Using Treatment Fidelity Measures to Understand Walking Recovery: A Secondary Analysis From the Community Ambulation Project

Abstract

Objectives

Physical therapist intervention studies can be deemed ineffective when, in fact, they may not have been delivered as intended. Measurement of treatment fidelity (TF) can address this issue. The purpose of this study was to describe TF of a home-based intervention, identify factors associated with TF, and examine whether components of TF were associated with the outcome of change in 6-minute walk distance (∆6MWD).

Methods

This is a secondary analysis of community-dwelling hip fracture participants who completed standard therapy and were randomly assigned to the active intervention (Push). Push was 16 weeks of lower extremity strengthening, function, and endurance training. TF was defined as delivery (attendance rate, exercise duration) and receipt (progression in training load, heart rate reserve [HRR] during endurance training, and exercise position [exercise on floor]). The outcome was ∆6MWD. Independent variables included baseline (demographic and clinical) measures. Descriptive statistics were calculated; linear and logistic regressions were performed.

Results

Eighty-nine participants were included in this analysis; 59 (66%) had attendance of 75% or greater. Participants walked for 20 minutes or more for 78% of sessions. The average training load increased by 22%; the mean HRR was 35%; and 61 (69%) participants exercised on the floor for at least 75% of sessions. Regression analyses showed that a higher body mass index and greater baseline 6MWD were related to components of TF; 4 out of 5 components of TF were significantly related to ∆6MWD. The strongest TF relationship showed that those who exercised on the floor improved by 62 m (95% CI = 31–93 m) more than those who did not get on the floor.

Conclusions

Measures of TF should extend beyond attendance rate. This analysis demonstrates how measures of TF, including program attendance, progression in training load, endurance duration, and exercising on the floor were significantly related to improvement in 6MWD in participants post hip fracture.

Impact

This careful analysis of treatment fidelity assured that the intervention was delivered and received as intended. Analysis of data from a large trial with participants after hip fracture showed that regular attendance, frequent endurance training for 20 minutes, increases in lower extremity training loads, and exercising on the floor were associated with improvements in the outcome of 6-minute-walk distance. The strongest association with improvement was exercising on the floor.

Introduction

Older adults demonstrate great heterogeneity in functional recovery following hip fracture.¹ Physical therapist interventions associated with hip fracture rehabilitation are complex behaviors that take time, physical effort, and motivation. Systematic reviews have shown that exercise can improve muscle strength and mobility,^2–5 but there is great variability in the types of exercise, the dose prescribed, and participant response to a given exercise program. For example, the dose associated with high-intensity strength training (2–3 times per week for 8–12 weeks) produces large positive effects for muscle strength,⁶ but produces much smaller effects for functional activities in mobility-limited older adults,⁷ such as those post hip fracture. It is important to differentiate if a proposed exercise dose is insufficient to improve function and mobility or if the participants are unable to achieve the dosages intended by the intervention. When “intention to treat approaches” are used, as commonly found in randomized controlled trials, there is limited information related to treatment fidelity (TF) of the interventions tested. Robust measures of TF would begin to address this question.

Of the 5 domains of TF (study design, training, delivery, receipt, and enactment), most literature describes delivery, which focuses on demonstrating that the intervention has been delivered by providers/researchers to participants as intended.⁸ Equally important, but less often described, is intervention receipt, which focuses on the participant and whether the participant understands and can perform the delivered treatment skills,⁸ for example, the participant understands how to do a quadriceps strengthening exercise and is able to perform it. In complex interventions, several aspects of delivery and receipt can be examined such as duration, intensity, and specific type of exercise delivered. In a recent systematic review, intensity was the least likely reported aspect of intervention delivery, whereas attendance was the most commonly reported.⁹ In most exercise trials, attendance is a proxy for adherence, and adherence implies that the intervention was delivered and received as intended,¹⁰ but this assumption is untested.

Demographic factors (eg, age, sex), physical health (eg, gait, balance, pain, and functional status), mental health (eg, mood, resilience, cognitive status), fear of falling, beliefs about ability to engage in exercise activities and its benefits, social support, and the structural environment within communities have been reported to influence participation in regular exercise in older adults.^11–17 In the context of interventions to improve the benefits of exercise, it is important to understand factors that are related to the domains of TF (eg, delivery, receipt) and which domains of TF are associated with specific health outcomes. In doing so, interventions can be revised appropriately so that if delivery is poor, the intervention approach can be adjusted. If receipt is poor, additional motivational interventions may be added, for example, to more fully engage participants.

