Responsiveness of the Numeric Pain Rating Scale in Patients With Shoulder Pain and the Effect of Surgical Status

Transcription

1 Journal of Sport Rehabilitation, 2011, 20, Human Kinetics, Inc. Responsiveness of the Numeric Pain Rating Scale in Patients With Shoulder Pain and the Effect of Surgical Status Lori A. Michener, Alison R. Snyder, and Brian G. Leggin Context: The Numeric Pain Rating Scale (NPRS) is commonly used to assess pain. Change in the NPRS across time can be interpreted with responsiveness indices. Objective: To determine the minimal clinically important difference (MCID) of the NPRS. Design: Single-group repeated measures. Setting: Outpatient rehabilitation clinics. Patients: Patients with shoulder pain (N = 136). Main Outcome Measures: At the initial evaluation patients completed the Penn Shoulder Score (PSS), which includes pain, satisfaction, and function sections. Pain was measured using an 11-point NPRS for 3 conditions of pain: at rest, with normal daily activities, and with strenuous activities. The NPRS average was calculated by averaging the NPRS scores for 3 conditions of pain. The final PSS was completed after 3 4 wk of rehabilitation. To determine the MCID for the NPRS average, the minimal detectible change of 8.6 points for the PSS function scale (0 60 points) was used as an external criterion anchor to classify patients as meaningfully improved ( 8.6 point change) or not improved (<8.6-point change). The MCID for the NPRS average was also determined for subgroups of surgical and nonsurgical patients. Cohen s effect sizes were calculated as a measure of group responsiveness for the NPRS average. Results: Using a receiver-operating-characteristic analysis, the MCID for the average NPRS for all patients was 2.17, and it was 2.17 for both the surgical and nonsurgical subgroup: area-under-the-curve range (95%CI:.55.95). The effect size for all patients was 1.84, and it was 1.51 and 1.94 for the surgical and nonsurgical groups, respectively. Conclusions: The NPRS average of 3 pain questions demonstrated responsiveness with an MCID of 2.17 in patients with shoulder pain receiving rehabilitation for 3 4 wk. The effect sizes indicated a large effect. However, responsiveness values are not static. Further research is indicated to assess responsiveness of the NPRS average in different types of patients with shoulder pain. Keywords: surgery, minimal clinically important difference, rehabilitation, outcomes Pain ratings are used to evaluate the effects of treatment, and pain intensity at baseline can predict treatment outcome in patients with shoulder pain. 1 8 Pain is commonly assessed in outpatient rehabilitation in patients with shoulder pain using Michener (lamichen@vcu.edu) is with the Dept of Physical Therapy, Virginia Commonwealth University, Richmond, VA. Snyder is with the Athletic Training Program, A.T. Still University, Mesa, AZ. Leggin is with Good Shepherd Penn Partners, Penn Presbyterian Medical Center, Philadelphia, PA. 115

