Background and Purpose: The purpose of this study was to determine whether physical therapy interventions predicted meaningful short-term improvement in 4 measures of physical health, pain, and function for patients diagnosed with adhesive capsulitis.
Participants: Data were examined from 2,370 patients (mean age=55.3 years, SD=12.4; 65% female, 35% male) classified into ICD-9 code 726.0 who had completed an episode of outpatient physical therapy.
Methods: Principal components factor analysis was used to define intervention categories from specific treatments applied during the episode of care. A nested logistic regression model was used to identify intervention categories that predicted a 50% or greater change in Physical Component Summary-12 (PCS-12), physical function (PF), bodily pain (BP), and hybrid function (HF) scores.
Results: None of the patients achieved a 50% or greater improvement in PCS-12 scores. Improvement in BP scores was more likely in patients who received joint mobility interventions (odds ratio=1.35, 95% confidence interval=1.10–1.65). Improvement in HF scores was more likely in patients who received exercise interventions (odds ratio=1.50, 95% confidence interval=1.03–2.17). Use of iontophoresis, phonophoresis, ultrasound, or massage reduced the likelihood of improvement in these 3 outcome measures by 19% to 32%.
Limitations: The authors relied on clinician-identified ICD-9 coding for the diagnosis. Impairment measures were not available to support the diagnosis, and some interventions were excluded because of infrequent use by participating therapists.
Discussion and Conclusions: These results are consistent with findings from randomized clinical trials that demonstrated the effectiveness of joint mobilization and exercise for patients with adhesive capsulitis. Ultrasound, massage, iontophoresis, and phonophoresis reduced the likelihood of a favorable outcome, which suggests that use of these modalities should be discouraged.
Adhesive capsulitis, or “frozen shoulder,” is a soft tissue disorder that results in pain, stiffness, and progressive loss of active and passive range of motion (AROM and PROM) in the glenohumeral joint.1,2 The patient frequently has difficulty dressing, grooming, and performing overhead reaching activities for a period of several months to several years.3 The disorder predominantly affects women aged 40 to 60 years and occurs in roughly 2% to 4% of the adult population.1,4–6 In addition to age and sex, risk factors commonly identified include diabetes, cervical disk disease, immobilization of the shoulder, shoulder trauma, cardiovascular and pulmonary disease, hyperthyroidism, and autoimmune diseases.2,4,7
Three distinct clinical stages with overlapping timelines have been proposed: (1) an acute stage, with painful shoulder motion and sleep interruption; (2) a frozen stage, characterized by stiffness, reduced pain, and loss of joint motion, and; (3) a thawing stage, with resolution of pain and gradual recovery of joint motion.8 Left untreated, adhesive capsulitis appears to be a self-limiting, but frequently disabling, disorder that lasts an average of 3 years. The economic impact of this protracted condition has been estimated in randomized controlled trials (RCTs) in Europe and Australia.9,10 When converted to US dollars, the total combined annual health care and non–health care costs are approximately $7,000 to 8,000 per episode. Despite the individual and societal burden associated with adhesive capsulitis, relatively little is known about which nonsurgical interventions increase or decrease the likelihood of a successful outcome.
Management options for adhesive capsulitis include oral corticosteroids or glenohumeral joint corticosteroid injection,6,11–18 manipulation under anesthesia,1 arthroscopic capsular release,19 and a variety of interventions provided by physical therapists. Physical therapy interventions most commonly described include therapeutic exercise, manual therapy techniques such as joint mobilization and manipulation, electrotherapeutic modalities, and thermal modalities.10,11,13–17,20–22 Green and colleagues’ meta-analysis of clinical trials evaluating the effectiveness of physical therapy interventions for shoulder pain revealed beneficial effects from laser therapy for patients with adhesive capsulitis or rotator cuff pathology.23 Exercise also produced positive outcomes in patients with “mixed shoulder disorders.” Ultrasound, however, provided no additional benefit, and there was insufficient evidence to support or refute the effectiveness of physical therapy alone for adhesive capsulitis. Most of the studies included in the review were methodologically weak, had small sample sizes, and did not investigate combinations of interventions. The variability in diagnostic classification across studies also made it difficult to isolate treatment effects for patients with confirmed adhesive capsulitis.