Interventions might be deemed ineffective when, in fact, they may not have been delivered or received as intended. A recent large, multicenter, randomized controlled trial post hip fracture, the Community Ambulation Project (CAP), failed to show statistically significant differences in community ambulation between participants randomly assigned to strength and endurance training (Push) compared with those randomly assigned to seated range-of-motion exercise and sensory-level electrical stimulation.¹⁸ To gain a better understanding of whether the intervention was ineffective or if participants failed to perform the recommended exercise regimen, a careful evaluation of multiple components of TF was completed. The purpose of the study was to: (1) describe TF (delivery and receipt) of the Push intervention, (2) identify the factors associated with TF, and (3) examine whether components of TF were associated with the outcome of change in 6-minute walk distance (∆6MWD). Gaining a better understanding of TF will help guide therapists to address modifiable factors that can influence TF and improve the ability to optimally implement potentially useful interventions.

Methods

Design

This was a secondary data analysis using a subset of data from CAP, a multicenter randomized controlled trial of older adults who were randomly assigned within 26 weeks post admission to a hospital for hip fracture (clinicaltrials.gov identifier: NCT01783704). CAP was designed to determine the superiority of a specific multicomponent physical therapy program (strength, endurance, and balance—Push) over a nonspecific multicomponent physical therapy program (seated range-of-motion exercises and sensory level stimulation—Pulse). The Push intervention is the intervention of interest because its components are supported in the literature for older adults with and without hip fracture.^2,3,6 Data were collected at 3 clinical sites with randomization beginning September 2013 and final follow-up assessment ending in October 2017. The rationale, design (including the TF training and monitoring plan), and results of this trial have been presented elsewhere.^18,19 The study protocol was approved at all sites, and all participants signed informed consent. This article presents analyses that were not prespecified in the protocol.

Participants

Individuals were eligible to participate if they were aged 60 years and older, ambulatory while living in the community, sustained a minimal trauma, nonpathologic hip fracture with surgical repair, and walked less than 300 m in 6 minutes. Exclusion criteria included medical conditions that precluded safe participation in exercise, body mass index (BMI) less than 17.9 or greater than 40 kg/m², and cognitive impairment.¹⁹ Separate randomization schedules per clinical site were created in randomly ordered blocks, with equal numbers of participants assigned to each treatment within each block. CAP included 210 participants, of whom 105 were allocated to Push. Among the Push participants, 103 received more than 1 visit, and of these, 14 participants did not have a measured 6MWD at follow-up. A total of 89 participants were, therefore, included in this analysis. Based on independent sample t tests for numeric variables and χ² tests of independence for categorical variables, the 89 participants were not significantly different from the 16 excluded on any baseline variables.

The Push Intervention

Push, defined as the training intervention,¹⁹ involved 32 or 40 home-based visits over 16 weeks from a physical therapist. Total visit number was based on the protocol version to which participants were randomly assigned. Participants wore heart rate monitors during all sessions, and the intervention consisted of lower extremity strengthening exercises and a walking endurance program. Specifically, Push included 4 exercises targeting muscles of the lower extremities important for function, namely the hip, knee, and ankle extensors and hip abductors. The leg press and hip abduction exercises were prescribed to be performed in supine on the floor (if able). Enabling and assisting participants to get on and off the floor each session was explicitly built into the exercise program to increase balance, skill, and confidence with this task; otherwise, the same exercises were performed on a bed or couch. Hip extension and plantarflexion exercises were performed in standing. The Shuttle MiniPress (ShuttleSystems, Bellingham, WA, USA) was used to provide resistance (external load) for all exercises except plantarflexion, for which body weight was the resistance load. The participants were supervised during all sessions and encouraged to perform 3 sets of 8 repetitions at an intensity of the 8-repetition maximum for each leg. External load was progressed by adding an additional elastic cord to the exercise and was evaluated every 2 weeks. After the strength training portion, the participants performed 20 minutes of walking, which occurred indoors and outside (weather permitting). Heart rate (HR) was recorded every 5 minutes in the endurance portion. The target intensity was 50% of heart rate reserve (HRR). The actual intensity achieved was calculated by the following formula: (HR_{during endurance training} – HR_{baseline at rest})/((220 − AGE) – HR_{baseline at rest}) × 100.

