Advertisement

Secondary Mediation and Regression Analyses of the PTClinResNet Database: Determining Causal Relationships Among the International Classification of Functioning, Disability and Health Levels for Four Physical Therapy Intervention Trials

Sara J. Mulroy, Carolee J. Winstein, Kornelia Kulig, George J. Beneck, Eileen G. Fowler, Sharon K. DeMuth, Katherine J. Sullivan, David A. Brown, Christianne J. Lane
DOI: 10.2522/ptj.20110024 Published 1 December 2011

Abstract

Background Each of the 4 randomized clinical trials (RCTs) hosted by the Physical Therapy Clinical Research Network (PTClinResNet) targeted a different disability group (low back disorder in the Muscle-Specific Strength Training Effectiveness After Lumbar Microdiskectomy [MUSSEL] trial, chronic spinal cord injury in the Strengthening and Optimal Movements for Painful Shoulders in Chronic Spinal Cord Injury [STOMPS] trial, adult stroke in the Strength Training Effectiveness Post-Stroke [STEPS] trial, and pediatric cerebral palsy in the Pediatric Endurance and Limb Strengthening [PEDALS] trial for children with spastic diplegic cerebral palsy) and tested the effectiveness of a muscle-specific or functional activity–based intervention on primary outcomes that captured pain (STOMPS, MUSSEL) or locomotor function (STEPS, PEDALS).

Objective The focus of these secondary analyses was to determine causal relationships among outcomes across levels of the International Classification of Functioning, Disability and Health (ICF) framework for the 4 RCTs.

Method and Design With the database from PTClinResNet, we used 2 separate secondary statistical approaches—mediation analysis for the MUSSEL and STOMPS trials and regression analysis for the STEPS and PEDALS trials—to test relationships among muscle performance, primary outcomes (pain related and locomotor related), activity and participation measures, and overall quality of life.

Results Predictive models were stronger for the 2 studies with pain-related primary outcomes. Change in muscle performance mediated or predicted reductions in pain for the MUSSEL and STOMPS trials and, to some extent, walking speed for the STEPS trial. Changes in primary outcome variables were significantly related to changes in activity and participation variables for all 4 trials. Improvement in activity and participation outcomes mediated or predicted increases in overall quality of life for the 3 trials with adult populations.

Limitations Variables included in the statistical models were limited to those measured in the 4 RCTs. It is possible that other variables also mediated or predicted the changes in outcomes. The relatively small sample size in the PEDALS trial limited statistical power for those analyses.

Conclusions Evaluating the mediators or predictors of change between each ICF level and for 2 fundamentally different outcome variables (pain versus walking) provided insights into the complexities inherent across 4 prevalent disability groups.

Jette and Latham, in a 2010 editorial in PTJ, stated, “People with disabilities comprise the largest and arguably the most important health care consumer group in the United States; however, as the Institute of Medicine and others have warned,1,2 far too little progress has been made over the past 2 decades in removing obstacles that limit what many people with physical and cognitive impairments can achieve.”3(p324) If we are to address this concern in a systematic way, we need to understand the nature of the obstacles that limit various disability groups. As Jette4 and Iezzoni and Freedman5 have argued, the World Health Organization's International Classification of Functioning, Disability and Health (ICF)6 framework has the potential to become a standard for disability language that looks beyond mortality and disease to focus on how people carry out their lives with these chronic conditions.

An evolving trend in health care, prompted to some degree by a more patient-centered approach, is the shift to a greater focus on treating disability and improving participation and quality of life (QOL). Although impairment-focused interventions are still viable approaches in some cases, the outcomes of such interventions are evaluated by the degree to which an improvement at the impairment level affects an individual's ability to participate in chosen meaningful activities, thereby also affecting his or her QOL. More than 15 years ago, Jette7 challenged the physical therapy community to make a paradigm shift from descriptive studies to those that provide direct evidence of the degree to which physical therapy that affects an impairment such as muscle performance also will reduce disability and improve functional outcomes of the individual. Thus, diseases and treatment interventions are still primarily focused on impairment mitigation, but with the additional considerations of reducing disability and improving participation and QOL. The question of how well our interventions accomplish these additional goals can be addressed by including these progressively more distal ICF levels as outcomes in our research designs.

The Physical Therapy Clinical Research Network (PTClinResNet) was established in 2002 with funding from the Foundation for Physical Therapy of the American Physical Therapy Association.8 The network was established to accomplish 3 specific aims: (1) create the infrastructure necessary to develop and sustain clinical trials research in rehabilitation, (2) generate evidence to evaluate the efficacy of exercise-based physical therapist interventions designed to improve muscle performance and movement skills, and (3) provide education and training opportunities for present and future clinician-researchers in rehabilitation research and for the rehabilitation community at large in its support of evidence-based practice. In its original proposal, PTClinResNet used the Nagi disablement model9 as a framework for outcome selection. However, early on, the network adopted the ICF model, in large part because of its acceptance by the broader community of health care professionals and its greater research potential to identify important interactions and links among outcome measures.4,6 One of the aims of PTClinResNet was to generate evidence to evaluate the efficacy of physical therapist interventions, particularly related to populations with characteristic force generation impairments. To accomplish this aim, we conducted 4 projects: (1) Strength Training Effectiveness Post-Stroke (STEPS), (2) Muscle-Specific Strength Training Effectiveness After Lumbar Microdiskectomy (MUSSEL), (3) Strengthening and Optimal Movements for Painful Shoulders in Chronic Spinal Cord Injury (STOMPS), and (4) Pediatric Endurance and Limb Strengthening for children with spastic diplegic cerebral palsy (PEDALS).

Once the primary intervention outcomes from the 4 trials had been published,10–13 we sought to conduct a set of secondary analyses using the PTClinResNet database with a focus on determining the causal relationships among outcomes across levels of the ICF framework. Because we adopted the ICF early in our design, we chose several common measures of activity and participation (eg, 36-Item Short-Form Health Survey questionnaire [SF-36])14 and QOL (ie, single-item QOL)15 across the 4 studies in anticipation of the kind of secondary analyses described here. Thus, the focus of these secondary analyses was to uncover the causal relationships among the ICF levels in the context of a common set of physical therapy interventions that targeted 2 distinct classes of primary outcomes—pain-related primary outcomes (MUSSEL, STOMPS) and walking-related primary outcomes (STEPS, PEDALS). The secondary analyses described here leverage the PTClinResNet database to include 4 disability groups and an ICF model design to better understand the following causal relationships: (1) the relationship between impairment mitigation at the muscle performance level on pain-related and locomotor-related primary outcomes, (2) the relationship between the primary outcome variables and activity and participation outcomes, and (3) the relationship between activity and participation outcomes and overall QOL.