2 116 Michener, Snyder, and Leggin various patient-rated numeric pain-rating scales (NPRSs). An NPRS is described as an 11-point scale with scores from 0 to 10 and anchors of 0 = no pain and 10 = worst possible pain. 9 The 11-point NPRS has been used to assess shoulder pain under various conditions such as pain at rest, pain with normal activities, average pain, best pain, and worst pain. 10,11 When pain is assessed under various conditions, multiple NPRS scores are obtained. The Pennsylvania Shoulder Score (PSS) is a patient-rated outcomes instrument used to measure pain under 3 conditions by rating each pain question using the NPRS. The PSS also has sections to measure function and satisfaction in people with shoulder conditions. 11 Although completing the entire PSS will yield information about pain, satisfaction, and function, clinicians may choose to use only the pain portion of the PSS in certain situations, such as when a patient s pain level is very high and the patient s primary short-term goal is to decrease pain. Pain is typically assessed at various time points in the course of a program of care to determine progress. However, to interpret the change in pain over time, determining the amount of change in pain considered meaningful is required. Clinically meaningful change can be estimated by determining the responsiveness of an instrument. Responsiveness, defined as the ability of an instrument to detect clinically meaningful or clinically important change over time, has been advocated as a critical measurement property of outcome instruments Responsiveness indices can be calculated to assess responsiveness of a measure applied to a group of patients or individual patients. Cohen s effect size 16 is a recommended responsiveness index used to indicate responsiveness for a group of patients; it cannot be applied to individual patients. It is less vulnerable to inappropriate variation because it is calculated using a pooled standard deviation in the denominator. Moreover, there are reference standards for interpretation;.2 is a small effect,.5 is a medium effect,.8 is a large effect. Responsiveness indices indicated for use at the individual patient level, or the amount of change in a measure that is considered meaningful to a group of patients, can be determined through distribution- or anchor-based methods. Distributionbased methods provide information regarding the statistical error associated with a change score, commonly known as the minimal detectable change (MDC) or smallest detectable difference. The anchor-based method uses an external criterion, typically based on patient perception, to operationally define the clinical meaningfulness of the observed change, 14,17 which is often referred to as the minimally clinically important difference (MCID) or the minimal important change. Both the distribution- (MDC) and the anchor-based (MCID) methods provide useful information, but only the anchor-based method can provide an estimate of clinically meaningful change. The distribution-based method determines the statistical magnitude of change, thus providing no information about clinical meaningfulness. Lack of clinical meaning hampers the interpretation of outcomes instruments and limits their usefulness for both directing individual patient care and evaluating clinical studies of treatment effectiveness. The amount of change that is considered meaningful, the MCID, is based on an external measure of change that assesses important change to the patient. A common external criterion measure for the detection of patient-perceived improvement is the Global Rating of Change (GRC), also known as global perceived effect. The GRC has been criticized because it is a transitional scale that requires recall of prior health status Measurement of change that does not require recall may be more

3 Responsiveness of the NPRS in Shoulder Pain 117 appropriate to determine important or clinically meaningful change in patient status because these methods are not vulnerable to recall bias. An anchor-based method, unlike a transitional scale that involves recall, determines meaningful improvement by calculating change in an external criterion measure such as function or overall rating of function or disability. 21 Calculating important change on independent measures such as function or disability reduces the likelihood of inaccurate reflection by the patient regarding a change in health status. Responsiveness values of various pain conditions using the 11-point NPRS have been reported for a variety of musculoskeletal disorders. The MCID for a single NPRS rating or the average of multiple NPRSs for various types of pain conditions over a 1- to 4-week treatment interval ranges from 1 to 3 points Recently, in patients with shoulder pain the MCID for an NPRS of the average of 3 ratings of current pain, worst pain in 24 hours, and least pain in 24 hours was 1.1 points over a 2- to 4-week interval of rehabilitation treatment. 10 This indicates that the average of 3 NPRSs for the 3 pain conditions of the shoulder is responsive to monitor change over time. However, in this study a transitional scale, the GRC, was used as the external criterion to define patient improvement. The GRC may not accurately reflect change in patient-perceived important improvement Furthermore, the effect of surgical status on responsiveness of pain scores in patients with shoulder pain has not been examined. Surgical status may affect responsiveness, because it has influenced the reliability of patient-rated measures of shoulder function. 25 The purpose of this study was to determine responsiveness of an average of 3 NPRS scores from the 3 PSS pain questions in patients undergoing rehabilitation for shoulder pain in outpatient clinics. Specifically, we were interested in determining the MCID of the NPRS average, which was generated from the 3 questions of the PSS on pain at rest, during activities of daily living, and during strenuous activities using a nontransitional anchor-based method to determine meaningful change. Furthermore, we wanted to determine whether the responsiveness index of the MCID for patients who present to rehabilitation postsurgery was different from those who do not have surgery. We hypothesized that patients who had surgery would have a larger MCID than those who did not. Finally, we aimed to determine the group responsiveness index of Cohen s effect size for all patients and for those who had and had not had surgery. These results will provide guidelines for interpreting change scores for the NPRS average for patients with shoulder pain postsurgery and for nonoperative outpatient rehabilitation. Methods A sample of convenience was used. Consecutive patients who presented between 2002 and 2006 with shoulder pain (N = 136) to 17 outpatient clinics across 4 states (Delaware, New Jersey, New York, and Pennsylvania) for the purpose of shoulder rehabilitation were included in the study as described in Table 1. Inclusion criteria included a primary diagnosis of shoulder pain, the ability to read and comprehend English, and age ranging from 18 to 85 years. Patients presented with a variety of shoulder diagnoses with both surgical and nonsurgical shoulder conditions (Table 2). At the initial evaluation, patients were asked to provide demographic information and complete the PSS. 11 The PSS contains 3 questions to assess pain, scored on