Four RCTs published since Green and colleagues’ systematic review23 examined different combinations of physical therapy interventions in people with adhesive capsulitis. Guler-Uysal and Kozanoglu20 randomly assigned 40 patients with idiopathic adhesive capsulitis (35% painful stage, 65% frozen stage) to a deep friction massage and Cyriax mobilization group or a heat modality and therapeutic exercise group. Ninety-five percent of the participants who received manual therapy techniques achieved 80% of normal PROM and greater pain reduction after 2 weeks of treatment compared with 65% of those who received modalities and exercise. The authors did not assess functional change and did not repeat the outcome measures beyond the 2-week treatment period.
Buchbinder et al10 evaluated the effectiveness of a combination of manual therapy techniques and exercise compared with sham ultrasound (placebo) in 144 patients who had adhesive capsulitis for an average of 6 months. Both groups received 8 treatment sessions over 6 weeks following arthrographic glenohumeral joint distension. The experimental group achieved a statistically significant improvement in AROM (ie, an increase in active shoulder abduction of 10.6°, P=.006) and patient-perceived improvement (relative risk=1.4, 95% confidence interval [CI]=1.1–1.7) that were sustained over the 6-month follow-up period. However, no differences were detected between the groups with respect to pain, function, or quality of life at 6, 12, and 26 weeks.
Two RCTs21,22 compared the effectiveness of specific joint mobilization techniques in patients with adhesive capsulitis who had documented PROM deficits. Both studies showed improvements in PROM for either Maitland grade III or IV procedures (ie, an increase of 36.3°-46.3° of passive abduction)21 or with posteriorly directed Kaltenborn grade III procedures (ie, an increase of 31.3° of passive external rotation).22 However, subsequent changes in pain and function were comparable to those of the respective comparison groups. In addition, the improvement in functional outcomes in Vermeulen and colleagues’ subjects occurred months after treatment concluded.21
The 4 RCTs10,21–22 examined different types and combinations of treatments over different time periods and used a variety of self-report instruments to assess pain, function, and quality of life. Two of the studies used small samples with a wide range of symptom duration. As a result, it is unclear which interventions may be most effective for patients with adhesive capsulitis. In addition, we found no cohort studies evaluating outcomes of physical therapy interventions for this disorder in large samples of subjects. Longitudinal outcome studies have limitations but can provide insight into the usefulness of interventions delivered in “real-world” clinical conditions, the influences of which are purposefully constrained in RCTs. Large sample cohort studies also have the advantage of comparing outcomes for many combinations of interventions, unlike RCTs, which typically compare a limited number of interventions. Predictive models using large cohorts may identify convincing relationships between specific interventions and patient outcomes and help narrow the focus of future effectiveness and efficacy studies. Therefore, the purpose of our study was to determine whether physical therapy interventions, either in isolation or in combination, predicted meaningful short-term change in pain and physical health for a large cohort of patients diagnosed with adhesive capsulitis of the shoulder.
Focus On Therapeutic Outcomes Inc (FOTO),* a private rehabilitation outcomes management company, provided data for this study. Clinics contracting with FOTO during the study used standardized questionnaires24 with several demographic and disorder-related questions as well as the Physical Component Summary-12 (PCS-12).25 The PCS-12 is a 12-item generic health measure derived from the Physical Component Summary of the Medical Outcomes Study 36-Item Short-Form Health Survey questionnaire (SF-36). Transformed scores on the PCS-12 range from 0 to 100, with 50 representing the mean score for the population of adults in the United States. Higher scores indicate less physical disability.26 Physical function (PF) and bodily pain (BP) subscale scores also can be calculated from this survey.
