Invited commentary.

Prognosis is an important aspect of evidence-based care provided by physical therapists. The Guide to Physical Therapist Practice defines prognosis as the "determination of the level of optimal improvement that might be obtained from intervention and the amount of time required to reach that level of improvement." (1) According to Dorland's Illustrated Medical Dictionary, prognosis is "a forecast as to the probable outcome of an attack of disease; the prospect as to recovery from a disease as indicated by the nature of symptoms of the case." (2) The ability to accurately predict the outcomes of a health condition and the likelihood that each outcome will occur could greatly enhance the care provided by physical therapists. An accurate method to predict the outcome of care would allow clinicians to distinguish between those patients who are most likely to have a favorable outcome from those who are at risk for a less favorable outcome. More specifically, development and validation of a clinical prediction rule to guide clinical decision making could provide physical therapists with a valuable tool for quantifying the contributions that various components of the history and physical examination make toward predicting the likely response to treatment for an individual patient.

The purpose of the study by Kennedy et al was to determine prognostic factors affecting response in soft tissue disorders of the shoulder in patients receiving physical therapy. As such, this study has the potential to enhance clinical decision making, thus contributing to physical therapist practice.

The level of evidence provided by any research study is dependent on the research design relative to the underlying question addressed by the study. For prognostic studies, the optimal study design is a prospective cohort study in which one or more groups of patients who have not yet experienced the outcome of interest are followed forward over time. Potential prognostic factors or predictors of outcome are collected when patients are enrolled in the study, and patients are followed forward over time to determine who develops the outcome of interest. An analysis is performed to determine which prognostic factors predict the outcome of interest. For the study to be valid, it must include a well-defined sample of patients who are representative of the population of interest and must make use of well-established criteria to determine when the outcome of interest has occurred. (3) Furthermore, the predictors of outcome must be reliably and accurately measured.

The study by Kennedy and colleagues has met many of these criteria. They have included a well-defined sample of patients who are representative of patients with soft tissue disorders of the shoulder that are managed by physical therapists. The outcome of interest was patient-reported disability, and this was measured with the Disability of the Arm, Shoulder, and Hand (DASH) outcomes questionnaire, which has demonstrated psychometric properties (including reliability, validity, and responsiveness) in patients with shoulder disorders who are receiving outpatient physical therapy. The authors measured a number of potential predictors of outcome at the time of enrollment of each patient into the study. Broadly, these prognostic factors included demographic, disorder-related and disability measures, medication use, clinical findings, and expectations for recovery.

Several of the decisions made by the authors raise some interesting questions that warrant further discussion. One of these decisions was how to operationally define the outcome of interest. Specifically, the authors defined outcome in terms of the level of disability at the end of care as well as the amount of change in disability from the beginning to the end of care. The results indicated that disability at the end of care and change in disability over the course of care were predicted by a different set of prognostic variables. Higher levels of disability at the end of care were predicted by a higher level of disability at the beginning of care, presence of a workers' compensation claim, the physical therapist's prediction of restricted activities at the end of care, the patient's increasing age, and whether the patient was female. On the other hand, a larger change in disability from the beginning to the end of care was predicted by a higher initial level of pain, surgery within 6 months of the start of care, a shorter duration of symptoms before the start of physical therapy, younger age, and worse physical health. The regression models were able to account for approximately 36% of the variability in the level of disability at the end of care and 23% of the variability of the change in disability from the start to end of care. Based on these results, the authors concluded that, in the future, consumers of the literature must consider not only what instrument was used to measure disability, but also whether disability was defined in terms of the level of disability at the end of care or the change in disability over the course of care.

Given that different sets of predictors were identified dependent on how disability was defined, an interesting question arises: What is more important from the patient's perspective--the level of disability at the end of care, or the change in disability from the start to the end of care? We could argue that the most important outcome from the patient's perspective is the level of disability at the end of care. Lower levels of disability at the end of care would be expected to be viewed as more important by the patient. However, in the discussion, the authors have presented other perspectives of what may be most important from the patient's perspective. These include change in the level of disability, achievement of some threshold level of disability where the patient can cope with the disability, achievement of a "normal" or "functional" range of disability, or change in disability that is greater than measurement error. Further research will be needed to determine what is most important from the patient's perspective.