Data Sources/Measurement

Data were obtained from intervention sessions (TF variables), baseline testing (predictors of TF), and 16-week follow-up testing (outcome: ∆6MWD). Daily treatment forms provided the data for measures of TF. Delivery of the intervention was based on the following measures: (1) attendance rate and (2) exercise duration of endurance training. Attendance was counted when the physical therapists performed a home visit. Attendance rate was calculated as number of visits divided by the total number of sessions the participant was randomly assigned to receive (32 or 40 visits). Attendance was dichotomized to those who did or did not perform 75% or more of the possible sessions. Visual inspection of the distribution showed a distinct break at 75%; below that rate, there was a low number in each decile of attendance (eg, 50%–60%, 60%–70%, 70%–80%, etc) and 75% is comparable to other reported values of acceptable attendance rates.¹⁰ Exercise duration was measured during the endurance training portion of the intervention with the goal for the physical therapist to engage the participant for 20 minutes per session. The number of times that each participant achieved 20 minutes of endurance training divided by the number of visits completed over the 16 weeks was calculated.

Receipt was based on the following measures: (1) progressions in training load, (2) achieving HRR target during endurance training, and (3) attaining proposed exercise position for supine exercises. Progression in the external load for strength training was expected over the 16-week period with twice-weekly training. For simplicity, we chose 1 exercise (hip extension) to illustrate progression. Progression was operationally defined by estimating the increase in the number of elastic bands over the course of the study. To calculate progression, we plotted the number of elastic bands used on the y-axis and visit number on the x-axis. Some sessions were missed, in which case the data were approximated as the number of elastic bands from the prior session. The slope of the line was rescaled to indicate the estimated change over the 16 study weeks by multiplying by 40 (ie, extrapolating all visits out to 40). The slope accurately captures the variable pattern of change over 16 weeks.

HRR was the intensity measure for the endurance component of the study. The target started at 30% HRR and progressed to 50% HRR by week 3, remaining at that level for the rest of the sessions. The mean percentage of the HRR value for all sessions was calculated in each participant. Exercising on the floor was a key feature of the design of the intervention because being unable to rise from the floor is associated with poor recovery and is potentially modifiable with practice.²⁰ Exercise position was calculated as the percentage of sessions in which exercises were completed on the floor divided by the number of sessions completed. Exercise position was dichotomized to those who performed exercises on the floor at least 75% of sessions compared with those who did not. We used the same visual inspection methods as described for attendance rate to guide the decision for the dichotomy.

Predictors of TF were collected at baseline and included demographic variables, BMI, and a variety of self-reported clinical variables. Demographic variables included age, sex, and education level. Clinical measures included a self-reported history of arthritis, back disease, osteoporosis, stroke or transient ischemic attack, peripheral vascular disease, cardiac/pulmonary/neurologic diseases, and psychiatric disorders. The Brief Resilience Scale was used to evaluate self-reported ability to bounce back from a stressful event.²¹ This 6-item test is scored from 1 (low resilience) to 5 (high resilience). Depressive symptoms were evaluated based on the 20-item Centers for Epidemiologic Studies Depression scale.²² It is scored from 0 to 60, with scores of 16 or more indicating risk of clinical depression. The “Pain” subscale score and the “Physical Component Score” were derived from the Medical Outcomes Study Short Form 36 (SF-36). Transformed scores range from 0 to 100, with higher scores indicating a better health state (ie, freedom from pain and higher physical function).²³ Nutritional status was measured by the Mini-Nutritional Assessment—Short Form.²⁴ Scores range from 0 to 14, with higher scores indicating better nutritional status. Cognitive status was measured by the Modified Mini-Mental State examination. Scoring is from 0 to 100, with higher scores indicating less cognitive impairment (in this sample scores ranged from 73 to 100).²⁵

The outcome variable of interest was ∆6MWD from baseline to 16-week follow-up. The test was conducted according to the American Thoracic Society guidelines with an 18-m pathway length.²⁶ The reliability of the aforementioned baseline and outcome measures have been previously reported.¹⁹

Statistical Methods

Descriptive statistics were used to describe the demographic, clinical, and baseline physical performance of the sample as well as components of TF. For categorical data, we report the percentage within each category and the total sample size. For numeric data, we report the mean and SD.