Several authors have examined the usefulness of self-report health status instruments in patients with shoulder disorders. Gartsman et al27 compared the SF-36 scores of 100 patients diagnosed with adhesive capsulitis. The average Physical Component Summary (PCS) score was 37.6 (SD=8.8), and the mean PF and BP subscale scores were 67.2 (SD=22.4) and 37.6 (SD=20.6). However, the authors did not evaluate responsiveness to change of the composite or subscale scores in this patient group. Beacon and Richards28 found the SF-36 PCS and PF subscale scores to be less responsive to change compared with shoulder-specific health status instruments in 99 patients undergoing total shoulder arthroplasty or rotator cuff surgery. The BP subscale, however, demonstrated a standardized response mean comparable to those of the other surveys. Schmitt and Di Fabio29 reported comparable findings regarding the responsiveness of the PCS-12 in 58 patients with a variety of “proximal UE [upper-extremity] musculoskeletal disorders.” Taken together, these studies27–29 suggest that generic measures such as the 36-Item Short-Form Health Survey questionnaire (SF-12) or SF-36 may require supplementation by region-specific measures in order to detect change in health status in patients with adhesive capsulitis.
Setting and Participants
The initial sample consisted of 3,383 patients who received outpatient physical therapy services for signs and symptoms that were classified by ICD-9 code 726.0 as adhesive capsulitis30 between January 1, 1998, and December 31, 2000. All patients were at least 18 years of age and did not have secondary diagnostic or procedural ICD-9 codes related to the shoulder. These criteria were selected to ensure that our sample consisted of adults in whom we could confirm that the interventions provided were only for adhesive capsulitis. Consenting patients completed the FOTO questionnaires at intake and at discharge from physical therapy. Therapists were not informed of their patients’ initial or discharge PCS-12 scores. Patients were excluded if they did not complete their episode of care (n=832), if outcome scores were missing (n=101), or if categorical or ordinal variables were coded with scores not included among the survey options (n=80).
All FOTO data were masked to prevent recognition of individual patients, clinicians, or clinics. FOTO's research oversight committee approved the use of these data following review of the study proposal.
The subjects were 2,370 patients coded as having adhesive capsulitis who received outpatient physical therapy services from a specific admission date to a specific discharge date within the study time frame. The majority of the patients were female (65.4%), with a mean age of 55.3 years (SD=12.4). Information regarding patient race or ethnicity was not available in the database. On admission, the mean PCS-12 score was 39.6 (SD=9.2), the mean PF score was 69.1 (SD=22.3), and the mean BP score was 43.1 (SD=17.7). These scores are comparable to those reported by Gartsman et al27 using the SF-36. Patients included in the analysis were slightly older, averaged 2 more visits, were less likely to be working, and had slightly lower overall mental health scores (Mental Component Summary-12 [MCS-12]) than those excluded according the study criteria (Tab. 1).
A total of 1,175 physical therapists with an average of 8.6 years (SD=6.9) of clinical experience participated in the study (Tab. 2). Three percent of the therapists reported that they were clinical specialists in orthopedic physical therapy, as defined by the American Board of Physical Therapy Specialties, or specialists in manual therapy, as defined by the American Academy of Orthopaedic Manual Physical Therapists.
When a patient was discharged, the physical therapist completed a form that indicated the interventions provided over the episode of care. The therapist checked relevant items from a list of 60 treatment options. Definitions for each treatment are provided in the FOTO training manual.31 We included 21 interventions that were provided to at least 5% of the patients in the study (eTab. 1). We reasoned that interventions provided less frequently would not be represented adequately in the data set in order to estimate their relationship to outcomes with any precision. Therapists used an average of 7.88 (SD=2.8) different interventions over an episode of care. We used factor analysis to group interventions to more accurately reflect the multidimensional nature of the plan of care. We have described this analytic approach in previous work using FOTO data for patients with sciatica.32
Seven factors explaining 48% of the common variance were extracted in the factor analysis. Factors 1 through 6 each contained at least 2 interventions that met the ≥0.500 inclusion criterion recommended by Sharma.33 We applied the following category labels based on the definitions of the interventions (eTab. 2) in each of the groups: “postural correction and stabilization,” “joint mobilization and mobility,” “ultrasound and massage,” “exercise,” “ice and electrical stimulation,” and “iontophoresis and phonophoresis.” Factor 7 was defined by only 1 intervention (moist heat) and was labeled accordingly. These treatment groups are similar to those previously reported and appear to describe combinations of interventions with clinically sensible bases. Five of the 21 interventions did not meet the ≥0.500 criterion and, therefore, were excluded from the analysis: closed-chain exercises (−0.376), endurance exercises (0.471), flexibility exercises (0.494), pain modulation (0.470), and myofascial release (0.429). The intervention categories and relevant factor loading scores obtained in the analysis are summarized in eTable 3.