4 118 Michener, Snyder, and Leggin Table 1 Demographic and Self-Report Measures of All Subjects, N = 136 Variable (N = 136) Value Age, y (SD) 51.7 (16.4) Sex, n (%) male 66 (48.5) female 70 (51.5) Surgery, n (%) no 104 (76.5) yes 32 (23.5) Symptoms duration, n (%) less than 4 wk 15 (11.0) 4 12 wk 57 (41.9) more than 12 wk 60 (44.1) missing 4 (2.9) NPRS-average a (0 10; 0 = no pain), mean (SD) initial 5.0 (1.6) final 2.0 (1.7) PSS total pain (0 30; 30 = no pain), mean (SD) initial 15.1 (6.3) final 23.7 (5.3) PSS satisfaction (0-10; 10 = completely satisfied), mean (SD) initial 2.3 (2.4) final 7.1 (2.7) PSS function (0 60; 60 = full function), mean (SD) initial 28.6 (12.8) final 44.3 (12.0) NPRS, Numeric Pain-Rating Scale; PSS, Penn Shoulder Score. a (Rest + activities of daily living + strenuous)/3. the 11-point standard NPRS. At the final time point, which was after 3 to 4 weeks of treatment, patients completed the same forms again. At this time, patients were blinded to their scores from the initial evaluation. This study was approved by the University of Pennsylvania Medical Center and Hahnemann University. PSS The PSS is a self-report outcomes tool designed to assess shoulder pain, satisfaction, and function. 11 It contains 3 sections: pain (30 points maximum), satisfaction (10 points maximum), and function (60 points maximum). The PSS total score ranges from 0 to 100 points, with 100 indicating no pain and full satisfaction and

5 Responsiveness of the NPRS in Shoulder Pain 119 Table 2 Frequency of Diagnoses in the Surgical and Nonsurgical Groups Group Condition Frequency (n) Surgical Nonsurgical rotator-cuff repair 10 labral-tear repair 7 arthroscopic debridement 4 fracture: surgical repair 4 stabilization surgery 3 hemiarthroplasty 2 subacromial decompression 2 total 32 rotator-cuff disease impingement syndrome/partial- to full-thickness rotator-cuff tears, tendinitis 58 frozen shoulder 23 frozen shoulder and rotator-cuff disease 5 instability 5 instability and labral tear 1 labral tear 5 fracture 4 acromioclavicular separation 2 shoulder weakness 1 total 104 function. The PSS scale has demonstrated acceptable test retest reliability (ICC 2,1 =.94, 95%CI =.89,.97) and standard error of the measurement with 90% confidence intervals (SEM 90 ) of 9 points and an MDC with 90% confidence intervals (MDC 90 ) of 12.1 points. The pain section of the PSS contains 3 pain questions, specifically asking for a score on an NPRS rating of pain at rest, with normal daily activities, and with strenuous activities. Each item is rated on an 11-point NPRS, with the anchors of 0 = no pain and 10 = worst possible pain. 9 For the PSS pain subscale, each of the 3 pain items is scored by subtracting the raw score from 10, and then total PSS pain subscale score is calculated by summing the 3 individual pain items for a total pain score of 0 to 30 (30 = no pain). For this study, we summed the 3 NPRS scores from the 3 pain questions from the PSS and then divided by 3 to calculate the NPRS average. The average of the 3 PSS pain items was used for data analysis. Scores for each pain item were labeled as NPRS at rest, NPRS normal activities, and NPRS strenuous activities, and the average as NPRS average. The PSS function subscale contains 20 questions (items) rated on a 4-point Likert scale for level of difficulty (0 = cannot do at all, 1 = much difficulty, 2 = some