Simple and multiple regression techniques were used for determining relationships with each TF variable (TF models). Linear regression models were employed for numeric outcomes, and coefficient (intercept and slope) estimates were reported along with 95% CIs. Logistic regression was used for dichotomous outcomes, and estimated odds ratios were reported along with 95% CIs. All clinical variables were examined as potential covariates in each TF model.

We also examined variables associated with the primary outcome of the CAP study (Δ6MWD models). Because Δ6MWD is a numeric variable, we used multiple linear regression models. The “demographic” model examines the relationship between the significant variables identified by the TF models with Δ6MWD. We then considered all clinical variables for inclusion into the model using a stepwise (both directions) variable selection procedure. A stepwise procedure considers adding or removing 1 variable at a time. This model was called “base.” We then considered separate models for each TF component including the “base” model on Δ6MWD.

Role of the Funding Source

The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Results

Table 1 shows baseline characteristics of the 89 participants. The sample was predominantly female (76%), with approximately 60% being 80 years or older, and 47% having a BMI greater than 25 kg/m². Of the self-reported comorbidities, arthritis (74%), osteoporosis (38%), psychological conditions (30%), and back disease (29%) were the most commonly reported. At baseline, the average score for pain (SF-36 pain subscale) during the previous 4 weeks was 60 (SD = 25), and 29% showed depressive symptoms on the Centers for Epidemiologic Studies Depression scale. Baseline 6MWD was 189 m (SD =55 m).

Table 1.

Baseline Characteristics of Participants^a

Characteristic (N = 89)	Mean [SD] or Number (%)
Demographic
Age, y	80.2 [8.0]
Age ≥80	53 (60%)
Females	68 (76%)
College degree or higher	32 (36%)
Clinical
BMI, kg/m2	25.2 [4.7]
BMI ≥25	42 (47%)
Presence of comorbidities
Back disease	26 (29%)
Cardiac disease	18 (20%)
Pulmonary disease	12 (13%)
Diabetes	13 (15%)
Neurological	20 (22%)
Arthritis	66 (74%)
Osteoporosis	34 (38%)
PVD	14 (16%)
Psychological	27 (30%)
Stroke	10 (11%)
MNA-SF score	9.0 [2.1]
CES-D score	10.7 [8.0]
CES-D score ≥16	26 (29%)
3MS score	91.7 [6.7]
Brief Resilience Score	19.7 [2.1]
Bodily Pain Index (SF-36)	59.9 [24.9]
Physical Function Score (SF-36 PCS)	36.5 [7.8]
Baseline physical performance
6-Minute Walk Test distance, m	188.7 [54.8]

^a BMI = body mass index; CES-D = Center for Epidemiological Studies Depression; MNA-SF = Mini Nutritional Assessment-Short Form PCS = Physical Component Score; PVD = peripheral vascular disease; SF-36 = Medical Outcomes Study Short Form 36; 3 MS = Modified Mini-Mental State Examination.

Table 2 describes the distribution of delivery and receipt components of TF from this sample. For delivery, attendance of at least 75% of the expected sessions was found in 66% of the sample. Participants were able to perform the full duration of endurance exercise (20 minutes) in 78% of the sessions. Receipt aspects of TF showed that the average increase in load for the fractured leg was 1.3 elastic bands for the hip extension exercise estimated over 40 visits, which equates to approximately 14–18 pounds (6.4 to 8.2 kg) (K.K. Mangione, PT, PhD, unpublished data, 2010). The average HRR for the participants across all sessions was 35%. Sixty-nine percent of the sample exercised on the floor for at least 75% of the sessions.

Table 2.

Description of Components of Treatment Fidelity

Component	Description	Data Type	Distribution
Delivery
Attendance rate	The percentage of all intended home visits the patient actually received	Dichotomous—did the patient attend at least 75% of the intended home visits? (yes/no)	N = 89 • 59 participants (66% of sample) had attendance rate ≥75%
Exercise duration	The percentage of sessions the patient attended that were ≥20 min for endurance training	Numeric	N = 89 • Mean [SD]: 78% [23%] of the sessions were ≥20 min • Range: 0%-100%
Receipt
Progression in training load	The estimated change in number of bands used each session over the course of the 16-wk study	Numeric	N = 88 (1 missing value) • Mean [SD]: increase of 1.3 [1.3] bands • Range: −2.5 to 5.3
Heart rate reserve (HRR) intensity	The mean of reported HRR values over all sessions for the patient	Numeric	N = 87 (2 missing values) Range: 0%-100% • Mean [SD] HRR was 35.2 [18.3]% • Range: −18.4% to 86.2%
Exercise position	The percentage of sessions a patient completed in which exercises were performed on floor	Dichotomous—did the patient exercise on the floor ≥75%? (yes/no)	N = 89 • 61 participants (69% of sample) exercised on floor ≥75%

