing. The subject identified the involved shoulder for the therapist, but no diagnosis or clinical history was provided. After evaluating all of the subjects for the first time, the therapists were given a second and different randomized order of patients for their second evaluation. The time between the first and second evaluations was 15 to 20 minutes. Subjects were evaluated in 4 groups of 6 subjects and 2 groups of 5 subjects over 3 separate days of testing at approximately 1-month intervals.

Data Analysis

Intrarater and interrater reliability were assessed using the ICC (2,1).[19] For each of the movement diagram variables, a 2-way analysis of variance (ANOVA)[30] was performed to generate mean square values, which were used to calculate the ICCs and their 95% 2-sided confidence intervals.[19] The ANOVA calculations for the intrarater ICCs identified the subjects as one factor and the testing occasion (ie, first or second assessment) as the second (repeated) factor. The ANOVA calculations for the interrater ICCs identified the subjects as one factor and the physical therapist (ie, therapist A or therapist B) as the second (repeated) factor. Interrater reliability statistics for generalizing to other raters ([[Rho].sub.inter_ran]), with their 95% one-sided lower-limit confidence interval and standard error of measurement (SEM) following the method of Eliasziw et al,[21] were also calculated[22] to compare the results of 2 alternative approaches for calculating reliability. When 2 raters perform repeated measurements on a sample of subjects, [[Rho].sub.inter_ran] estimates reliability with more precision than the ICC because it is derived from multiple (in this case, 2) observations per subject.[21] The SEM was calculated because it allows the expression of measurement error in the same units as those of the measurement tool. Intrarater and interrater agreement for the Cyriax[14] end-feel categorical data were evaluated using the kappa statistic ([Kappa]).[23] Moderate reliability (ICC, [[Rho].sub.inter_ran], or [Kappa] [is greater than] .40) was considered the minimum required for clinical utility. Because kappa is a chance-corrected statistic, lower values were anticipated.

Because both movement diagrams and end-feel categories describe a clinical construct about resistance, pain, and passive ROM, the relationship among these variables from the movement diagram were of interest. Using the Pearson product-moment correlation coefficient (r),[20] the correlations of pain and resistance variables with the ROM variable (horizontal distance to the limit of lateral rotation) from the movement diagram were calculated. These correlations provided insight into how the therapists associated pain and resistance findings with joint ROM.

Results

Movement Diagram

Table 2 lists the intrarater ICCs with their 95% confidence intervals. Therapist B generally yielded consistently better values than did therapist A. Table 3 shows the interrater ICCs and the SEM values for drawing the movement diagram variables. The ICC values listed for the first and second assessments demonstrate that interrater agreement improved for all variables with a second round of assessments. Interrater reliability statistics ([[Rho].sub.inter_ran]) and SEMs, calculated according to the method described by Eliasziw et al,[21] are included in Table 3 for comparison. The [[Rho].sub.inter_ran] reliability values tend to lie between or slightly below the ICC values for each assessment and, as noted by Eliasziw et al,[21] provide a more precise estimate of interrater reliability. Lower confidence limits associated with these coefficients vary from a minimum of 0.35 to a maximum of 0.79.

Table 2.
Intrarater Reliability for Movement Diagram Variables

                                     ICC(b)
Variable(a)                          Therapist A     Therapist 8

Pain
  Horizontal distance to [P.sub.1]   .58 (.35-75)    .86 (.75-.92)
  Horizontal distance to [P.sub.2]   .73 (.55-.84)   .87 (.77-.93)
  Vertical distance to [P.sub.2]     .78 (.64-.88)   .87 (.78-.93)
Resistance
  Horizontal distance to [R.sub.1]   .71 (.51-.83)   .69 (.48-.82)
  Horizontal distance to [R.sub.2]   .78 (.63-.87)   .89 (.81-.94)
  Vertical distance to [R.sub.2]     .73 (.56-.85)   .80 (.65-.88)
Lateral rotation ROM
  Horizontal distance to L           .74 (.57-.85)   .86 (.76-.92)

(a) [P.sub.1] = onset of pain, [P.sub.2] = maximum pain, [R.sub.1] = onset of resistance, [R.sub.2] = maximum resistance, ROM = range of motion, L = limit of passive lateral rotation ROM.

(b) Interclass correlation coefficient (ICC[2,1])[19] (95% confidence interval).