There is no doubt that “disability is complex, multifactorial, and difficult to define.”3(p324) Some of the complexity is inherent to the not so well-understood relationships among the levels of the ICF and the time course for change. For example, Lai et al16 found that therapeutic exercise administered to survivors of subacute stroke resulted in accelerated gains in daily functioning and QOL, compared with usual care, but additional benefit of the intervention had dissipated by 6 months after the intervention, thus underscoring the inherent complexity and multifaceted nature of disability. Careful consideration of previous research findings, fundamental understanding of muscle adaptation to exercise, and considerable clinical expertise were used together to formulate the design of each intervention hosted by the PTClinResNet.

The nature of disability is known to be dynamic, multifactorial, and, in some cases, poorly defined with respect to physical impairment and participation.17 For example, a recent systematic review showed that a variety of factors, including gross motor function, manual ability, cognitive ability, communicative skills, age, and sex, are the most important variables associated with participation in leisure activities for children and youth with physical disabilities.18 Despite this complexity, the investigative teams based their approach upon many assumptions that, in retrospect, relied on a relatively simplistic view of the disability condition. For example, we assumed that if specific muscles were strengthened through our exercise or activity protocols, walking function would improve (STEPS, PEDALS) or pain would be reduced (STOMPS, MUSSEL). Furthermore, we assumed that if pain was reduced or functional activity (eg, walking, wheelchair propulsion) was enabled, this improvement would lead to improved participation in meaningful activities. Finally, we assumed that if the consequences of the disability were reduced, this change would affect the overall QOL of the individual.

Therefore, the purpose of this study was to determine the degree to which our simplistic assumptions were supported and, in so doing, to provide a deeper understanding of the inherent complexity and multifaceted nature of 4 of the most common chronic disabilities among patients seeking professional care from a physical therapist.

Method

This study consisted of a set of secondary analyses of the existing data sets from the 4 randomized controlled trials (RCTs) hosted by the PTClinResNet. A detailed description of the study sample, interventions, and outcome variables is available in the primary outcome paper for each trial.10–13 Below we provide an overview of each study design, and a summary of specific variables measured in each trial is provided in Table 1 for easy comparison among studies.

View this table:
Table 1.

Study Design and Outcome Variables for the 4 Randomized Controlled Trials Hosted by the Physical Therapy Clinical Research Networka

MUSSEL

The primary objective of the MUSSEL10 trial was to assess the early effects of muscle-specific strengthening on decreasing low back pain-related disability and improving function and QOL in people recovering from lumbar microdiskectomy. A total of 98 individuals who were 4 to 6 weeks post-microdiskectomy were enrolled into the study and randomly assigned to 1 of 2 intervention groups: education only (one session of back care education) or education + exercise (one back care education session followed by 12 weeks of back extensor strength and endurance training delivered 3 times per week). At the end of the 12-week intervention, 26 participants had not adhered to their assigned group, mostly those enrolled in the education only group who sought outside care. This outcome resulted in 3 intervention groups: education only, usual care, and education + exercise. For this secondary analysis, we used data from these 3 intervention groups.

STOMPS

The primary objective of the STOMPS11 trial was to determine the efficacy of a combined intervention comprising strengthening exercises and instruction in how to optimize movements to reduce shoulder pain in people with chronic spinal cord injury who routinely use a manual wheelchair for mobility. A total of 80 individuals with paraplegia resulting from spinal cord injury and shoulder pain were enrolled and randomly assigned to 1 of 2 intervention groups: an attention-control education group and a group that received the combined program of shoulder exercises performed 3 times per week for 12 weeks paired with instruction to optimize movement techniques of upper-extremity weight-bearing activities. Outcomes were assessed immediately after intervention and 4 weeks following the end of the intervention.

STEPS

The primary objective of the STEPS12 trial was to determine whether functional walking outcomes in individuals with chronic stroke are improved with exercise programs that include task-specific training (ie, body-weight–supported treadmill training) or lower-extremity strength training, or a combined exercise program that includes both task-specific and strength training. A total of 80 individuals who were at least 6 months poststroke were enrolled and randomly assigned to 1 of 4 intervention groups: (1) limb-loaded leg cycling + arm cycling (sham), (2) body-weight–supported treadmill training (BWSTT) + arm cycling (sham), (3) BWSTT + limb-loaded leg cycling, and (4) BWSTT + muscle-specific strengthening exercises. Participants received each of the 2 intervention components twice weekly for 6 weeks (4 days per week). Outcomes were assessed immediately after intervention and 6 months following the end of the intervention.

PEDALS

The primary objective of the PEDALS13 trial was to determine the efficacy of a stationary cycling intervention for treatment outcomes in children with cerebral palsy. A total of 64 children (aged 7–18 years) with spastic diplegia were enrolled and randomly assigned to either a group that received a 12-week program of limb-loaded leg cycling performed 3 times per week or to a control group that received no intervention. Outcomes were assessed immediately after intervention.

All participants in each of the 4 trials read and signed an informed consent statement that had been approved by the institutional review boards of each of the individual study-specific institutions.

Overview of Data Analysis

We first evaluated the viability of mediation modeling to characterize the relationships between changes in outcomes induced by the interventions. Mediation modeling is a causal analysis method that identifies whether changes in an outcome (dependent variable) produced by an intervention (independent variable) were caused by (or could be explained by) changes in a mediator variable that also was affected by the intervention (Fig. 1).19 This approach requires intervention group–related differences in both the mediator and outcome variables. Both the MUSSEL and STOMPS trials were designed with placebo group comparisons and resulted in intervention group–related differences in outcome variables across all ICF levels. Consequently, these 2 trials met the criteria for mediation analysis, which we used to test the following 3 hypotheses:

  1. Changes in muscle performance would be statistically significant mediators of the intervention-induced changes in the primary outcome variables (ie, pain and pain-related disability).

  2. Changes in the pain-related primary outcome variables would be statistically significant mediators of the intervention-induced changes in activity and participation outcome variables.

  3. Changes in activity and participation variables would be statistically significant mediators of the intervention-induced changes in overall QOL.

Figure 1.
Figure 1.

Theoretical model of mediation analysis. An independent variable or intervention group has an impact on a dependent variable or outcome (path C) that is caused by the effect of the independent variable on a mediator variable (path A), which, in turn, produces the change in the dependent variable (path B). The mediation is complete if the intervention group is no longer a statistically significant predictor of the outcome variable after accounting for the effect of the change in the mediator (path C′). The mediation is partial if the intervention group remains a significant but weaker predictor of the outcome variable after accounting for the effect of the change in the mediator.