The TF models showed that BMI and 6MWD at baseline had significant associations with measures of TF delivery and receipt, whereas gender showed an association with 1 component. (Tab. 3) To simplify comparisons between models, we included the following variables in all remaining models. (1) Attendance rate: those with high BMI had 4.3 times the odds of being in the high attendance category (95% CI = 1.5 to 13.6), and for each 50-m increment in baseline 6MWD, participants had 1.8 times the odds of high attendance (95% CI = 1.1 to 3.1). (2) Exercise duration: being female and having a BMI of 25 or greater were associated with a 12.5% (95% CI = 1.6 to 23.4) and 16% (95% CI = 6.7 to 25.6) increase, respectively, in the percentage of sessions they walked at least 20 minutes for endurance training. (3) Progressions in training load: those with higher BMI improved by 1 elastic band (95% CI = 0.5 to 1.6), and for each 50-m increment in baseline 6MWD, participants improved by 0.3 elastic band (95% CI = 0.1 to 0.6). (4) HRR: for each 50-m increment in baseline 6MWD participants had a 4.2% higher HRR (95% CI = 0.7 to 7.7). (5) Exercise position: participants with higher BMIs had 7.2 times the odds of exercising on the floor (95% CI = 2.2 to 28.2), and for each 50-m increment in baseline 6MWD, participants had 3.3 times the odds of exercising on the floor for at least 75% of the time (95% CI = 1.8 to 6.6). The only clinical variable that was statistically significant in the TF models, when controlling for gender, BMI, and 6MWD at baseline, was depression. In the model predicting HRR, participants with depression had an HRR that was 10.2% lower on average than those without depression (95% CI = −18.6 to −1.9). (Tab. 3).

Table 3.

Baseline and Clinical Factors Related to Treatment Fidelity^a

Model	(Intercept)	Female	BMI ≥25 kg/m 2	Baseline 6MWD in 50-m Increments	Depression	R 2  or C-statistic
Delivery
Attendance rate—logistic	0.05 (0 to 0.45)	2.96 (0.97 to 9.54)	4.28 (1.52 to 13.58)	1.83 (1.14 to 3.13)		C = 0.724
Exercise duration—linear	48.4 (28.4 to 68.4)	12.5 (1.6 to 23.4)	16.1 (6.7 to 25.6)	3.23 (−1.1 to 7.5)		R 2 = 0.155
Receipt
Progressions in training load—linear	−0.5 (−1.6 to 0.7)	0.2 (−0.5 to 0.8)	1 (0.5 to 1.6)	0.3 (0.1 to 0.6)		R 2 = 0.165
Heart rate reserve - linear	11.1 (−5.4 to 27.6)	7.9 (−1 to 16.8)	4.6 (−3.3 to 12.4)	4.2 (0.7 to 7.7)		R 2 = 0.101
Heart rate reserve—linear (with depression)	18.4 (1.3 to 35.5)	8.0 (−0.7 to 16.7)	3.5 (−4.1 to 11.2)	3.2 (−0.4 to 6.7)	−10.2 (−18.6 to −1.9)	R 2 = 0.162
Exercise position—logistic	0.01 (0 to 0.1)	3.17 (0.9 to 12.0)	7.2 (2.2 to 28.2)	3.3 (1.8 to 6.6)		C = 0.787

^a Linear models show the slope coefficients (95% CI) and R² whereas the logistic models show the odds ratio (95% CI) and C-statistic.