Table 3.
Interrater Reliability for Movement Diagram Variables

Variable(a)                           1st ICC(b)      2nd ICC(b)
Pain
  Horizontal distance to [P.sub.1]   .59 (.35-.75)   .88 (.80-.93)
  Horizontal distance to [P.sub.2]   .85 (.73-.91)   .91 (.85-.95)
  Vertical distance to [P.sub.2]     .80 (.66-.89)   .86 (.76-.92)
Resistance
  Horizontal distance to [R.sub.1]   .65 (.44-.79)   .76 (.61-.86)
  Horizontal distance to [R.sub.2]   .90 (.83-.95)   .91 (.84-.95)
  Vertical distance to [R.sub.2]     .34 (.05-.57)   .82 (.69-.90)
Lateral rotation ROM
  Horizontal distance to L           .83 (.70.-.90)   .90 (.83-.95)

                                     [[Rho].sub.
Variable(a)                          inter_ran]     SEM(mm)(d)
Pain
  Horizontal distance to [P.sub.1]   .71 (.59)         6.7
  Horizontal distance to [P.sub.2]   .84 (.76)         4.2
  Vertical distance to [P.sub.2]     .82(.73)          4.3
Resistance
  Horizontal distance to [R.sub.1]   .64(.51)          6.5
  Horizontal distance to [R.sub.2]   .86(.79)          3.8
  Vertical distance to [R.sub.2]     .53(.35)          3.3
Lateral rotation ROM
  Horizontal distance to L           .83(.75)          4.2

(a) [P.sub.1] = onset of pain, [P.sub.2] = maximum pain, [R.sub.1] = onset of resistance, [R.sub.2] = maximum resistance, ROM=range of motion, L=limit of passive lateral rotation ROM.

(b) Intraclass correlation coefficient (ICC[2,1])[19] (95% confidence interval).

(c) Random effects reliability statistic (lower one-sided 95% confidence interval), Eliasziw et at method.[21]

(d) SEM= standard error of measurement, Eliasziw et al method.[21]

Results of the correlational analyses between the pain and resistance variables and joint ROM measured with the movement diagram are summarized in Table 4. The results show that the relationship between movement diagram horizontal distance to maximum pain and resistance and the horizontal distance to L was strong and linear regardless of factor (therapist or assessment), since the Pearson correlation coefficients varied from .914 to .999 (P [is less than] .0001). For the movement diagram variables indicating the onset of pain or resistance, this relationship was less marked, with Pearson correlation coefficients varying from .573 (P [is less than] .001) to .956 (P [is less than] .0001). Correlational analyses for the relationship between movement diagram vertical pain and resistance variables and the horizontal distance to L were generally poor, varying in strength and nature of association as well as-in statistical significance.

Table 4.
Correlation of Pain and Resistance With Horizontal Distance
to Limit of Lateral Rotation From Movement Diagram(a)

                                     Therapist A         2nd
Variable(b)                          1st Assessment   Assessment

Pain
  Horizontal distance to [P.sub.1]      .573(c)           .936
  Horizontal distance to [P.sub.2]      .914              .999
  Vertical distance to [P.sub.2]       -.275(d)          -.467(e)

Resistance
 Horizontal distance to [P.sub.1]       .899              .940
 Horizontal distance to [P.sub.2]       .976              .999
 Vertical distance to [P.sub.2]         .327(d)           .390(f)

                                     Therapist B        2nd
Variable(b)                          1st Assessment   Assessment

Pain
  Horizontal distance to [P.sub.1]      .947              .956
  Horizontal distance to [P.sub.2]      .998              .998
  Vertical distance to [P.sub.2]       -.581(c)          -.506(e)

Resistance
 Horizontal distance to [P.sub.1]       .782              .885
 Horizontal distance to [P.sub.2]       .982              .994
 Vertical distance to [P.sub.2]         .173(d)           .282(d)

(a) Values are Pearson product-moment correlation coefficients (r).[20] Shaded area indicates r significantly different than zero, P < .001.

(b) [P.sub.1] = onset of pain, [P.sub.2] = maximum pain, [R.sub.1] = onset of resistance, [R.sub.2] = maximum resistance.

(c) r significantly different from zero, P < .001.

(d) r not significantly different from zero, P > .05.

(e) r significantly different from zero, P < .01.

(f) r significantly different from zero, P < .05.