Dependent and Independent Variables
We evaluated change in 4 dependent variables: PCS-12 score, PF score, BP score, and a hybrid function score (HF). We used the composite physical health score (PCS-12) because Gartsman et al27 demonstrated that patients with adhesive capsulitis have reduced PCS-36 scores relative to the adult population. We included the 2 subscale scores (PF and BP) because we could not determine the stage of patients’ adhesive capsulitis and, therefore, did not know whether primary impairments for individual patients were related to pain, loss of range of motion (ROM), or both. We reasoned that the PF and BP scores may be more responsive to change for patients in the “frozen” and “painful” stages, respectively.
The HF score was created for this study to provide an additional measure of functional status that was theoretically more sensitive to change than the PCS-12 or PF scores. We developed this score using exploratory factor analysis of FOTO physical health survey items that we hypothesized were most related to shoulder function (Tab. 3). We anticipated that this hybrid score might be sensitive to change because it more specifically focused on upper-extremity function similar to shoulder-specific self-report instruments evaluated by other authors.28,29
We defined clinically meaningful improvement as an increase of 50% or more in a dependent variable. We chose this criterion over other options such as the minimum detectable change (MDC) or minimal clinically important difference (MCID) because these thresholds have not been established for the PCS-12 and related subscales in patients with adhesive capsulitis. Schmitt and Di Fabio29 calculated MDC and MCID thresholds for PCS-12 scores in patients with proximal UE conditions, and these thresholds were smaller than our 50% improvement criterion; therefore, we acknowledge that our selection of a 50% threshold may be too conservative. Some of the patients in our study may have changed by less than 50% but still had clinically meaningful improvement. However, given the protracted nature of adhesive capsulitis, we feel confident that a 50% improvement reflects meaningful change in individuals with the disorder. Data in our study were coded as “0” when less than 50% improvement was found and “1” when 50% or greater improvement was found.
The primary independent variables in our study were the 7 intervention groups derived from the factor analysis (eTab. 3). We scored the factor as present when a patient received 50% or more of the interventions represented by the factor. Each factor was scored as absent if a patient received less than 50% of the interventions represented by the factor. We reasoned that when at least half of the interventions for a factor were provided to a patient, the factor was adequately represented.
Potential collinearity among the demographic variables was assessed using Pearson product moment and Phi statistics. Analysis of the intake scores for the dependent variables indicated that a proportion of patients (PF=53.5%, HF=48%, BP=10.3%, PCS-12=0%) were too close to the upper limit of the scale (“no disability”) to demonstrate a 50% or greater improvement in scores. We eliminated this ceiling effect by transforming the intake and discharge scores for all of the dependent variables so that the lowest value on the scale equaled “no disability” and the highest value on the scale equaled “maximum disability.” A meaningful improvement, therefore, was redefined as a 50% or greater decrease in each of the disability scores.
We used multiple logistic regression to identify the treatment variables that independently predicted meaningful clinical improvement. In order to account for variance from multiple levels of data, we used a model that nested patient data within therapists. We were unable to perform a 3-level analysis because the majority of clinics had only one physical therapist. To protect against violation of assumptions of the model, we used a robust variance estimation method.34 The calculated intra-cluster correlations for this study ranged from .0578 to .1414; these values represent the proportion of total variation in meaningful improvement in the dependent variables that can be attributed to the differences among therapists. The probabilities for entry of the intervention variables into and removal from the model were set at .20 and .10, respectively. A P value ≤.05 for the Wald test was used to determine the statistical significance of the predictor variables. To adjust for confounders, patient age and sex, length of time since onset of condition (<90 days), employment status (working), use of prescription medications for the condition, number of visits, intake mental health composite (MCS-12) score, and scores on admission for physical therapy for each of the dependent variables were forced into each model. We tested for all 2-way interactions of any independent variables included in the final models. Initial descriptive and correlational statistical analyses were conducted using SPSS 16.0 software for Windows.† Logistic regression analysis was generated using SAS 9.1.3,‡ a statistical software program designed to analyze cluster-correlated data like that used in this study.