The STEPS trial did not have a placebo control group, as it was designed to compare combinations of task-specific and muscle strengthening intervention approaches. Participants in the 3 intervention groups who received body-weight–supported treadmill training demonstrated greater improvement in customary walking speed, but not in muscle strength, activity, participation, or QOL, compared with those who received leg cycling paired with upper-extremity cycling.4 By contrast, the PEDALS trial did have a placebo control group. Although outcome variables improved for participants in the experimental group of the PEDALS trial and not for those in the control group, between-group differences in outcomes were not statistically significant.13 Consequently, the mediation modeling approach was not appropriate for the STEPS and PEDALS RCTs. Instead, we used a regression analysis for these 2 trials to evaluate the relationships between changes in outcome variables without testing the impact of intervention group. We tested the following 3 hypotheses:

  1. Changes in muscle performance would be significant predictors of the change in the locomotor-related primary outcome variables: customary walking speed and walk/run speed.

  2. Changes in the locomotor-related primary outcome variables would be significant predictors of the changes in activity and participation outcome variables.

  3. Changes in activity and participation variables would be significant predictors of the changes in overall QOL.

Statistical Analysis

Table 1 lists all of the outcome measures that were potential variables in the statistical models for the 4 RCTs. The SF-3614 and the Subjective Quality of Life Scale15 were included as activity and participation variables for the 3 studies with adult populations (MUSSEL, STOMPS, and STEPS). Subscales of the SF-36 were tested rather than composite scores to avoid inclusion of domains that either were not expected to be affected by the interventions (general health) or were too closely related to the primary outcome variables (bodily pain for the MUSSEL and STOMPS trials). The Pediatric Quality of Life Inventory (PedsQL)20 total score and physical health and psychosocial health subscales were included for the PEDALS analysis as activity and participation variables. The response on the single-item parental satisfaction with physical condition from the Pediatric Outcomes Data Collection Instrument (PODCI) was used as an analog of overall QOL.21

Both mediation modeling and regression analysis are causal models that are specified based on content knowledge.22 Variables were selected for evaluation in the statistical models based on predictions of mechanisms underlying each intervention. In addition, for the MUSSEL and STOMPS trials, only variables that demonstrated statistically significant differences among intervention groups in the primary outcome studies were selected for testing. Tables 2 and 3 list the variables selected for evaluation in the mediation and regression models, respectively. If 2 or more variables were significant mediators or predictors of the change in outcome for each model, the mediator (or predictor) with greatest change in the R2 value and the largest standardized regression coefficient was selected for inclusion in the model.

View this table:
Table 2.

Variables Evaluated for Mediation Modeling Analysis at Each Level of the International Classification of Functioning, Disability and Health (ICF) for the Muscle-Specific Strength Training Effectiveness After Lumbar Microdiskectomy (MUSSEL) and Strengthening and Optimal Movements for Painful Shoulders in Chronic Spinal Cord Injury (STOMPS) Clinical Trialsa

View this table:
Table 3.

Variables Evaluated for Regression Modeling Analysis at Each Level of the International Classification of Functioning, Disability and Health (ICF) for the Strength Training Effectiveness Post-Stroke (STEPS) and Pediatric Endurance and Limb Strengthening (PEDALS) for Children With Spastic Diplegia Cerebral Palsy Clinical Trialsa

Separate meditational analyses were used to test each of the 3 hypotheses for the MUSSEL and STOMPS trials.19 An intervention group–related change was first confirmed for both mediator and dependent variables by demonstrating that intervention group was a significant predictor of the change in each variable after controlling for baseline values (Fig. 1, paths A and C). Linear regression analysis then was used to determine whether the change in a potential mediator variable was a significant predictor of the change in the dependent variable after controlling for baseline values of both the mediator and dependent variables (Fig. 1, path B). The significance of both change in R2 value and regression coefficient was assessed. Finally, the addition of intervention group assignment to the model determined whether the mediation was complete or partial. If group assignment was no longer a statistically significant predictor of the outcome variable after controlling for the mediator (Fig. 1, path C′), the change in the mediator explained all of the impact of the intervention on the change in the dependent variable. Conversely, if intervention group remained a significant predictor of the change in the dependent variable after controlling for the mediator, the mediator explained only part of the intervention-related change in the dependent variable, and other factors associated with group assignment were presumed to contribute to the change in the dependent variable.

A similar approach using 3 separate linear regression analyses was used to test the hypotheses for the STEPS and PEDALS trials, but without testing the impact of intervention group. Participants were included from all 4 intervention groups of the STEPS trial, but only participants in the cycling group of the PEDALS trial were included, as those in the control group did not receive an intervention.

Role of the Funding Source

This study was supported by grants from the Foundation for Physical Therapy (American Physical Therapy Association) to establish the PTClinResNet, the National Institute for Disability and Rehabilitation Research through the Research and Rehabilitation Training Center (RRTC) on Aging With a Disability (grant H133B081002), and the National Institute for Disability and Rehabilitation Research through the Rehabilitation Engineering Research Center (RERC) on Successful Aging With a Disability (grant H133E080024). The PTClinResNet was a clinical research network to evaluate the efficacy of physical therapist practice.

Results

Mediation Models Identifying Relationships Between Changes in Variables for MUSSEL and STOMPS

Table 2 identifies (with an asterisk) the variables that were statistically significant mediators of the outcome variables and which was the strongest mediator (in bold type) that was included in each model testing the 3 hypotheses for the MUSSEL and STOMPS RCTs. Figure 2 depicts the sequential increases in R2 values in each of the mediation models. In the analysis of the first hypothesis for the MUSSEL trial, both Sorenson and modified Sorenson test23 (lumbar extensor muscle endurance) scores were complete mediators of the change in the Oswestry Disability Index24 (back pain and pain-related disability) score (Tab. 2). The modified Sorenson test score was the stronger of the 2 mediators and was included in the final model (R2 change=.157, P<.001) (Fig. 2). After controlling for the change in lumbar extensor muscle endurance, intervention group was no longer a predictor of the pain-related primary outcome. In the second model for the MUSSEL trial, change in the Oswestry Disability Index score was a complete mediator of the intervention-related change in the SF-36 role–physical subscale score (impact of physical disability on life roles) in the MUSSEL trial (R2 change=.241, P<.001). Finally, change in the SF-36 role–physical subscale score mediated the intervention-related change in overall QOL in the model of the third hypothesis for the MUSSEL trial (R2 change=.212, P<.001).

Figure 2.
Figure 2.