End-Feel

Therapist A agreed with the initial decision for end-feel category in 79.4% of subjects over 2 testing occasions. Similarly, therapist B agreed with the initial decision for end-feel category in 81.8% of subjects over 2 testing occasions. Because a high proportion of subjects were classified as having an abnormal capsular end-feel, the expected agreement by chance alone was relatively high. This chance-corrected agreement was reflected in kappa coefficients that were only moderate ([Kappa] = .48 for therapist A, [Kappa] = .59 for therapist B), as defined by Landis and Koch.[24]

Therapist A agreed with therapist B in 82.4% of the subjects during the first assessment. This agreement increased to 90.9% during the second assessment. Kappa statistic values for chance-corrected agreement between therapists ([Kappa] = .62 during the first assessment, [Kappa] = .76 during the second assessment) indicated "substantial" interrater reliability, as defined by Landis and Koch.[24]

Discussion

Movement Diagram

Level of training and experience may affect examiner error during manual therapy measurements.[17] The reliability values from our study, therefore, may be better than those found previously because the raters were certified at the highest level by their national professional association and had practiced manual therapy for a mean of 14 years and because we trained them in the joint measurement procedures prior to testing. Accordingly, the study results should only be generalized to physical therapists with similar levels of training, and then only with caution because by having training sessions and an independent measurer for the diagram, there were factors to reduce error that would not be available in clinical practice.

The study results, however, reinforce the importance of evaluating the reliability of peripheral joint assessments separately. Lower confidence limits for random rater effects [[Rho].sub.inter_ran] indicated moderate to substantial ability ([is greater than] .40 to [is greater than] .60, respectively[24]) for 6 of the 7 movement diagram variables. One explanation for obtaining relatively high ICCs for movement diagrams of lateral rotation of the shoulder may be the large amplitude of motion under investigation. If a small ROM had been studied, as the case has been with spinal motion, the total variability would be small. Because reliability measures are influenced by the proportion of the total variance that is due to error, ICC calculations for reliability would tend to be low.[25] Intervertebral joints are capable of only small movements, and clinicians may not be able to reliably measure the proportion of normal movement at which pain and resistance occur when assessing small amplitudes of movement. Our findings suggest that reliability data for measurements of peripheral joint movements such as lateral rotation of the shoulder may be fundamentally different than reliability data for measurements of smaller-amplitude intervertebral joint movements. Previous findings[3-7] of poor Movement diagram reliability may have been affected by the small amplitude of movement that was under investigation rather than by the scale itself. Data, however, will be needed to demonstrate whether reliability has been influenced by the range of movement available or by the degree of error.

Figure 2 illustrates a plot of the raw data for vertical distance to [R.sub.2] during the first assessment, which yielded the poorest interrater ICC of .34) (Tab. 3). Each diamond represents a value for vertical distance to [R.sub.2] that was obtained from one subject. The fines between the values obtained by therapist A and the values obtained from therapist B identify individual subjects. Fewer individual data points are visible because of the large number of subjects who yielded identical values for this variable. Data points above the thick horizontal line (identified by the arrow) represent overlapping values for 16 subjects and values that differed by a maximum of 4 mm for an additional 12 subjects. Thus, most of the data points were located near the end of the vertical scale, giving the visual impression of good agreement in the presence of a low ICC. Although most of the values were obtained near the top of the movement diagram (ie, low variability), therapist B marked [R.sub.2] much lower for 5 subjects than did therapist A. The lower variability between subjects demonstrated by therapist A, combined with substantial disagreement by therapist B for 5 subjects, resulted in a low interrater ICC of .34. In conformity with the visual impression of some consistency in Figure 2, however, the SEM of 3.3 nun for vertical distance to [R.sub.2] is the lowest of all SEM values for variables derived from the movement diagram (Tab. 3).

[Figure 2 ILLUSTRATION OMITTED]

This example illustrates one limitation of ICCs and how, in our view, the SEM can provide useful additional, but sometimes paradoxical, reliability information. The reader is reminded that a high ICC indicates that a measure is able to discriminate among subjects, whereas a low SEM reveals that measurement error is small.[26]

Although the reliability values from the movement diagram scale used in this study are generally favorable, we believe that clinicians should be aware that the level of reliability required for individual comparisons, such as those made with patients, needs to be higher than for situations where group comparisons are the primary objective. Looking at the error that can be anticipated when drawing a movement diagram illustrates this point in the same units as those of the measurement tool. The SEMs for the vertical measures varied from 3.3 to 4.3 mm, or 12% to 16% of the y-axis (27 mm) on the movement diagram. The SEMs for the horizontal measures varied from 3.8 to 6.7 mm, or 10% to 17% of the x-axis (39 mm) on the movement diagram. The minimum difference between pretreatment and posttreatment assessments that must be exceeded to be certain that a true change has occurred can be calculated using these SEM values and the following equation:

[Z.sub.[Alpha]]/2 [square root] 2(SEM)

where [Z.sub.[Alpha]/2] = 1.96 when [Alpha] = .05.[21] This minimum difference for the vertical measures varied from 9.2 to 12.0 mm, or 34% to 44% of the y-axis on the movement diagram. The minimum difference for the horizontal measures varied from 10.5 to 18.6 mm, or 27% to 48% of the x-axis on the movement diagram. Therefore, although our reliability values are encouraging, the error that may be present when drawing movement diagrams may encompass from 30% to almost 50% of the total range of the movement diagram scale. Dialogue with clinicians and further research are needed to assess the amount of error that therapists should accept when using movement diagrams.