All but one pair of subject demographic characteristics were weakly correlated to one another, with coefficients ranging from .000 to .102. Patient age and working status demonstrated a statistically significant relationship (point biserial coefficient=−.493, P<.01). Although this value represents a moderate correlation, we decided to enter both variables into the analysis because we anticipated that they potentially would influence outcomes differently.
The following proportions of patients achieved clinically meaningful improvement in the dependent variables: PCS-12=0%, PF=43.8% (n=1,037), BP=37.8% (n=895), and HF=40.4% (n=958). We eliminated the PCS-12 score from further analysis as a result of these findings.
Two independent variables reduced the odds of meaningful improvement in PF score: “iontophoresis and phonophoresis” (odds ratio [OR]=0.76, 95% confidence interval [CI]=0.60–0.96) and “ultrasound and massage” (OR=0.77, 95% CI=0.65–0.91). None of the intervention categories increased the odds of meaningful improvement in this measure (Tab. 4).
Application of “iontophoresis and phonophoresis” and “ultrasound and massage” interventions also reduced the odds of meaningful improvement in the BP score. The use of “joint mobilization and mobility” interventions, however, increased the odds of meaningful improvement in this measure (OR=1.35, 95% CI=1.10–1.65) (Tab. 5).
Similar results occurred with respect to the “iontophoresis and phonophoresis” and “ultrasound and massage” intervention categories in the model for predicting meaningful improvement in the HF score. However, application of the “exercise” intervention category increased the odds of improvement in this measure by 50% (OR=1.50, 95% CI=1.03–2.17) (Tab. 6).
We did not identify statistically significant interactions among independent variables in any of the models. Model fit using the likelihood ratio test indicated that the models with covariates fit the data better than the models with the intercept only (P<.0001).
Two intervention categories—“iontophoresis and phonophoresis” and “ultrasound and massage”—remained in all 3 regression models (P<.05) after adjusting for patient demographic and clinical characteristics. Odds ratios indicated that the use of these types of intervention reduced the likelihood of improvement in the dependent variables by 19% to 32%. Alternatively, the presence of 2 intervention categories—“joint mobilization and mobility” and “exercise”—increased the odds of a successful outcome in the BP and HF models, respectively. For example, after adjusting for confounders, the odds of a meaningful change in BP score were 34% higher in patients who received joint mobility interventions compared with patients who did not receive these interventions. The odds of a comparable change in HF score also were 50% greater in patients who received exercise interventions compared with those who did not receive these interventions.
We conducted a post hoc chi-square analysis to determine whether a difference in intervention use existed between subjects who demonstrated meaningful improvement in outcomes and subjects who did not demonstrate improvement. Fewer subjects with improved PF scores received “ultrasound and massage” interventions compared with those whose PF scores did not meet the threshold for improvement (48.4% vs 55.8%, P<.001). A greater proportion of subjects whose BP scores improved received “joint mobilization and mobility” interventions (73.5% vs 68.8%, P=.018). Fewer of these subjects, however, received “iontophoresis and phonophoresis” interventions (13.5% vs 17.4%, P=.015). Finally, more subjects with improved HF scores received “exercise” intervention (95.1% vs 92.9%, P=.037), whereas fewer of these subjects received “ultrasound and massage” interventions compared with subjects who did not achieve meaningful improvement (49.2% vs 54.8%, P=.006).
Our study suggests that the use of physical therapy interventions to administer anti-inflammatory or other agents across the skin (eg, iontophoresis and phonophoresis) decreases the likelihood of meaningful clinical improvement in pain or physical health in patients with adhesive capsulitis. We found 2 RCTs that examined the effectiveness of iontophoresis in patients with calcific tendinitis of the shoulder.35,36 Both studies demonstrated no added benefit compared with a placebo treatment in terms of pain, disability, or size of the calcium deposits. We found no studies that investigated the effectiveness of phonophoresis for any disorders of the shoulder. Our results extend the implications of this limited work by suggesting that application of these passive modalities may actually hinder recovery when emphasized over other treatment options such as joint mobilization and mobility.