Sequential increases in the R2 values predicting outcome variables in the 3 mediation models for the Muscle-Specific Strength Training Effectiveness After Lumbar Microdiskectomy (MUSSEL) and Strengthening and Optimal Movements for Painful Shoulders in Chronic Spinal Cord Injury (STOMPS) trials. Asterisk denotes a statistically significant increase in R2 value (P<.05). The blue bar represents the R2 value attributable to the baseline value of the outcome variable. The red bar represents the increase in R2 value attributable to adding the baseline value of the mediator variable to the model. The green bar represents the increase in R2 value attributable to adding the change of the mediator variable to the model (path B in Fig. 1). If the green bar is statistically significant, mediation is confirmed. The purple bar represents the increase in R2 value attributable to adding the intervention group to the model after accounting for the change in mediator (path C′ in Fig. 1). If the green bar is statistically significant but the purple bar is not, complete mediation is confirmed. If both the green and purple bars are statistically significant, partial mediation is confirmed. Models for MUSSEL trial: Model 1—change in the muscle performance variable (lumbar endurance, as assessed with the modified Sorenson test) was a significant mediator of change in the primary outcome variable (back pain and pain-related disability, as assessed by Oswestry Disability Index scores, 0–12 weeks). Model 2—change in back pain and pain-related disability (as assessed by Oswestry Disability Index scores) was a significant mediator of change in scores on the activity and participation measure (36-Item Short-Form Health Survey questionnaire [SF-36] role–physical subscale). Model 3—change in scores on the activity and participation measure (SF-36 role–physical subscale) was a significant mediator of overall quality of life. Models for STOMPS trial: Model 1—change in muscle performance variable (shoulder external rotation torque) was a significant partial mediator of the change in the primary outcome variable (shoulder pain, as assessed by Wheelchair User's Shoulder Pain Index [WUSPI] scores, 0–12 weeks). Intervention group (purple bar) was still a significant predictor of change in shoulder pain after accounting for the impact of change in external rotation torque. This finding indicates that factors in addition to improved muscle strength associated with being in the intervention group contributed to improvement in the primary outcome variable. Model 2—change in shoulder pain (WUSPI score) was a significant mediator of the change in scores on the activity and participation measure (SF-36 social function subscale). Model 3—change in scores on the activity and participation measure (SF-36 role–physical subscale) was a significant mediator of overall quality of life. SQOL=Subjective Quality of Life Scale.

For the STOMPS trial, both shoulder external rotation and shoulder elevation torques satisfied the criteria for partial mediation of the intervention-related change in shoulder pain as measured by the Wheelchair User's Shoulder Pain Index25 (WUSPI) (Tab. 2). Shoulder external rotation torque was the stronger of the 2 mediators (R2 change=.113, P<.01) and was included in the final model (Fig. 2). Intervention group assignment was still a significant (although weaker) predictor of the change in shoulder pain after controlling for external rotation torque (R2 change=.124, P<.001), indicating that other factors associated with the intervention also contributed to the change in shoulder pain.

In the model of the second hypothesis for the STOMPS trial, the change in shoulder pain (WUSPI score) was a significant mediator of the intervention-related changes in both SF-36 social function and role–physical subscale scores. The mediation was stronger for the social function subscale (R2 change=.207, P<.001). Similarly, in the third model for the STOMPS trial, changes in both SF-36 social function and role–physical subscale scores were significant mediators of the intervention-related change in Subjective Quality of Life Scale score, with the social function subscale as the stronger of the 2 mediators (R2 change=.143, P<.001). The mediations were complete for both the second and third models for the STOMPS trial, as intervention group no longer added to the predictions of change in the SF-36 social function subscale score or overall QOL score.

Regression Models Predicting Relationships Between Changes in Variables for STEPS and PEDALS

Table 3 identifies (with an asterisk) the variables that were statistically significant predictors of change in the outcome variables and the strongest predictor (in bold type) that was included in each model testing the 3 hypotheses for the STEPS and PEDALS RCTs. Figure 3 depicts the sequential increases in R2 values in each of the regression models. In the first model of the STEPS analysis, change in paretic ankle plantar-flexion torque was the only muscle performance variable that was a significant predictor of the change in customary walking speed at 6 months (R2 change=.066, P<.03) (Tab. 3, Fig. 3). Change in customary walking speed at 6 months was a significant predictor of the change scores for several of the subscales of both the Stroke Impact Scale (SIS)26 and the SF-36 (Tab. 3). The strongest predictive relationship was with the SIS activities of daily living (ADL) subscale (R2 change=.118, P<.01). In the third model for the STEPS trial, change in SF-36 mental health subscale score was the only statistically significant predictor of the change in overall QOL (R2 change=.090, P<.03).

Figure 3.
Figure 3.

Sequential increases in the R2 values in the regression models for Strength Training Effectiveness Post-Stroke (STEPS) and Pediatric Endurance and Limb Strengthening (PEDALS) for Children With Spastic Diplegia Cerebral Palsy trials. Asterisk denotes a statistically significant increase in R2 value (P<.05). The blue bar represents the R2 value attributable to the baseline value of the outcome variable. The red bar represents the increase in R2 value attributable to adding the baseline value of the predictor variable to the model. The green bar represents the increase in R2 value attributable to adding the change of the predictor variable to the model. Models for STEPS trial: Model 1—change in muscle performance variable (ankle plantar-flexion torque in paretic leg, 0–12 weeks) was a significant predictor of change in the locomotor primary outcome variable (customary walking speed, 0–6 months). Model 2—change in locomotor primary outcome variable (customary walking speed) was a significant predictor of change in scores on the activity and participation measure (Sickness Impact Scale activities of daily living [SIS ADL] subscale, 0–6 months). Model 3—change in scores on the activity and participation measure (36-Item Short-Form Health Survey questionnaire [SF-36] mental health subscale) was a significant predictor of overall Subjective Quality of Life Scale (SQOL) scores (0–6 months). Models for PEDALS trial: model 1—change in muscle strength (knee extension torque at 120°/s) was not a statistically significant predictor of the change in the primary locomotor outcome measure (600-yd walk/run speed). Model 2—change in the primary locomotor outcome measure (600-yd walk/run speed) was a significant predictor of the change in the activity and participation variable (Pediatric Quality of Life Inventory [PedsQL] total score, 0–12 weeks). Model 3—change in activity and participation measure (PedsQL total score, 0–12 weeks) was not a statistically significant predictor of the parental satisfaction item on the Pediatric Outcomes Data Collection Instrument (0–12 weeks).