6 120 Michener, Snyder, and Leggin difficulty, 3 = no difficulty) for a maximum score of 60 points, which is indicative of full function. There is also an option for did not do before injury for each function item. The function section is scored by totaling the scores of each of the 20 items. If a patient has indicated that he or she did not do an item before injury, then the total score is calculated by subtracting 3 points for every item that was not done before the injury from the maximum 60-point total score (number of didn t do before injury 3 = a; 60 a = b; function total score = total score/b = c, then c 60). The MDC 90 of the PSS function subscale is 8.6 points. 11 Data Analysis Patients were dichotomized as improved or not improved based on the PSS function scores. The PSS function scale was used as the external criterion measure of meaningful change in a patient s status because a change in use of the shoulder in daily activities, work, and recreation was considered a meaningful construct of change. Patients who improved from initial to final on the PSS function subscale by the MDC 90 of were classified as improved, and those with less than 8.6 points change as not improved. The MDC 90 value is an index of the error associated with 2 scores or change scores. Using the MDC 90 of 8.6 provides 90% confidence that a change of more than 8.6 PSS function points is indicative of true improvement and not error. This classification was done for the total group and the 2 subgroups: those who had surgery and those who did not. Means and standard deviations were calculated for NPRS average at initial and final time points, as well as change scores. Responsiveness indices, the MCID, and Cohen s effect sizes were calculated for the NPRS average for all patients and for surgical and nonsurgical subgroups. Receiver-operator-characteristic (ROC) curves were constructed by plotting the sensitivity and specificity values for the NPRS-average change scores to determine the MCID that differentiates patients who improved an important amount from those who did not. The area under the curve (AUC) was used to determine the probability that a patient who was identified as improved was correctly identified by the change in NPRS average. AUC values range from.5 to 1.0; measures with no discriminative ability beyond chance have AUC values of.5, whereas measures with perfect discriminative accuracy have an AUC of It is generally accepted that an AUC value for a test or measure is satisfactory when it exceeds.70, and values above.90 represent high accuracy. 24,27 The MCID was determined by identifying the upper left-hand point on the ROC curve where sensitivity and 1-specificity were maximized and the best cut point for distinguishing improved and unimproved patients was located. 28 Sensitivity and specificity were calculated for each of the selected cutoff scores. Cohen s effect sizes 16 were calculated as group responsiveness index: (mean final NPRS average mean initial-evaluation NPRS average)/ pooled SD. SPSS, version 15 (SPSS Inc, Chicago, IL), was used for data analysis. Results Means and standard deviations were calculated for the pain variables associated with all patients (Table 3) and the surgical and nonsurgical subgroups (Table 4) according to improved and unimproved status. The ranges in scores for the average change in the NPRS average for the improved groups were 2.8 ± 1.7 to 3.7 ±