The application of therapeutic ultrasound or massage also reduced the odds of meaningful improvement in pain or physical function. Dogru et al37 randomly assigned 49 patients with adhesive capsulitis of 5 to 6 months’ duration (range=3–12) to an ultrasound group or a sham ultrasound group. Both groups also received moist hot packs along with stretching and AROM exercises. These authors found that both groups improved similarly over time on the visual analog scale (VAS) for pain, the Shoulder Pain and Disability Index (SPADI), and the SF-36. Lack of effectiveness of therapeutic ultrasound also has been demonstrated in RCTs enrolling patients with other disorders of the shoulder.38,39 Studies evaluating the benefits of massage are limited in size and quality and provide conflicting results in mixed populations with general cervical or shoulder pain.40,41 Time spent using these interventions in patients with adhesive capsulitis may delay meaningful improvement in symptoms and functional abilities.
We identified 2 treatment categories that increased the likelihood of meaningful improvement in patients with adhesive capsulitis. The “joint mobilization and mobility” intervention category consists of manual joint mobilization techniques as well as activities the patient performs to improve joint mobility. Our findings are consistent with those of studies examining the effectiveness of interventions focusing on improving joint movement in this patient population. Vermeulen et al21 found that patients with adhesive capsulitis who received “high-grade” or “low-grade” Maitland joint mobilization techniques reported reductions in VAS pain scores at rest, with movement, and during sleep. Unlike our findings, these subjects also demonstrated a 13- to 14-point improvement in their PCS-36 scores following treatment. However, the experimental group averaged 18.6 treatment sessions, and the comparison group averaged 21.5 treatment sessions. Our subjects averaged only 11 treatment sessions. Although 71% of our sample received this intervention category (eTab. 3), it is impossible to determine whether mobilization or joint mobility was applied during every session. Johnson et al22 demonstrated comparable reductions in pain and mixed success with physical function following 6 treatment sessions of anterior or posterior glide joint mobilization techniques for patients with this condition. Our results, in conjunction with these findings, suggest that the effect of joint mobility techniques on physical function in people with adhesive capsulitis may be dependent upon the number of times these interventions are applied.
The results of this study also indicate that exercise interventions improve the likelihood of a meaningful clinical improvement in physical health, as measured by the HF score (OR=1.5). The “exercise” category consisted of strengthening, stretching, and home programs. None of the RCTs we found evaluated the effectiveness of exercise techniques alone in patients with adhesive capsulitis. Two studies examined the impact of physical therapy interventions that included exercise on self-report measures of pain and disability comparable to those used in our study. Buchbinder et al10 found that patients who received 8 sessions of passive stretching and active strength and coordination exercises, along with shoulder and spine mobilization techniques, demonstrated improvement over time similar to the comparison group in pain and disability, as measured by the VAS pain scale, the SF-36, and the SPADI. These results suggest that the combination of physical therapy interventions did not provide additional benefit beyond the capsular distension technique for these 3 self-report measures.
Carette et al16 randomly assigned 93 patients with adhesive capsulitis to 1 of 4 groups: corticosteroid injection plus physical therapy, corticosteroid injection alone, physical therapy plus saline injection, or saline injection alone (placebo). Physical therapy interventions consisted of electrical or thermal modalities, AROM and PROM exercises, joint mobilization techniques, and a home program. After 12 sessions, the patients in the corticosteroid injection plus physical therapy group and the corticosteroid injection alone group demonstrated a statistically significant improvement in their SPADI composite, pain, and disability scores compared with patients in the physical therapy plus saline injection and placebo groups. The corticosteroid injection plus physical therapy group also achieved greater improvement in joint AROM and PROM compared with the other 3 groups. Contrary to Buchbinder and colleagues’ findings,10 these results indicate that a combination of physical therapy interventions may provide additional benefit when applied along with a corticosteroid injection. However, the effect of the exercise interventions in either study cannot be isolated.