In the analysis of the PEDALS data, neither of the muscle performance variables was a statistically significant predictor of the change in 600-yd (548.64-m) walk/run time27 (Tab. 3). The results for knee extension torque are displayed in Figure 3, as the change in R2 value was greater than that for knee flexion torque (R2 change=.020, P=.495). For the second model of the PEDALS trial, change in 600-yd walk/run speed was a statistically significant predictor of the change in PedsQL total score20 (R2 change=.137, P<.01). Finally, none of the activity and participation variables were statistically significant predictors of the change in scores on the parental satisfaction item on the PODCI21 in the third analysis for the PEDALS trial. The results for the PedsQL total score as a predictor of the PODCI item are displayed in Figure, 3 as the change in R2 value was greater (R2 change=.045, P=.202) than those for the PedsQL physical health and psychosocial health subscale scores.

Discussion

The purpose of these secondary analyses was to elucidate the causal relationships in outcomes across the ICF framework. We analyzed 2 studies with pain-related primary outcome variables and 2 studies with locomotor-related primary outcomes. Overall, the predictive models were stronger for the 2 studies with pain-related primary outcomes. This trend underscores the significant negative impact of pain on participation and QOL.28,29 For both MUSSEL and STOMPS studies, the reduction in pain-related primary outcomes had a greater impact on participation outcomes than activity variables. None of the activity measures for either study listed in Table 1 (ie, walking speed, wheelchair propulsion speed, or activity scales) were different among intervention groups (or were only weakly affected) and were not mediated by change in the pain-related primary outcomes. For the most part, low back or shoulder pain did not make participants unable to perform these activities, just less willing to do them on a daily basis. The more general outcome measures of participation (SF-36 subscales) were more responsive to change with the intervention than measures of performance of specific activities. The presence of pain may have led participants to prioritize critical activities such as work or self-care and participate less frequently in optional leisure activities, which tend to vary greatly among individuals. Examination of the subscales of the more global activity and participation measures allowed us to identify specific domains that were affected by the change in scores for the primary outcome variables.

For the MUSSEL and STOMPS trials, reduction in pain-related primary outcomes mediated the intervention-related improvements in ability to fulfill life roles and social functions, which, in turn, mediated intervention-related improvements in overall QOL. For participants in these 2 trials, the intervention was addressing either the primary disability (ie, back pain) or a significant secondary disability (ie, shoulder pain). The reduction or alleviation of pain allowed them to increase their willingness to participate in more optional activities, which had a direct impact on QOL. In a cross-sectional study of multiple diagnostic groups, the presence of pain had the largest negative impact of all impairment variables on overall health-related QOL.28

Improvements in the pain-related primary outcomes of the MUSSEL and STOMPS trials were explained (or partially explained in the STOMPS trial) by muscle performance gains. Both of the experimental interventions were multifactorial, including strengthening exercises of multiple muscle groups, flexibility exercises, endurance training, and movement technique optimization. Despite the varied components of the interventions, the findings of relationships between increased performance in individual muscle groups and improvement in primary outcome variables provide some insights into possible mechanisms underlying the interventions. In the MUSSEL trial, participants in all 3 groups were experiencing recovery in back extensor endurance and function after surgery, and those in both the experimental and usual care intervention groups were performing some form of low back exercises. This study identified that improvement in lumbar extensor muscle endurance (modified Sorenson test scores) mediated the intervention group–related change in back pain and disability (Oswestry Disability Index scores), indicating that the experimental intervention resulted in a significant benefit in addition to recovery and usual care that was important in reducing back pain and disability. In the STOMPS trial, intervention-related reductions in shoulder pain were mediated by increased muscle torque production in the shoulder external rotators and abductors, but not in the internal rotators or adductors. The infraspinatus and supraspinatus are the 2 primary muscles involved in rotator cuff pathology,30 and our results indicate that strengthening of these muscles is critical to shoulder pain reduction.

The STEPS and PEDALS trials utilized functional activity–based interventions, which increased the capacity for activity-level, locomotor-related primary outcomes. Change in muscle strength either did not explain or was a weak influence on improvement in locomotor function. These interventions were multifaceted and likely worked through mechanisms other than just improved muscle force production. For the STEPS trial, the paretic ankle plantar flexors were the only muscle group with increased torque values that predicted improvement in customary walking speed. Moreover, the soleus was the only muscle to demonstrate change in activation during walking in participants who achieved the largest gains in walking speed in response to the STEPS interventions.31 These findings underscore the role of the ankle plantar flexors in increasing walking speed.32 For the PEDALS trial, we can conclude that improved walk/run speed following the cycling intervention was related to factors other than strengthening gains of the hamstring and quadriceps muscles. An increase in endurance may have been a factor underlying the improvements on the walk/run test, as the participants walked or ran a longer distance more rapidly.

The relationships were strongest in the second STEPS and PEDALS models and most support was demonstrated for hypothesis 2—that changes in the primary outcome variables were significant predictors of the changes in activity and participation outcome variables. Improved capacity for locomotor activity led to (predicted) improved participation, although specific intervention group per se was not a significant explanatory variable in the improved participation. For the STEPS and PEDALS trials, improvement in walking speed and walk/run endurance predicted gains in ADL function and overall health-related QOL, respectively.

The improved activity and participation modestly affected overall QOL for the STEPS trial, whereas no effect was found for the PEDALS trial. In both of the studies with locomotor-related primary outcomes, the intervention improved the participants' functional capacity marginally, but the overall disability remained. For most participants in these 2 studies, the onset of the disability was not acute (occurring years prior to intervention). People generally arrange their lives to participate using the capacities they have available.33 Therefore, a small improvement in capacity did lead to improved (different) participation options, but the overall impact on life satisfaction was small.

For participants in the STEPS trial, only improvement on the mental health subscale of the SF-36 (which primarily reflects mood) was related to increased overall QOL. After stroke, depression negatively influences QOL for many individuals.16 Lai and colleagues16 demonstrated that a home-based exercise intervention for individuals with subacute stroke decreased impairments and functional limitations regardless of baseline depression levels, but QOL was improved mostly in participants with greater initial depression. The authors suggested that a combination of improvements in physical and emotional functioning elicits the greatest enhancement of QOL after stroke. Our findings support this idea.

The lack of a relationship between change in activity and participation variables and overall QOL for the PEDALS trial may relate to the variable selected to represent overall QOL. The satisfaction item of the PODCI is scored by the parent rather than the child. This item was selected because of the wide range in ages and cognitive levels of the children in the PEDALS trial. Change in 600-yd walk/run endurance was related to the total score of the PedsQL, which was self-reported by the children who participated. This finding indicates that improvement in walk/run speed resulted in a positive change in self-reported, overall health-related QOL.