7 Responsiveness of the NPRS in Shoulder Pain 121 Table 3 Means (SD) for All Patients (n = 136) for the Numeric Pain- Rating Scale (NPRS) Average Improved (n = 93) Not Improved (n = 43) Pain variable IE Final IE Final Rest (0 10; 0 = no pain) 2.8 (2.7) 0.8 (1.5) 2.4 (2.3) 1.3 (2.0) Activities of daily living 5.0 (2.5) 1.4 (1.5) 4.2 (2.5) 2.5 (2.3) (0 10; 0 = no pain) Strenuous activity (0 10; 0 = 7.5 (2.5) 2.8 (2.0) 7.6 (2.4) 4.5 (2.5) no pain) NPRS average (0 10; 0 = no pain) 5.1 (1.6) 1.7 (1.4) 4.7 (1.6) 2.8 (1.9) NPRS-average change from IE to final (0 10; 0 = no pain) 3.4 (1.9), 90%CI = (1.6), 90%CI = IE, initial evaluation. NPRS average = (rest + activities of daily living + strenuous activity)/3. Improved defined as 8.6 points change on Penn Function Subscale, not improved as <8.6 points change on Penn Function Subscale. 1.9 points, and the ranges for the unimproved groups were 1.0 ± 1.8 to 2.1 ± 1.5 points. Figures 1 to 3 represent the ROC curves for all patients and the nonsurgical and surgical subgroups, respectively. The range in AUC values for the NPRS average of the 3 pain questions from initial evaluation to DC was.74 (.65.82) to.76 (.67.86; Table 5). Thresholds for MCID for all groups were 2.17 pain points; corresponding sensitivity and specificity values are reported in Table 5. Cohen s effect size ranged from 1.51 to 1.94 for the 3 groups (Table 5), each greater than 0.8, indicating a large effect. Discussion One challenge with patient self-report outcomes instruments is interpreting their scores. Determining the clinically meaningful change of measures used to assess outcome is critically needed to guide the interpretation of the effectiveness of interventions delivered to individual patients and those in clinical trials. We investigated the responsiveness of the NPRS average, the average of 3 NPRS scores from the 3 PSS pain questions. This study is the first to determine Cohen s effect sizes and MCIDs of the NPRS average in patients with shoulder pain and subgroups of surgical and nonsurgical patients. The effect sizes were all greater than.8, indicating a large effect and, therefore, responsiveness of the NPRS average. The MCID is 2.17 points for any patient presenting with shoulder pain and is also 2.17 NPRS-average points for both subgroups of surgical and nonsurgical patients. Clinically, this MCID value can be used to guide treatment of individual patients. When a patient has a change of more than 2.17 points (eg, a decrease from 9 to 5 points) on the NPRS average over a 3- to 4-week course of treatment, the clinician can be confident that the patient has experienced a change in pain that is meaningful with respect to his or her function. Researchers can use the 2.17 MCID for the NPRS average

8 Table 4 Means (SD) for the Surgical and Nonsurgical Subgroups for the Numeric Pain-Rating Scale (NPRS) Average Nonsurgical Surgical Improved (n = 67) Not Improved (n = 37) Improved (n = 26) Not Improved (n = 6) Pain variable IE Final IE Final IE Final IE Final Rest (0 10; 0 = no pain) 2.7 (2.7) 0.5 (0.9) 2.3 (2.2) 1.2 (1.7) 3.0 (2.6) 1.5 (2.3) 2.8 (2.9) 2.2 (3.3) Activities of daily living (0 10; 0 = no pain) 5.0 (2.6) 1.1 (1.3) 4.1 (2.3) 2.5 (2.0) 5.1 (2.5) 2.2 (1.8) 5.0 (3.8) 3.1 (3.8) Strenuous activity (0 10; 0 = no pain) 7.4 (2.6) 2.4 (2.0) 7.8 (2.3) 4.4 (2.3) 7.6 (2.2) 3.7 (1.8) 6.2 (2.2) 5.7 (3.4) NPRS average (0 10; 0 = no pain) 5.1 (1.6) 1.4 (1.2) 4.7 (1.6) 2.7 (1.7) 5.2 (1.6) 2.5 (1.7) 4.7 (1.8) 3.6 (3.1) NPRS-average change from IE to final (0 10; 0 = no pain) 3.7 (1.9), 90%CI = (1.5), 90%CI = (1.7), 90%CI = (1.8), 90%CI = 0.5 to 2.5 IE, initial evaluation. NPRS average = (rest + activities of daily living + strenuous activity)/3. Improved defined as 8.6 points change on Penn Function Subscale, not improved as <8.6 points change on Penn Function Subscale. 122

9 Responsiveness of the NPRS in Shoulder Pain 123 Table 5 Responsiveness Characteristics of the Change in Numeric Pain-Rating Scale (NPRS) Average From Initial Evaluation to Final (0 10; 0 = no pain) Group MCID AUC (95% CI) Sn Sp Cohen effect size All patients (.65.82)** Nonsurgical (.67.86)** Surgical (.55.95)* MCID, minimal clinically important difference; AUC, area under the curve; Sn, sensitivity; Sp, specificity. Cohen effect size = (mean final NPRS average mean initial NPRS average)/sd pooled. * P =.06. **P < Figure 1 Receiver-operating-characteristic curves for the average change in numeric pain-rating scale (NPRS) for all subjects. The circled value represents the minimal clinically important difference for the NPRS in patients with shoulder pain. for shoulder pain to determine whether the results of a clinical trial are clinically meaningful, and not just statistically significant. 29 Change that is meaningful to the patient should be the value used to interpret treatment effect for individual patient care and for studies investigating treatment effectiveness.