We used 4 self-report outcome measures in this study because we anticipated that they might differ in their abilities to detect clinically meaningful change. The consistency of results regarding the role of passive modalities that may not address the most impaired tissues is particularly compelling. The finding that different intervention categories improved the odds of meaningful improvement in pain compared with function is an important consideration for future effectiveness studies. None of our sample, however, achieved a 50% or greater change score on the PCS-12. One potential explanation for this result is that this composite score consists of items related to lower-extremity function and general health, neither of which may be affected by adhesive capsulitis.26 Another explanation is that our threshold for meaningful improvement was too high. However, when we reduced the value to 33% or greater improvement, we found that only 66 patients (2.8% of the sample) had this result. Our mean intake PCS-12 score was 39.6 (Tab. 1), which may not be low enough for significant change to be measured.
The ability to generalize the findings from this study is constrained by sample selection and composition. First, the patients were chosen in a nonrandomized fashion from a single outcomes management company that does not have a uniform distribution of clients across the United States. Selection bias within the FOTO data set also is suggested by the statistically significant differences in age, employment status, and intake MCS-12 scores between patients included in the study versus those excluded from the study (Tab. 1). Despite these differences, both groups had similar PCS-12, PF, BP, and HF scores, suggesting that these attributes had a limited influence on the impairments and functional limitations of these individuals. Commitment to job or better mental health, however, may explain why so many patients in the excluded group did not complete their episodes of care and may have influenced our findings.
Second, we were unable to confirm the diagnosis recorded by the physical therapists or the methods by which the physical therapists selected individual diagnostic labels. Most of the RCTs evaluating treatment effects in patients with adhesive capsulitis used a limitation in one or more planes of glenohumeral joint PROM as an inclusion criterion.6,10,11,13,14,16,17,20–22 The FOTO data set does not include quantitative impairment measures; therefore, the potential lack of reliability and validity of the ICD-9 codes themselves also should be considered.42
Coding and interpretation of treatment options pose another challenge in this study. For example, we cannot determine the specific interventions that were provided as part of the “joint mobility” intervention. FOTO defines joint mobility in the therapist training manual31 as “self-explanatory,” implying to us that therapists had to judge for themselves whether the interventions they applied were designed to affect joint mobility. In the absence of professional consensus on the operational definitions of all physical therapy interventions, joint mobility exercises may vary considerably. Details from actual patient records would be required to identify the specific techniques attributed to this label.
To our knowledge, this is the first study to use statistically created intervention categories to evaluate the effectiveness of various physical therapy interventions for people with adhesive capsulitis. Our approach provides an objective method for reducing a wide variety of interventions into sensible groups of related techniques. However, we may have missed some important categories because of the criteria we used to include and exclude specific types of treatments. Additional research is needed to confirm the stability of these intervention categories in additional samples of patients with this disorder.
A related issue is the inability to assess the timing or sequence of interventions applied over the episode of care. Physical therapists identified all interventions provided when the patients were discharged. We were unable to identify what role, if any, the timing or sequence of interventions may have played in patient outcomes. In addition, our factor analysis explained only 48% of the variance in treatment categories used by physical therapists in this study. The substantial portion of unexplained variance may be attributable to those interventions we excluded because of weak factor loading or infrequent representation in the data set. Future research should address the impact of the interventions that were excluded from our analysis.
Therapists should consider increasing their use of joint mobility and exercise interventions as defined in this study for patients with adhesive capsulitis. Ultrasound, massage, iontophoresis, and phonophoresis reduce the odds of improved outcomes in these patients. Future RCTs are needed to clarify treatment dosage for the more-effective interventions for people with adhesive capsulitis. Specifically, studies of the frequency with which joint mobility interventions should be applied in isolation or in combination with exercise are needed in order to optimize physical therapy care.
Dr Jewell and Dr Riddle provided concept/idea/research design and writing. Dr Jewell provided data collection. Dr Jewell and Dr Thacker provided data analysis. Dr Riddle provided consultation (including review of manuscript before submission). The authors acknowledge Dennis L Hart, PT, PhD, Director of Research and Consulting Services, Focus On Therapeutic Outcomes Inc, for his assistance on this project.
The Virginia Commonwealth University Institutional Review Board approved this study according to expedited review criteria.
↵* Focus on Therapeutic Outcomes Inc, PO Box 11444, Knoxville, TN 37939-1444.
↵† SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.
↵‡ SAS Institute Inc, SAS Campus Dr, Cary, NC 27513.
- Received August 14, 2008.
- Accepted January 14, 2009.
- American Physical Therapy Association