Our results are consistent with findings in the diagnosis-specific literature for all 4 populations. Cross-sectional studies in the individual populations have demonstrated that both pain-related and locomotor-related primary outcome variables are associated with limitations in activity and participation.29,34–36 Recently, intervention trials have begun including activity and participation outcomes and demonstrating improvements with intervention in each of these populations.11,37–39 An important contribution of this study is the documentation of how the changes in activity and participation outcomes were related to the changes in the primary outcome variables. This type of analysis provides insights into the relationships between outcomes and the life realms that are most strongly affected by the interventions. Evaluating these relationships across the 4 studies suggests areas of particular pertinence to different populations. For example, the link between reduced back pain and disability and the ability to carry out life roles for participants in the MUSSEL trial may relate to the fact that the majority of participants in that study were employed. In contrast, social function and ADL performance were the life realms most affected by the STOMPS and STEPS interventions, where the majority of the participants were not employed.

Limitations

The variables included in the mediation and regression models were limited to those measured in the 4 RCTs. It is possible that other variables also mediated or predicted the changes in outcomes. For example, self-efficacy of disease management is known to be related to pain interference40 and QOL41 in people with spinal injury and improves with exercise intervention.42

The muscle performance and activity and participation variables in the STEPS and PEDALS trials did not demonstrate intervention group–related differences. Our analyses in these 2 data sets examined the changes in outcomes over time without regard to treatment group, which limits the conclusions that we can draw regarding the impact of specific interventions. For the STEPS trial, our results reflect the influences of all 4 intervention groups. In our analysis of the PEDALS trial, we included only participants in the intervention group, but the lack of group-related differences suggests that some of the changes in outcome variables could be related to accommodation to the testing environment or a practice effect, as well as to the exercise intervention. Additionally, the sample size used in the PEDALS trial was small, as we included only participants in the exercise group. This sample size would only be adequate to detect a large effect size for the increase in R2 values using our regression analysis model, which may have reduced our ability to identify significant predictors in the models for the PEDALS trial.43,44

Conclusions

For the MUSSEL and STOMPS trials, these secondary analyses established the links among outcomes across the ICF domains. Improvement in muscle performance variables mediated the reduction in the pain-related primary outcomes. Reduction in pain mediated the intervention-related improvements in ability to fulfill life roles and social functions, which, in turn, mediated intervention-related improvements in overall QOL. The relationships among outcomes were not as strong for the 2 studies with locomotor-related primary outcomes (ie, STEPS and PEDALS). Improvement in walking function was related to enhanced participation outcomes for both trials, but the impact on overall QOL was relatively small for the STEPS trial and not significant for the PEDALS trial. A modest improvement in walking function may improve physical activity and participation but not result in a large impact on overall QOL for people with a long-standing disability, particularly if they previously engaged in social activities and life roles that accommodated their functional limitations.

Footnotes

  • Dr Mulroy, Dr Winstein, Dr Fowler, Dr DeMuth, Dr Sullivan, and Dr Brown provided concept/idea/research design. Dr Mulroy, Dr Winstein, Dr Kulig, Dr Fowler, Dr DeMuth, and Dr Brown provided writing. Dr Mulroy, Dr Kulig, Dr Beneck, Dr Fowler, Dr DeMuth, Dr Sullivan, and Dr Brown provided data collection. Dr Mulroy, Dr Fowler, Dr DeMuth, and Dr Lane provided data analysis. Dr Mulroy, Dr Winstein, Dr Kulig, Dr Fowler, and Dr Brown provided project management. Dr Brown provided fund procurement. Dr Mulroy, Dr Fowler, and Dr Brown provided participants. Dr Mulroy, Dr Beneck, Dr Fowler, and Dr Brown provided facilities/equipment. Dr Fowler provided institutional liaisons. Dr Kulig, Dr Sullivan, Dr Brown, and Dr Lane provided consultation (including review of manuscript before submission).

  • The authors acknowledge the Physical Therapy Clinical Research Network (PTClinResNet). The principal investigator was Carolee J. Winstein, PhD, PT, FAPTA, and the co-principal investigator was James Gordon, EdD, PT, FAPTA (both at University of Southern California, Los Angeles, California). Project principal and co-principal investigators were: David A. Brown, PT, PhD, (Northwestern University, Chicago, Illinois); Sara J. Mulroy, PT, PhD, and Bryan Kemp, PhD (Rancho Los Amigos National Rehabilitation Center, Downey, California); Loretta M. Knutson, PT, PhD, PCS (Missouri State University, Springfield, Missouri); Eileen G. Fowler, PT, PhD (University of California, Los Angeles, Los Angeles, California); and Sharon K. DeMuth, PT, DPT, Kornelia Kulig, PT, PhD, FAPTA, and Katherine J. Sullivan, PT, PhD (University of Southern California). The Data Management Center (University of Southern California) is directed by Stanley P. Azen, PhD. The 4-member Data Safety and Monitoring Committee were: Nancy Byl, PT, PhD, FAPTA, Chair (University of California, San Francisco, San Francisco, California), Hugh G. Watts, MD (Shriners' Hospital for Children—LA Unit, Los Angeles, California), June Issacson Kailes, MSW (Western University of Health Sciences, Pomona, California), and Anny Xiang, PhD (University of Southern California).

  • This study was approved by the institutional review boards of Los Amigos Research & Education Institute, Missouri State University, Northwestern University, and the University of Southern California.

  • This study was supported by grants from the Foundation for Physical Therapy (American Physical Therapy Association) to establish the PTClinResNet, the National Institute for Disability and Rehabilitation Research through the Research and Rehabilitation Training Center (RRTC) on Aging With a Disability (grant H133B081002), and the National Institute for Disability and Rehabilitation Research through the Rehabilitation Engineering Research Center (RERC) on Successful Aging With a Disability (grant H133E080024). The PTClinResNet was a clinical research network to evaluate the efficacy of physical therapist practice.

  • Received January 28, 2011.
  • Accepted September 6, 2011.