The Δ6MWD models tested the relationships between baseline variables and components of TF with the primary outcome, Δ6MWD (Tab. 4). The “demographic” model showed that a BMI of 25 or greater was significantly related to greater Δ6MWD. The “base” model identified back disease and pulmonary disease as significant predictors of Δ6MWD; those with the diseases walked less than those without the diseases. In the models that controlled for “base” and considered each TF component, participants who attended at least 75% of sessions walked 38 m (95% CI = 8 to 68 m) more than those who had lower attendance. Those with a higher percentage of sessions with 20 minutes of endurance training walked 0.8 m (95% CI = 0.2 to 1.4 m) more than those with a lower percentage of sessions at 20 minutes. Participants who had an increase of 1 resistance band walked an average of 13 m more (95% CI = 2 to 24 m) than those who did not increase in bands, and those who exercised on the floor walked 62 m (95% CI = 31 to 93 m) more than participants who did not exercise on the floor. HRR intensity was not associated with Δ6MWD.

Table 4.

Treatment Fidelity and Change in 6MWD^a

Model	(Intercept)	Female	BMI	Base 6MWD	Pulm	Back	Att Rate	Ex Dur	Prog Load	HRR	Ex Pos	R 2
Demographics	27 (−36 to 89)	29 (−5 to 63)	37 (8 to 67)	−3 (−17 to 10)	–	–	–	–	–	–	–	0.096
Base	57 (−6 to 119)	25 (−8 to 57)	39 (11 to 67)	−6 (−19 to 7)	−49 (−90 to −9)	−35 (−65 to −5)	–	–	–	–	–	0.199
Base with attendance rate	58 (−3 to 118)	17 (−15 to 49)	29 (0 to 57)	−10 (−23 to 3)	−48 (−87 to −9)	−34 (−63 to −5)	38 (8 to 68)	–	–	–	–	0.256
Base with exercise duration	18 (−49 to 86)	15 (−18 to 48)	26 (−3 to 55)	−9 (−21 to 4)	−49 (−88 to −10)	−35 (−64 to −6)	–	0.8 (0.2 to 1.4)	–	–	–	0.256
Base with progression of load	62 (0 to 124)	22 (−11 to 55)	25 (−6 to 55)	−10 (−23 to 3)	−48 (−89 to −7)	−29 (−59 to 0)	–	–	13 (2 to 24)	–	–	0.238
Base with HRR	57.1 (−6 to 120.2)	20.6 (−12.4 to 53.6)	32.8 (4.2 to 61.4)	−10.3 (−23.7 to 3.1)	−56.5 (−97 to −16)	−33.8 (−63.9 to −3.7)	–	–	–	0.7 (−0.1 to 1.5)	–	0.233
Base with exercise position	72 (14 to 130)	14 (−16 to 45)	20 (−8 to 48)	−17 (−31 to −4)	−53 (−90 to −16)	−29 (−57 to −1)	–	–	–	–	62 (31 to 93)	0.328

^a Data show slope coefficients and 95% CI. The last, row (in bold) shows the results when all TF variables were entered into the base model simultaneously. Exercise position variable was the first one to enter the model and, once it was in the model, no other TF variables were significantly related to ∆6MWD. Att rate = attendance rate; Back = back diseases; Base 6MWD = baseline 6MWD in 50-m increments; BMI = body mass index >25 m/kg²; Ex Dur = exercise duration; Ex Pos = exercise position; HRR = heart rate reserve; Prog Load = progressions in load; Pulm = pulmonary diseases; TF = treatment fidelity; ∆6MWD = change in 6-minute walk distance; 6MWD = 6-minute walk distance.

When all the TF components were considered in the base model simultaneously, exercise position was the first one to enter the model and, once it was in the model, no other TF variables were significantly related to ∆6MWD (bolded row in Tab. 4) Exercise position also had the highest R² value at 0.328. In this final model, those with pulmonary diseases and back diseases improved less (−53 m; 95% CI = −90 to −16 m) and (−29 m; 95% CI = −57 to −1 m) respectively, than those without these comorbidities, and each additional 50-m increment walked at baseline was associated with a reduction (−17 m; 95% CI = −31 to −4 m) in ∆6MWD (Tab. 4 and Figure). Variables not included in the table were not statistically significant.

Figure.

Fidelity and change in 6-minute walk distance models. Baseline 6-minute walk distance (in meters) is on the x-axis and change in 6-minute walk distance is on the y-axis. The dashed line shows no change (0 m). The lines are the loess smoothed curves. Exercise on the floor for 75% or greater of session (yes) is marked by the blue line, and less (no) is marked by the red line. Large markers indicate the presence of back disease, whereas small markers indicate the absence of back disease. Triangular markers indicate the presence of pulmonary disease, whereas circular markers indicate the absence of pulmonary disease. The color (blue or red) additionally represents the frequency with which those participants (with/without pulmonary disease or back disease) exercised on the floor (blue indicates yes, ≥75% of sessions; whereas red indicates no).