End-Feel

The kappa values for intrarater agreement were lower than those for interrater agreement. Regardless of the error noted with movement diagrams, however, the therapists seemed to agree on the concept of end-feet with this patient sample. This finding is clinically relevant, given the article by Haves et al[15] that questions the validity and reliability of measurements obtained for Cyriax[14] end-feel descriptors in patients with osteoarthritis of the knee. In their discussion, Hayes et al suggested that their reliability findings might have been affected because there was limited variability in the end-feel categories to which patients were assigned. We made the same observation. Most of our subjects were assigned to the abnormal capsular end-feel category. Calculation of the maximum kappa values ([Kappa].sub.max]) from our marginal probabilities generated intrarater values of .93 for therapist A and .74 for therapist B. The corresponding kappa values of .48 for therapist A and .59 for therapist B indicate that differences in the therapists' definitions of end-feel were not the primary source of disagreement. The interrater maximum kappa values of .80 for the first assessment and .77 for the second assessment, however, were closer to the interrater kappa values of .62 for the first assessment and .76 for the second assessment. This finding indicates that the therapists agreed as much as was possible, given their underlying rates of classification.

Taken together, the movement diagram and end-feel results extend previous arguments regarding the superior reliability of clinical judgments of pain compared with joint stiffness.[8] Two movement diagram variables that are not defined by end-feel (ie, judging the location of the onset of resistance and pain in joint ROM) were associated with the largest amount of error. Conversely, 2 movement diagram variables that are defined by end-feel (ie, determining the location of both the maximum resistance and pain in joint ROM) were more reliable, with less measurement error. Therefore, we believe that judgments of end-feel may be more reliable when clinicians' decisions about the location of pain and resistance are linked to a conceptual framework that specifies an association (mechanical or otherwise) with the end of joint motion. This belief is consistent with (1) our higher kappa values for end-feel agreement, (2) our high reliability for judging the location of maximum resistance and pain, and (3) our high correlation between these 2 variables and joint ROM.

Although examiner bias may be a source of error,[17] We believe that examiner bias may have contributed to the higher levels of agreement found in our study. For example, we believe that manual therapists are well aware of the mechanical characteristics of soft tissue. Therefore, they would expect the maximum resistance to be located at or near the end of the available ROM. Combined with our subdivision of "capsular end-feel" into normal and abnormal categories, this examiner bias may have contributed to the level of agreement noted for end-feel.

An analysis of reliability of data obtained from a movement diagram as a function of individual end-feel categories would help to explain the relationship between clinical constructs associated with the end of joint ROM and end-feel classification. Statistical power considerations limited this analytic strategy in our study. Perhaps the differentiation of stiffness (and pain) reliability assessment into an "onset" and "maximum" somewhere in range identifies two other dimensions in Maher and Adams' multidimensional concept of stiffness[10] that is not revealed through end-feel categories.

Conclusions

Because very skilled raters were used and our study had other elements not found in clinical practice, our results are generalizable only to physical therapists who are highly trained in the use of manual therapy techniques. In our sample of patients, however, the measurements obtained using movement diagrams and end-feel to describe shoulder joint function appear to be more reliable than the previous literature on other joints[3-11,15] has suggested. Physical therapists should be aware, however, that there are some limitations to ICCs. Calculation of the SEM shows there is a need to determine how much documentation error is clinically acceptable when using movement diagrams.

Both of the movement diagram variables representing location of maximum pain and resistance in the available ROM were strongly associated with ROM as measured by the movement diagram and exhibited high levels of reliability. This finding is consistent with the levels of agreement found for recording joint status by Cyriax[14] end-feel, because some end-feel definitions include the concept of pain and resistance related to ROM. Additional end-feel categories introduced in the study design may have been a factor in our reliability findings, and their use in future research on joint assessment may help to clarify the clinical utility of end-feel.

Acknowledgments

We thank Anne Edgell-Kennedy and Beverley Padfield for giving their time to evaluate patients in the study and Dr Michael Eliasziw and Binu Lamba for their assistance with the statistical analysis.

References

[1] Commission on Accreditation in Physical Therapy Education. 1997-1998 Accreditation Handbook Rev ed. Alexandria, Va: American Physical Therapy Association; 1997:70.