References

  1. ↵
    1. Field MJ,
    2. Jette AM
    , eds; Committee on Disability in America. The Future of Disability in America. Washington, DC: The National Academies Press; 2007.
  2. ↵
    Disability: beyond the medical model [editorial]. Lancet. 2009;374:1793.
    OpenUrlCrossRefPubMedWeb of Science
  3. ↵
    1. Jette AM,
    2. Latham N
    . Improving the evidence base for physical therapy disability interventions [editorial]. Phys Ther. 2010;90:324–325.
    OpenUrlFREE Full Text
  4. ↵
    1. Jette AM
    . Toward a common language for function, disability, and health. Phys Ther. 2006;86:726–734.
    OpenUrlAbstract/FREE Full Text
  5. ↵
    1. Iezzoni LI,
    2. Freedman VA
    . Turning the disability tide: the importance of definitions. JAMA. 2008;299:332–334.
    OpenUrlCrossRefPubMedWeb of Science
  6. ↵
    International Classification of Functioning, Disability and Health: ICF. 2001, Geneva, Switzerland: World Health Organization; 2001.
  7. ↵
    1. Jette AM
    . Outcomes research: shifting the dominant research paradigm in physical therapy. Phys Ther. 1995;75:965–970.
    OpenUrlAbstract/FREE Full Text
  8. ↵
    1. Winstein CJ,
    2. Pate P,
    3. Ge T,
    4. et al.
    ; for the Physical Therapy Clinical Research Network (PTClinResNet). The Physical Therapy Clinical Research Network (PTClinResNet): methods, efficacy, and benefits of a rehabilitation research network. Am J Phys Med Rehabil. 2008;87:937–950.
    OpenUrlCrossRefPubMedWeb of Science
  9. ↵
    1. Pope AM,
    2. Tarlov AR
    1. Nagi SZ
    . Disability concepts revisited: implications for prevention. In: Pope AM, Tarlov AR, eds. Disability in America: Toward a National Agenda for Prevention. Washington, DC: The National Academies Press; 1991:309–327.
  10. ↵
    1. Kulig K,
    2. Beneck GJ,
    3. Selkowitz DM,
    4. et al.
    ; for the Physical Therapy Clinical Research Network (PTClinResNet). An intensive, progressive exercise program reduces disability and improves functional performance in patients after single-level lumbar microdiskectomy. Phys Ther. 2009;89:1145–1157.
    OpenUrlAbstract/FREE Full Text
  11. ↵
    1. Mulroy SJ,
    2. Thompson L,
    3. Kemp B,
    4. et al.
    ; for the Physical Therapy Clinical Research Network (PTClinResNet). Strengthening and Optimal Movements for Painful Shoulders (STOMPS) in chronic spinal cord injury: a randomized controlled trial. Phys Ther. 2011;91:305–324.
    OpenUrlAbstract/FREE Full Text
  12. ↵
    1. Sullivan KJ,
    2. Brown DA,
    3. Klassen T,
    4. et al.
    ; for the Physical Therapy Clinical Research Network (PTClinResNet). Effects of task-specific locomotor and strength training in adults who were ambulatory after stroke: results of the STEPS randomized clinical trial. Phys Ther. 2007;87:1580–1602.
    OpenUrlAbstract/FREE Full Text
  13. ↵
    1. Fowler EG,
    2. Knutson LM
    . Demuth SK, et al; for the Physical Therapy Clinical Research Network (PTClinResNet). Pediatric Endurance and Limb Strengthening (PEDALS) for children with cerebral palsy using stationary cycling: a randomized controlled trial. Phys Ther. 2010;90:367–381.
    OpenUrlAbstract/FREE Full Text
  14. ↵
    1. Ware JE Jr,
    2. Sherbourne CD
    . The MOS 36-Item Short-Form Health Survey (SF-36), I: conceptual framework and item selection. Med Care. 1992;30:473–483.
    OpenUrlCrossRefPubMedWeb of Science
  15. ↵
    1. Kemp BJ,
    2. Ettelson D
    . Quality of life while living and aging with a spinal cord injury and other impairments. Top Spinal Cord Inj Rehabil. 2001;6:116–127.
    OpenUrlCrossRef
  16. ↵
    1. Lai SM,
    2. Studenski S,
    3. Richards L,
    4. et al
    . Therapeutic exercise and depressive symptoms after stroke. J Am Geriatr Soc. 2006;54:240–247.
    OpenUrlCrossRefPubMedWeb of Science
  17. ↵
    1. Liu CJ,
    2. Latham N
    . Can progressive resistance strength training reduce physical disability in older adults? A meta-analysis study. Disabil Rehabil. 2011;33:87–97.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Bult MK,
    2. Verschuren O,
    3. Jongmans MJ,
    4. et al
    . What influences participation in leisure activities of children and youth with physical disabilities? A systematic review. Res Dev Disabil. 2011;32:1521–1529.
    OpenUrlCrossRefPubMed
  19. ↵
    1. Baron RM,
    2. Kenny DA
    . The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol. 1986;51:1173–1182.
    OpenUrlCrossRefPubMedWeb of Science
  20. ↵
    1. Varni JW,
    2. Seid M,
    3. Kurtin PS
    . PedsQL 4.0: reliability and validity of the Pediatric Quality of Life Inventory version 4.0 generic core scales in healthy and patient populations. Med Care. 2001;39:800–812.
    OpenUrlCrossRefPubMedWeb of Science
  21. ↵
    1. Daltroy LH,
    2. Liang MH,
    3. Fossel AH,
    4. Goldberg MJ
    ; for the Pediatric Outcomes Instrument Development Group, Pediatric Orthopaedic Society of North America. The POSNA pediatric musculoskeletal functional health questionnaire: report on reliability, validity, and sensitivity to change. J Pediatr Orthop. 1998;18:561–571.
    OpenUrlCrossRefPubMedWeb of Science
  22. ↵
    1. MacKinnon DP,
    2. Fairchild AJ,
    3. Fritz MS
    . Mediation analysis. Annu Rev Psychol. 2007;58:593–614.
    OpenUrlCrossRefPubMedWeb of Science
  23. ↵
    1. Flanagan SP,
    2. Kulig K
    . Assessing musculoskeletal performance of the back extensors following a single-level microdiscectomy. J Orthop Sports Phys Ther. 2007;37:356–363.
    OpenUrlPubMedWeb of Science
  24. ↵
    1. Fritz JM,
    2. Irrgang JJ
    . A comparison of a modified Oswestry Low Back Pain Disability Questionnaire and the Quebec Back Pain Disability Scale. Phys Ther. 2001;81:776–788.
    OpenUrlAbstract/FREE Full Text
  25. ↵
    1. Curtis KA,
    2. Roach KE,
    3. Applegate EB,
    4. et al
    . Development of the Wheelchair User's Shoulder Pain Index (WUSPI). Paraplegia. 1995;33:290–293.
    OpenUrlCrossRefPubMedWeb of Science
  26. ↵
    1. Duncan PW,
    2. Wallace D,
    3. Lai SM,
    4. et al
    . The Stroke Impact Scale version 2.0: evaluation of reliability, validity, and sensitivity to change. Stroke. 1999;30:2131–2140.
    OpenUrlCrossRefPubMedWeb of Science
  27. ↵
    1. Fernhall B,
    2. Pitelli KH,
    3. Vukovich MD,
    4. et al
    . Validation of cardiovascular fitness field tests in children with mental retardation. Am J Ment Retard. 1998;102:602–612.
    OpenUrlCrossRefPubMedWeb of Science
  28. ↵
    1. Cieza A,
    2. Geyh S,
    3. Chatterji S,
    4. et al
    . Identification of candidate categories of the International Classification of Functioning, Disability and Health (ICF) for a Generic ICF Core Set based on regression modelling. BMC Med Res Methodol. 2006;6:36.
    OpenUrlCrossRefPubMed
  29. ↵
    1. Gutierrez DD,
    2. Thompson L,
    3. Kemp B,
    4. Mulroy SS
    ; for the Physical Therapy Clinical Research Network and the Rehabilitation Research and Training Center on Aging-Related Changes in Impairment for Persons Living With Physical Disabilities. The relationship of shoulder pain intensity to quality of life, physical activity, and community participation in persons with paraplegia. J Spinal Cord Med. 2007;30:251–255.
    OpenUrlPubMedWeb of Science
  30. ↵
    1. Kim HM,
    2. Dahiya N,
    3. Teefey SA,
    4. et al
    . Location and initiation of degenerative rotator cuff tears: an analysis of three hundred and sixty shoulders. J Bone Joint Surg Am. 2010;92:1088–1096.
    OpenUrlCrossRefPubMed
  31. ↵
    1. Mulroy SJ,
    2. Klassen T,
    3. Gronley JK,
    4. et al
    . Gait parameters associated with responsiveness to treadmill training with body-weight support after stroke: an exploratory study. Phys Ther. 2010;90:209–223.
    OpenUrlAbstract/FREE Full Text
  32. ↵
    1. Richards CL,
    2. Malouin F,
    3. Bravo G,
    4. et al
    . The role of technology in task-oriented training in persons with subacute stroke: a randomized controlled trial. Neurorehabil Neural Repair. 2004;18:199–211.
    OpenUrlAbstract/FREE Full Text
  33. ↵
    1. Kemp BJ,
    2. Mosqueda L
    , eds. Aging With a Disability: What the Clinician Needs to Know. Baltimore, MD: Johns Hopkins University Press; 2004:48–67.
  34. ↵
    1. Abbott AD,
    2. Hedlund R,
    3. Tyni-Lenné R
    . Patients' experience post-lumbar fusion regarding back problems, recovery and expectations in terms of the International Classification of Functioning, Disability and Health. Disabil Rehabil. 2011;33:1399–1408.
    OpenUrlCrossRefPubMed
  35. ↵
    1. Oeffinger DJ,
    2. Tylkowhi CM,
    3. Rayens MK,
    4. et al
    . Gross Motor Function Classification System and outcome tools for assessing ambulatory cerebral palsy: a multicenter study. Dev Med Child Neurol. 2004;46:311–319.
    OpenUrlPubMedWeb of Science
  36. ↵
    1. van de Port IG,
    2. Kwakkel G,
    3. Lindeman E
    . Community ambulation in patients with chronic stroke: how is it related to gait speed? J Rehabil Med. 2008;40:23–27.
    OpenUrlCrossRefPubMedWeb of Science
  37. ↵
    1. Galvin R,
    2. Cusack T,
    3. O'Grady E,
    4. et al
    . Family-Mediated Exercise Intervention (FAME): evaluation of a novel form of exercise delivery after stroke. Stroke. 2011;42:681–686.
    OpenUrlAbstract/FREE Full Text
  38. ↵
    1. Newsome RJ,
    2. et al
    . A prospective, randomized trial of immediate exercise following lumbar microdiscectomy: a preliminary study. Physiotherapy. 2009;95:273–279.
    OpenUrlCrossRefPubMedWeb of Science
  39. ↵
    1. Verschuren O,
    2. Ketelaar M,
    3. Gorter JW,
    4. et al
    . Exercise training program in children and adolescents with cerebral palsy: a randomized controlled trial. Arch Pediatr Adolesc Med. 2007;161:1075–1081.
    OpenUrlCrossRefPubMedWeb of Science
  40. ↵
    1. Pang MY,
    2. Eng JJ,
    3. Lin KH,
    4. et al
    . Association of depression and pain interference with disease-management self-efficacy in community-dwelling individuals with spinal cord injury. J Rehabil Med. 2009;41:1068–1073.
    OpenUrlCrossRefPubMed
  41. ↵
    1. Middleton J,
    2. Tran Y,
    3. Craig A
    . Relationship between quality of life and self-efficacy in persons with spinal cord injuries. Arch Phys Med Rehabil. 2007;88:1643–1648.
    OpenUrlCrossRefPubMed
  42. ↵
    1. Dionigi R
    . Resistance training and older adults' beliefs about psychological benefits: the importance of self-efficacy and social interaction. J Sport Exerc Psychol. 2007;29:723–746.
    OpenUrlPubMed
  43. ↵
    1. Cohen J,
    2. Cohen C,
    3. West SG,
    4. Aiken LS
    . Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. 3rd ed. Mahwah, NJ: Lawrence Erlbaum Associates Inc; 2003.
  44. ↵
    1. Soper DS
    . The Free Statistics Calculator Website, Online Software. Available at: http://www.danielsoper.com/statcalc/. 2011. Accessed July 1, 2011.
View Abstract
Back to top
Vol 96 Issue 12 Table of Contents
Physical Therapy: 96 (12)