10 Figure 2 Receiver-operating-characteristic curves for the average change in numeric painrating scale (NPRS) for the nonsurgical subgroup. The circled value represents the minimal clinically important difference for the NPRS in patients with shoulder pain who have not had surgery. Figure 3 Receiver-operating-characteristic curves for the average change in numeric painrating scale (NPRS) for the surgical subgroup. The circled value represents the minimal clinically important difference for the NPRS in patients with shoulder pain who have had surgery. 124

11 Responsiveness of the NPRS in Shoulder Pain 125 Using ROC-curve analysis, the MCID for the NPRS average was 2.17 for all patients and the subgroups of surgical and nonsurgical diagnoses. All 3 ROC-curve analyses had acceptable levels of diagnostic accuracy (sensitivity.69.83, specificity.60.83) and demonstrated the ability to discriminate between improved and unimproved patients 74% to 76% of the time using a change of 2.17 NPRS-average points. For the surgical group, the AUC s significance level fell below the typical cutoff value of.05, with a probability value of.06. However, the AUC value for the surgical group was.75, with 95%CI of.55 to.95. The confidence bounds for the AUC for the surgical group are wide, which is likely because of a small sample of 6 subjects in the unimproved group. Given the similarities of the diagnostic characteristics of the 3 groups, the MCID of 2.17 can be used, but with less confidence, to distinguish improved versus unimproved surgical patients. In a prior study by Mintken et al, 10 the MCID for a different NPRS average of 3 pain questions in patients with shoulder pain was 1.1 points. This NPRS average was calculated from the 3 pain questions of average pain, worst pain, and best pain, which were different questions than those used in our study. Patients in the prior study by Mintken et al were receiving outpatient rehabilitation, with the time interval between data-collection points of 2 to 4 weeks. This interval of treatment is comparable to that of our study of 3 to 4 weeks, but Mintken et al used the GRC as the anchor to classify patients as improved and stable. As previously described, the GRC is a transitional scale that is subject to recall bias We used a change in function on the PSS function subscale, specifically the MDC value of 8.6 points, as the external criterion measure of change. Using an external criterion for change based on a measure of current health status reduces the likelihood of inaccurate reflection and recall bias regarding a change in health status. In addition, our decision to use an external criterion based on function was based on a theory that when there is a change in function greater than error, the patient is improving. The discrepancies in defining when meaningful change has occurred and the use of 3 different pain questions for calculating the Mintken et al 10 defined NPRS average may account for the different MCIDs of Mintken et al 10 (1.1 points) and our study (2.17 points).an interesting finding of this study is that the MCID for the surgical and nonsurgical subgroups was the same, 2.17 points, indicating that meaningful change in pain, when measured with the NPRS average, is perceived the same, regardless of surgical status. We hypothesized that the MCID value would be greater in the surgical group than the nonsurgical group because of a likely higher initial pain intensity in the surgical group. 25 Baseline scores for the surgical group had an NPRS average of 5.1 and the nonsurgical group a 4.9, indicating slightly higher pain for the surgical group. For the surgical subgroup, there was an overlap in 90% confidence intervals for the improved and unimproved NPRS-average change scores, indicating a limited ability of the MCID for the NPRS average to differentiate between those who improve and those who do not in the surgical group. The 11-point NPRS is a common method of assessing pain in patients being treated in outpatient rehabilitation centers for musculoskeletal disorders. Prior studies have reported MCID values for NPRS or NPRS average derived from of a variety of pain conditions ranging from 1.0 to 2.76 points in these settings over a treatment interval ranging from 1 to 4 weeks. 10,22 24 However, other studies have reported a range of MCID values of 0.5 to 4.5 with intervals of measurement from 2 days to 3 months. 10,22 24,27,30 With different time intervals, and hence different periods of change, the MCID may vary.