Discussion

The findings from this study demonstrate the importance of careful consideration of TF, specifically delivery and receipt of the intervention, in examining the results of a trial. Intervention-specific fidelity measures, as we described, included evaluation of key components of the Push intervention. This is among the few trials to have examined TF in hip fracture rehabilitation.^27,28 In fact, a recent systematic review for older adults with fragility fractures revealed that only 12% of randomized controlled trials reported a complete dose of exercise, which is needed to evaluate TF.²⁹

This study examined both TF delivery and receipt. With regard to delivery, attendance in the Push group was not as high as desired but similar to attendance rates reported in home-based exercise trials with older adults and those post hip fracture.^28,30,31 With regard to exercise duration, participants were able to walk for at least 20 minutes in 78% of the sessions. There was strong fidelity in many of the TF receipt components as well. Progressions in training load for the hip extensors exceeded 1 band, and 69% of the participants got on the floor to exercise regularly. Because progression in training load and exercising on the floor are intervention-specific to the CAP study, comparisons are difficult to make, but these are important aspects of the intervention that impacted outcomes. The participants, on average, did not achieve the intended target of 50% HRR. Of the trials that reported HR during endurance exercise in older adults, the average HRs we noted during training (95 bpm; 35% HRR) were similar to those reported for 85-year-old women post hip fracture enrolled in a home exercise program (88 bpm; 69% of maximal HR),³⁰ and for frail older adults (mean age 83) who had a mean HR of 100 bpm during training.^32,33 We cannot determine whether the intensity goal was too high for older adults after hip fracture or if medication and preexisting cardiac conditions affected the HR goals.

Predictors of TF included baseline 6MWD and BMI. In older adults, 6MWD is related to strength, balance, and age and considered a measure of overall mobility and function.³⁴ Our findings on baseline 6MWD predicting TF are consistent with literature reports that physical health factors are related to participation.^12,14,16,17 Higher BMIs have also been reported to have a positive association with participation in exercise.³⁵ The literature reports a U-shaped relationship between baseline BMI and the risk of developing a major mobility disability in the older adult. Specifically, the risk of developing a major mobility disability (failure to complete a 400-m walk in <15 minutes) for sedentary older adults, 70 to 89 years of age, with a BMI of 25 to 29 kg/m² was half that predicted for those with less than 25 kg/m² or more than 30 kg/m².³⁶ Although the range of 25 to 29 kg/m² is reported as harmful related to mortality in the middle-aged adult, it appears that this range of BMI in the older adult offers a protective effect.³⁷ Our findings provide additional support for this protective finding.

Four components of TF were significantly related to ∆6MWD. Those who attended more sessions, had frequent sessions of 20 minutes of endurance training, had greater progression in load, and exercised on the floor regularly had significantly larger ∆6MWD, suggesting participants in CAP who performed the intervention as intended, had greater improvements in 6MWD. Our data suggest that a woman who walked 188 m at baseline, had a BMI of 26 kg/m², did not report back disease or pulmonary disease, and was able to exercise on the floor for most sessions, would be likely be able to walk 355 m after 16 weeks of the intervention. If that same woman did not exercise on the floor for at least 75% of the session, her predicted walking distance would be 293 m after 16 weeks, a 62-m difference. This difference is considered a meaningful difference³⁸ because even small increases in physical activity performed by older adults can impact morbidity and quality of life.^39,40

An unexpected finding was that exercising on the floor was the most significant contributor to change in 6MWD. Although assisting a patient or providing supervision during the task of getting on and off the floor could be related to the physical therapist–patient therapeutic relationship, therapist confidence, and/or the patient’s self-efficacy or confidence in his or her abilities, it could also be related to more basic physiological status and raises the issue of classifying persons who will be most responsive to the Push intervention. Ardali et al⁴¹ reported that community-dwelling older adults who were independent in getting up and down off the floor were categorized as physically nondisabled, nonfrail, and functionally independent in contrast to those not independent in the activity. The inability to get up from the floor after a fall has been associated with increased risk of hospitalization, poor physical function, health conditions, age, and increased mortality.^20,42,43 These characteristics suggest a dichotomy in ability and present a possibility of tailoring treatments for those who can and cannot perform floor transfers. At a minimum, our findings suggest floor exercise is associated with improved outcome following intervention, and we suggest that clinicians examine the ability to complete floor transfers to guide clinical decision making.