Issue highlights

Email
Print
Secondary Mediation and Regression Analyses of the PTClinResNet Database: Determining Causal Relationships Among the International Classification of Functioning, Disability and Health Levels for Four Physical Therapy Intervention Trials
Sara J. Mulroy, Carolee J. Winstein, Kornelia Kulig, George J. Beneck, Eileen G. Fowler, Sharon K. DeMuth, Katherine J. Sullivan, David A. Brown, Christianne J. Lane
Physical Therapy Dec 2011, 91 (12) 1766-1779; DOI: 10.2522/ptj.20110024

Citation Manager Formats

Save to my folders

Secondary Mediation and Regression Analyses of the PTClinResNet Database: Determining Causal Relationships Among the International Classification of Functioning, Disability and Health Levels for Four Physical Therapy Intervention Trials
Sara J. Mulroy, Carolee J. Winstein, Kornelia Kulig, George J. Beneck, Eileen G. Fowler, Sharon K. DeMuth, Katherine J. Sullivan, David A. Brown, Christianne J. Lane
Physical Therapy Dec 2011, 91 (12) 1766-1779; DOI: 10.2522/ptj.20110024
del.icio.us logo Digg logo Reddit logo Technorati logo Twitter logo CiteULike logo Connotea logo Facebook logo Google logo Mendeley logo

Subjects