12 126 Michener, Snyder, and Leggin The MCID can be affected by a variety of factors that include, but are not limited to, time interval of change, demographics of patients, type of treatment, types of questions used to rate pain with an NPRS, the external criterion measure used to define change, and the definition of change for the external criterion measure. 14 Of these, the definition of the amount of change is critical because the wording used to represent clinically important change can vary the MCID value. For example, definitions for meaningful change have included minimal change versus somewhat different versus a great deal different. These different definitions of change can produce different values of MCID. As a result, MCID and effect size should not be viewed as single static values. The MCID is not a single value (eg, 2.17) for shoulder pain but rather a range of values dependent on factors that influence the measurement of the MCID. There are limitations to this study. The MCID and effect-size values in this study are generalizable to patients undergoing outpatient rehabilitation for surgical or nonsurgical shoulder pain, evaluating change using the 3 pain questions in this study for the NPRS average after 3 to 4 weeks of rehabilitation. In addition, the external criterion measure of change used in this study was MDC for the PSS function subscore, and although based on theory, it is not the only external criterion that could be used. Shorter or longer intervals of time between measurements of pain or the use of a different criterion measure of change could result in lower or higher MCID values. In addition, we do not know whether MCID and effect-size values would differ across other variables such as ethnicity. Finally, generalizability of values for the surgical group may be limited by our small sample size in the unimproved group and overlap of the confidence bounds of the NPRS-average change scores for improved and unimproved groups. Future studies should continue the investigation of MCID for pain in patients with shoulder pain, because the MCID may be higher or lower in a different mix of shoulder-pain diagnoses, treatment settings, and time intervals of change. Moreover, further investigation of meaningful change in pain for patients undergoing surgery is warranted. Investigations using different anchors of change are needed. Further studies will improve interpretability of pain scores in patients with shoulder pain. Conclusions The NPRS average as calculated in this study as an average of 3 questions of pain at rest, pain with normal activities, and pain with strenuous activities using the external criterion of change on the PSS function subscale is responsive in patients with shoulder pain who are undergoing rehabilitation for 3 to 4 weeks. A large effect was demonstrated with Cohen s effect size for all patients and for the subgroups of surgical and nonsurgical patients. A 2.17-point change in the NPRS average is clinically meaningful. A change in NPRS average of 2.17 can discriminate between patients who have improved and those who have not improved in shoulder function in all patients, as well as surgical and nonsurgical groups. Our study has contributed to the interpretability of the NPRS average based on the 3 pain items from the PSS, but not to the interpretability of the PSS pain subscale. The PSS subscale is the sum of the 3 pain items. This is the first study to examine the responsiveness of the NPRS average, as defined in this study, in patients with