We did not collect detailed information regarding the reasons why Push participants did not perform the intervention as intended, and therefore, it remains unclear. It is possible that they were limited by physical capacity, motivation, environmental constraints, or a combination of factors. Although Push was an intensive intervention that may have been difficult for some to complete, we believe that therapists should encourage full participation to help improve outcomes among patients post hip fracture.

The presence of pulmonary disease and back disease was associated with decreased ∆6MWD and remained significant when TF was considered in the model. Pulmonary disease is strongly associated with decreased endurance,⁴⁴ whereas evidence for exercise effectiveness in older adults with low back pain is weak at best.⁴⁵ Participants with depression were noted to achieve lower HRR than those without depression. Depression post hip fracture is associated with poorer activities of daily living (ADL) and slower walking speed.⁴⁶ Optimizing management of these clinical problems may have a positive impact on the expected recovery in patients post hip fracture.

Limitations

Results of secondary analyses need to be interpreted cautiously. Our measurements of TF were obtained post randomization. The findings are reported for the Push group of the CAP study; we did not include the active control group so our findings may be misleading.⁴⁷ There were challenges in obtaining HR data given the device and interpretation of the findings in light of medications and arrhythmias. Although the data were checked for accuracy,⁴⁸ missing data may have influenced the HRR findings. Finally, caution should be used in interpreting a causal association between TF and improved performance because this association may be due, at least in part, to unmeasured participant characteristics that affect both fidelity and performance.

Conclusion

We described treatment fidelity delivery and receipt of an exercise intervention from the CAP trial. BMI and baseline 6MWD were associated with TF. Attendance rate of 75% or greater, endurance training for at least 20 minutes, progressions in training load, and exercising on the floor were significantly associated with improvements in 6MWD. The strongest relationship was found with exercising on the floor. These findings provide some support that when the CAP Push intervention was delivered and received as intended, walking recovery improved following hip fracture. Measurement and monitoring of TF may improve our understanding of clinical outcomes from exercise trials.

Author Contributions

Concept/idea/research design: K.K. Mangione, R.L. Craik, D.L. Orwig, J. Magaziner, B. Resnick

Writing: K.K. Mangione, M.A. Posner, R.L. Craik, B.A. Beamer, D.L. Orwig, B. Resnick

Data collection: K.K. Mangione, R.H. Fortinsky, B.A. Beamer, D.L. Orwig

Data analysis: K.K. Mangione, M.A. Posner, E.F. Wolff, B. Resnick

Project management: K.K. Mangione, R.L. Craik, R.H. Fortinsky, B.A. Beamer, D.L. Orwig, J. Magaziner

Fund procurement: R.L. Craik, D.L. Orwig, J. Magaziner

Providing participants: K.K. Mangione, R.H. Fortinsky

Providing facilities/equipment: K.K. Mangione, R.H. Fortinsky, J. Magaziner

Providing institutional liaisons: K.K. Mangione, R.H. Fortinsky, J. Magaziner

Clerical/secretarial support: J. Magaziner

Consultation (including review of manuscript before submitting): E.F. Wolff, R.H. Fortinsky, E.F. Binder, J. Magaziner, B. Resnick

Funding

This work was supported by grants from the National Institute of Child Health and Human Development and the National Institute on Aging at the National Institutes of Health (R21 HD043269, R01 AG035009, R37 AG09901 MERIT Award, R01 AG029315, T32 AG00262, and P30 AG028747).

Disclosures

K.K. Mangione reports receiving fees for serving as a data safety monitoring board member for Pluristem and for a National Institutes of Aging trial at the University of Maryland. R.L. Craik is a member of the board of directors of the Foundation for Physical Therapy Research and a member of the APTA Centennial Steering Committee. B.A. Beamer is a Veterans Administration Geriatric Research Education and Clinical Center investigator. J. Magaziner is a member of the American Orthopedic Association Own the Bone Multidisciplinary Advisory Board and a consultant for Novartis, Pluristem, and UCB. The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no other conflicts of interest.