[2] Chesworth BM, Vandervoort AA, Koval JJ. A pilot study to compare the subjective and objective evaluation of passive ankle joint stiffness. Physiotherapy Canada. 1991;43:13-18.

[3] Matyas TA, Bach TM. The reliability of selected techniques in clinical arthrometrics. Australian Journal of Physiotherapy. 1985;31: 175-199.

[4] Bond PB. Reliability of compliance assessment with passive physiological intervertebral movements: a comparison of cervical and lumbar tests. Australian Journal of Physiotherapy. 1986;32:68. Abstract.

[5] Carty G. A comparison of the reliability of manual tests of compliance using accessory movements in peripheral and spinal joints. Australian Journal of Physiotherapy. 1986;32:68. Abstract.

[6] Melville D. A comparison of the reliability of manual tests of compliance using passive physiological movements in spinal and peripheral joints. Australian Journal of Physiotherapy. 1986;32:69. Abstract.

[7] Young L. The reliability with which manual therapists detect changes in the behaviour of resistance using passive intervertebral movements. Australian Journal of Physiotherapy. 1986;32:69. Abstract.

[8] Maher C. Adams R. Reliability of pain and stiffness assessments in clinical manual lumbar spine examination. Phys Ther. 1994;74: 801-809.

[9] Binkley J, Stratford PW, Gill C. Interrater Reliability of lumbar accessory motion mobility testing. Phys Ther. 1995.75:786-792.

[10] Maher C, Adams R. Is the clinical concept of spinal stiffness multidimensional? Phys Ther. 1995;75:854-860.

[11] Gonnella C, Paris SV, Kutner M. Reliability in evaluating passive intervertebral motion. Mrs Ther. 1982;62:436-444.

[12] Maitland GD. Peripheral Manipulation. 2nd ed. London, England: Butterworth & Co (Publishers) Ltd; 1977:338-353

[13] Maitland GD. Vertebral Manipulation. 5th ed. London. England: Butterworth & Co (Publishers) Ltd; 1986:351-3772

[14] Cyriax J. Textbook of Orthapaedic Medicine, Volume 1: Diagnosis of Soft Tissue Lesions. 6th ed. London, England: Bailiere Tindall; 1975:76-77.

[15] Haves KW, Petersen C, Falconer J. An examination of Cyriax's passive motion tests with patients having osteoarthritis of the knee. Phys Ther. 1994,74:697-707.

[16] Donner A, Eliasziw M. Sample size requirements for reliability studies. Stats Med. 1987;6:441-448.

[17] Riddle DL. Measurement of accessory motion: critical issues and related concepts. Phys They. 1992;72:865-874.

[18] Orthopaedic Division Education Syllabus. Toronto, Ontario, Canada: Canadian Physiotherapy Association; 1991.

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

[20] Sigmastat Statistical Software, Version 1.0. San Rafael, Calif: Jandel Scientific; 1994.

[21] Eliasziw M, Young SL, Woodbury MG, Fryday-Field K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability, using goniometric measurements as an example. Phys Ther. 1994;74:777-788.

[22] BMDP Statistical Software, 1990 Release. Berkeley, Calif. University of California Press; 1990.

[23] Cohen J. A coefficient of agreement for nominal scales. Educational and Psychological Measurement. 1960;20:37-46.

[24] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159-174.

[25] Portney LG, Watkins MP. Foundations of Clinical Research: Applications to Practice. East Norwalk, Conn: Appleton and Lange; 1993:505-528.

[26] Stratford PW, Goldsmith CH. Use of the standard error as a reliability index of interest: an applied example using elbow flexor strength data. Phys Ther. 1997;77:745-750.

BM Chesworth, MClSc, is Research Associate, Patient Care Support-Physiotherapy, London Health Sciences Centre-Victoria Campus, 800 Commissioners Rd E, Rm 205C/MU2, London, Ontario, Canada N6A 4G5 ([e-mail]). Address all correspondence to Mr Chesworth.

JC MacDermid, MSc, is Co-Director, Clinical Research Laboratory, Hand and Upper Limb Centre, St Joseph's Health Centre, London, Ontario, Canada.

JH Roth, MD, is Medical Director, Hand and Upper Limb Centre, St Joseph's Health Centre.

SD Patterson, MBChB, is Surgeon, Hand and Upper Limb Centre, St Joseph's Health Centre.

This study was approved by the Review Board for Health Sciences Research Involving Human Subjects at the University of Western Ontario, London, Ontario, Canada.

This study was funded by the JAL TATA Research Fund of the London District Ontario Physiotherapy Association and the Research Fund, London Health Sciences Centre-Victoria Campus Physical Therapy Department.

This article was submitted June 17, 1997, and was accepted December 17, 1997.