A limitation of defining outcome in terms of the change in disability from the start to the end of care is that the magnitude of change may be dependent on the initial level of disability. Those with initial higher levels of disability have "more room" for improvement and thus a greater change score compared with those who have lower initial levels of disability. In essence, this creates a ceiling effect for those with a lower initial level of disability. In support of this, the authors found a correlation coefficient of -.60 between the initial level of disability and the change in the level of disability from the start to the end of care. This finding suggests that greater change was associated with higher levels of initial disability.

The question arises as to whether the baseline disability scores should have been included in the regression model to predict the change in disability from the start to end of care. The authors elected not to include the initial level of disability in the regression model to avoid the dependency that this would create between the independent and dependent variables in the regression model. Some biostatisticians, however, would recommend inclusion of the initial level of disability in the regression model to predict the change in disability to remove the covariance due to the initial scores (personal communication; James E Bost, PhD, Biostatistician, Center for Research on Health Care Data Center, University of Pittsburgh, Pittsburgh, Pa; May 11, 2006). Further analysis of the results to determine how inclusion of the initial level of disability would affect the prediction of the change in disability would be interesting to explore.

The explained variation in the models to predict the final level of disability and change in disability from the start to the end of care was 36% and 23%, respectively. This implies that 64% of the variability in the final level of disability and 77% of the change in disability was unexplained by the set prognostic variables that were included in the models. Given this, the prognostic factors included in the model are relatively inaccurate in predicting the outcome of physical therapy management of soft tissue disorders of the shoulder in terms of either the final level of disability or the change in disability from the start to the end of care. This limits the usefulness of the prediction model when attempting to apply it clinically to an individual patient.

Given the relatively large amount of unexplained variation, additional research is needed to identify other predictors of outcome. To manage soft tissue disorders of the shoulder, physical therapists often focus on impairments in range of motion (ROM) and muscle function. Therefore, it would be informative to know how impairments in ROM and muscle performance relate to outcome in terms of disability. The authors included measurements of ROM and muscle performance in the prediction models; however, these impairments were not precisely measured. Because of the large number of physical therapists participating in the study, the authors elected to dichotomize these variables as "none" versus some degree of limited motion or muscle performance. There were significant univariate relationships between ROM and disability at the end of care, and the change in disability and muscle performance was related to disability at the end of care; however, these variables were not included in the final prediction models. It is likely that the decision to dichotomize the ROM and muscle performance variables improved the reliability of those measurements; however, it is also likely that dichotomizing these variables resulted in some loss of measurement precision, which may have attenuated the relationship of these variables to outcome. In the future, ROM and muscle performance should be quantified using goniometry and handheld dynamometry, respectively. Furthermore, to account for individual variation among patients, ROM and muscle performance should be compared between the involved and noninvolved extremities. Establishing the relationship between measurements of impairment of ROM and muscle performance with the outcome of care may allow physical therapists to better direct their intervention to focus on managing the underlying impairments.

Other variables not measured by the authors of the study may be important predictors of treatment outcome for patients with soft tissue disorders of the shoulder. For example, in patients with acute low back pain, fear-avoidance behavior was found to be a significant predictor of the outcome of manipulation therapy, (4) and education level has been found to be a significant predictor of outcome after anterior cruciate ligament reconstruction. (5) Therefore, future research should consider the effects of psychosocial variables, such as fear-avoidance behavior, depression, anxiety, and education level, on the outcome of physical therapy management of soft tissue disorders of the shoulder.

Another limitation of the prediction models that were developed is that the identified prognostic factors are not under the direct control of the physical therapist. The authors addressed this in the introduction, indicating that even though the prognostic factors are not under the control of the physical therapist, the prediction model can still be helpful in predicting the outcome of the episode of care. This information may provide both the patient and the physical therapist with knowledge about the expected outcome.

In the future, it would be beneficial to investigate the effects that specific interventions have on the outcome of care provided by physical therapists. This will not be a simple task. It is not likely that a single intervention will prove to be beneficial for all patients. Rather, it is likely that the effects of specific interventions will be dependent on the patient's signs and symptoms. For example, it could be hypothesized that patients with limited motion will achieve greater improvements in ROM and a greater reduction in disability if they receive interventions directed at improving ROM (eg, ROM and stretching exercises, joint mobilization) than if they do not receive these interventions directed at restoring motion. Furthermore, the effects of ROM and stretching exercises in patients with limited ROM may be dependent on the acuteness of the condition. For example, patients with acute adhesive capsulitis (ie, in the "freezing" phase) may respond better to ROM exercises within the available ROM as opposed to stretching exercises. Conversely, patients in the "thawing" phase may benefit more from stretching exercises as opposed to ROM exercises.