13 Responsiveness of the NPRS in Shoulder Pain 127 shoulder pain and subgroups of surgical and nonsurgical diagnoses of shoulder pain. These findings will facilitate the interpretation of NPRS-average scores for clinicians and researchers. Acknowledgments This article is part of a special issue of JSR on clinical outcomes assessment. References 1. Reilingh ML, Kuijpers T, Tanja-Harfterkamp AM, van der Windt DA. Course and prognosis of shoulder symptoms in general practice. Rheumatology (Oxford). 2008;47(5): Kuijpers T, van der Heijden GJ, Vergouwe Y, et al. Good generalizability of a prediction rule for prediction of persistent shoulder pain in the short term. J Clin Epidemiol. 2007;60(9): Kuijpers T, van der Windt DA, van der Heijden GJ, Bouter LM. Systematic review of prognostic cohort studies on shoulder disorders. Pain. 2004;109(3): Kennedy CA, Manno M, Hogg-Johnson S, et al. Prognosis in soft tissue disorders of the shoulder: predicting both change in disability and level of disability after treatment. Phys Ther. 2006;86(7): Croft P, Pope D, Silman A. The clinical course of shoulder pain: prospective cohort study in primary care. Primary Care Rheumatology Society Shoulder Study Group. BMJ. 1996;313(7057): van der Windt DA, Koes BW, Boeke AJ, Deville W, de Jong BA, Bouter LM. Shoulder disorders in general practice: prognostic indicators of outcome. Br J Gen Pract. 1996;46(410): Bot SD, van der Waal JM, Terwee CB, et al. Predictors of outcome in neck and shoulder symptoms: a cohort study in general practice. Spine (Phila Pa 1976). 2005;30(16):E459 E Thomas E, van der Windt DA, Hay EM, et al. Two pragmatic trials of treatment for shoulder disorders in primary care: generalisability, course, and prognostic indicators. Ann Rheum Dis. 2005;64(7): Jensen MP, Karoly P, Braver S. The measurement of clinical pain intensity: a comparison of six methods. Pain. 1986;27(1): Mintken PE, Glynn P, Cleland JA. Psychometric properties of the shortened disabilities of the Arm, Shoulder, and Hand Questionnaire (QuickDASH) and Numeric Pain Rating Scale in patients with shoulder pain. J Shoulder Elbow Surg. 2009;18(6): Leggin BG, Michener LA, Shaffer MA, Brenneman SK, Iannotti JP, Williams GR Jr. The Penn Shoulder Score: reliability and validity. J Orthop Sports Phys Ther. 2006;36(3): Guyatt G, Walter S, Norman G. Measuring change over time: assessing the usefulness of evaluative instruments. J Chronic Dis. 1987;40: Angst F, Aeschlimann A, Michel BA, Stucki G. Minimal clinically important rehabilitation effects in patients with osteoarthritis of the lower extremities. J Rheumatol. 2002;29(1): Beaton DE, Bombardier C, Katz JN, Wright JG. A taxonomy for responsiveness. J Clin Epidemiol. 2001;54(12): Jaeschke R, Singer J, Guyatt GH. Measurement of health status. ascertaining the minimal clinically important difference. Control Clin Trials. 1989;10(4): Norman GR, Wyrwich KW, Patrick DL. The mathematical relationship among different forms of responsiveness coefficients. Qual Life Res. 2007;16(5):

14 128 Michener, Snyder, and Leggin 17. Ostelo RW, Deyo RA, Stratford P, et al. Interpreting change scores for pain and functional status in low back pain: towards international consensus regarding minimal important change. Spine (Phila Pa 1976). 2008;33(1): Norman GR, Stratford P, Regehr G. Methodological problems in the retrospective computation of responsiveness to change: the lesson of Cronbach. J Clin Epidemiol. 1997;50(8): Guyatt GH, Osoba D, Wu AW, Wyrwich KW, Norman GR. Methods to explain the clinical significance of health status measures. Mayo Clin Proc. 2002;77(4): Beaton DE, Tarasuk V, Katz JN, Wright JG, Bombardier C. Are you better? a qualitative study of the meaning of recovery. Arthritis Rheum. 2001;45(3): Schmitt J, Di Fabio RP. The validity of prospective and retrospective global change criterion measures. Arch Phys Med Rehabil. 2005;86(12): Grotle M, Brox JI, Vollestad NK. Concurrent comparison of responsiveness in pain and functional status measurements used for patients with low back pain. Spine (Phila Pa 1976). 2004;29(21):E492 E Farrar JT, Young JP, Jr, LaMoreaux L, Werth JL, Poole RM. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain. 2001;94(2): Childs JD, Piva SR, Fritz JM. Responsiveness of the Numeric Pain Rating Scale in patients with low back pain. Spine (Phila Pa 1976). 2005;30(11): Cook KF, Roddey TS, Olson SL, Gartsman GM, Valenzuela FF, Hanten WP. Reliability by surgical status of self-reported outcomes in patients who have shoulder pathologies. J Orthop Sports Phys Ther. 2002;32(7): Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143(1): Salaffi F, Stancati A, Silvestri CA, Ciapetti A, Grassi W. Minimal clinically important changes in chronic musculoskeletal pain intensity measured on a numerical rating scale. Eur J Pain. 2004;8: Stratford PW, Binkley FM, Riddle DL. Health status measures: strategies and analytic methods for assessing change scores. Phys Ther. 1996;76(10): Man-Son-Hing M, Laupacis A, O Rourke K, et al. Determination of the clinical importance of study results. J Gen Intern Med. 2002;17(6): Cleland JA, Childs JD, Whitman JM. Psychometric properties of the Neck Disability Index and Numeric Pain Rating Scale in patients with mechanical neck pain. Arch Phys Med Rehabil. 2008;89(1):69 74.