Similar arguments could be made for targeted interventions for other impairments--such as rotator cuff weakness, scapular muscle weakness, or glenohumeral joint laxity--that are commonly experienced by individuals with soft tissue disorders of the shoulder. In essence, this argues for the development and testing of a treatment classification scheme for individuals with soft tissue disorders of the shoulder. A treatment classification scheme for management of low back pain has been proposed by Delitto et al, (6) and recent evidence suggests that individuals who receive treatments matched to their classification have a more optimal outcome. (4,7-9) Development of a similar treatment classification scheme for management of soft tissue disorders of the shoulder may lead to improved treatment outcomes.

Using cluster analysis, Winters et al (10) proposed a classification system for patients with complaints of shoulder pain that was based on clusters of symptoms and signs with unique interventions for each category. The results revealed 3 classifications based on the magnitude and duration of pain and ROM limitations. (10) Furthermore, there is some evidence to suggest that these classifications are somewhat helpful in directing interventions. (11) Further development of a treatment classification system for patients with soft tissue disorders of the shoulder may lead to improvements in the ability to predict the outcomes of physical therapy management and should be the subject of future research.

In summary, Kennedy and coauthors should be congratulated for their efforts in developing a model to predict the outcome of physical therapy management of patients with soft tissue disorders of the shoulder. The methodology used by these authors lays the ground work for others to follow to address this important question. In particular, the authors have made an important point that careful consideration should be given to how outcome is defined, that is, either as the level of disability at the end of care or as the change in disability from the start to end of care. The usefulness of the prediction model developed by these authors will be determined by replication of this work in a second independent sample. Future efforts to determine prognosis for soft tissue disorders of the shoulder should consider more precise measurements of impairment, the effects of the specific interventions provided, and the interaction between the patient's signs and symptoms and the interventions provided. Furthermore, the prognostic value of psychosocial variables, which may indicate the need for consultation with other health care professionals, should be explored.

References

(1) Guide to Physical Therapist Practice 2nd ed. Phys Ther. 2001;81:9-744.

(2) Dorland's Illustrated Medical Dictionary. 30th ed. Philadelphia, Pa: WB Saunders Co; 2003.

(3) Sackett DL, Straus SE, Richardson WS, et al. Evidence-Based Medicine: How to Practice and Teach EBM. 2nd ed. London, United Kingdom: Churchill Livingstone; 2000.

(4) Childs JD, Fritz JM, Flynn TW, et al. A clinical prediction rule to identify patients with low back pain most likely to benefit from spinal manipulation: a validation study. Ann Intern Med. 2004;141:920-928.

(5) Spindler KP, Warren TA, Claiborne Callison J Jr, et al. Clinical outcome at a minimum of five years after reconstruction of the anterior cruciate ligament. J Bone Joint Surg Am. 2005;87:1673-1679.

(6) Delitto A, Erhard RE, Bowling RW. A treatment-based classification aproach to low back syndrome: identifying and staging patients for conservative treatment. Phys Ther. 1995;75:470-489.

(7) Hicks GE, Fritz JM, Delitto A, McGill SM. Preliminary development of a clinical prediction rule for determining which patients with low back pain will respond to a stabilization exercise program. Arch Phys Med Rehabil. 2005;86:1753-1762.

(8) George SZ, Fritz JM, Bialosky JE, Donald DA. The effect of fear-avoidance based physical therapy intervention for patients with acute low back pain: results of a randomized clinical trial. Spine. 2003;28:2551-2560.

(9) Brennan GP, Fritz JM, Hunter SJ, et al. Identifying subgroups of patients with acute/subacute "nonspecific" low back pain. Spine. 2006;31:623-631.

(10) Winters JC, Groenier KH, Sobel JS, et al. Classification of shoulder complaints in general practice by means of cluster analysis. Arch Phys Med Rehabil. 1997;78:1369-1374.

(11) Winters JC, Sobel JS, Groenier KH, et al. The course of pain and the restriction of mobility in patients with shoulder complaints in general practice. Rheumatol Int. 1997;166:219-225.

James J Irrgang, PT, PhD, ATC 

Associate Professor and Director of Clinical Research

Department of Orthopaedic Surgery

University of Pittsburgh School of Medicine 

Room 911, Kaufman Building

34 71 Fifth Ave

Pittsburgh, PA 15260

[e-mail]