Predictors of Adherence to a Falls Prevention Exercise Program for People with Parkinson's Disease

View the entire article with references and supplemental information on the Wiley Online Library. 
Note:  Reference links embedded in the article below will also take you to the article on the Wiley Online Library.

Return to Table of Contents

Authors:  Natalie E. Allen, Jooeun Song, Serene S. Paul, Catherine Sherrington, Susan M. Murray, Sandra D. O'Rourke, Stephen R. Lord, Victor S.C. Fung, Jacqueline C.T. Close, Kirsten Howard and Colleen G. Canning

Article first published online:   30 JUN 2015 | DOI: 10.1002/mdc3.12208



Long-term benefits of exercise for people with Parkinson's disease (PD) require regular and sustained participation. This study aimed to investigate predictors of adherence to a minimally supervised exercise program designed to reduce falls in people with PD.


People with idiopathic PD who participated in the exercise arm of a randomized, controlled trial were included. Exercises were prescribed three times per week for 6 months. Adherence was defined as the percentage of prescribed sessions participants reported as having undertaken. Potential predictors of adherence included baseline measures of demographic variables, disease severity and duration, falls and fear of falling, pain, self-reported health and quality of life, cognition, physical activity levels, freezing of gait, functional mobility and balance, and knee extensor strength.


The 108 participants included undertook a mean of 72% (standard deviation: 38%) of prescribed sessions. Participants had higher levels of adherence if they had shorter disease duration, less bodily pain, and better self-reported health and quality of life. A multivariate model (including disease duration, severity of bodily pain, self-reported physical well-being, the Frontal Assessment Battery, the Short Physical Performance Battery, and maximum walking time) explained 9% of the variance in exercise adherence, with shorter disease duration and less pain the strongest predictors (both predictors standardized β = −0.2; P = 0.04).


Disease duration and pain are likely to negatively influence exercise participation in people with PD. Given that most of the variance in adherence is unexplained, further work is required to determine other predictors of adherence to long-term exercise programs.

Despite advances in medical management, people with Parkinson's disease (PD) experience mobility decline from early in the disease process.[1, 2] Recent randomized, controlled trials (RCTs) and systematic reviews have provided good evidence that regular, task-specific physical activity can improve walking,[3, 4] balance,[3, 5] and muscle strength[6] and may reduce falls[7, 8] in people with PD. Additionally, regular exercise is associated with better quality of life, physical function, and mobility as well as less caregiver burden.[9] These results suggest that physiological changes are occurring as a result of physical activity. However, optimal long-term benefits from exercise will only be achieved by regular, sustained participation.

Establishing long-term exercise participation in people with a degenerative neurological condition, such as PD, is likely to be challenging. People with PD are less active than people of a similar age without PD,[10] and recent work shows that it may be difficult to instigate long-term increases in physical activity in sedentary people with PD.[11] In the general older population, accountability is a motivating factor for participation in organized physical activity programs.[12] Accountability is likely to be increased when there is more exercise supervision. However, health care systems are often unable to provide high levels of supervision in the long term, resulting in people with PD often being prescribed exercise programs to be undertaken with only occasional supervision from their clinician. This suggests that although clinicians seek to provide optimal exercise programs for people with PD, long-term adherence to these programs may be limited[13] and a greater understanding of what factors influence adherence to long-term exercise among people with PD is required.

Previous work has reported adherence to only fully supervised exercise programs[14, 15] or investigated adherence to a short-term (6-week) semisupervised program.[16] This study therefore aimed to investigate the predictors of adherence to a pragmatic, 6-month minimally supervised exercise program designed to reduce falls in community-dwelling people with PD. We hypothesized that a range of psychosocial, health, and lifestyle factors would influence adherence to the exercise program.

Patients and Methods

Study Design

This study utilized data from the intervention (exercise) group of an RCT[8, 17] of a minimally supervised exercise program. The trial protocol was approved by The University of Sydney Human Research Ethics Committee, and written informed consent was obtained from all participants. The methods relevant to the present study are outlined briefly.


Participants were cognitively intact (Mini–Mental State Examination [MMSE] ≥24) community-dwelling people with idiopathic PD who were over 40 years of age, able to walk independently with or without an aid, and had experienced one or more falls in the previous year or were deemed to be at risk of falling.[8] Participants were excluded if they suffered from an unstable medical condition. All participants were required to gain medical clearance before enrolling and those who were on antiparkinsonian medications were required to be on a stable medication regime for at least 2 weeks before entering the trial.

Exercise Program

Participants undertook the PD version of the Weight-bearing Exercise for Better Balance (PD-WEBB; program, which was provided free of charge, three times per week for 6 months. The PD-WEBB is an individualized, progressive program consisting of 40 to 60 minutes of lower-limb strength and balance exercises, along with evidence-based cueing strategies for participants with freezing of gait (FOG).[8]

Participants attended a monthly exercise class and were prescribed a home exercise program to complete three times per week. The classes included 2 to 6 participants and were run by an experienced physiotherapist. The physiotherapist also conducted 2 to 4 home visits over the 6 months to ensure that each participant could perform the exercises appropriately in their home. When an exercise class was not available in a particular geographical area, participants performed all exercise sessions individually at home, with 8 to 10 of these sessions supervised by a physiotherapist. Thirty-five participants performed home-based exercise only. All participants received a similar number of supervised exercise sessions from the physiotherapist (i.e., 8–10 sessions). Where appropriate, carers were instructed to supervise the home-based exercise. Participants were also provided with an exercise booklet containing written and pictorial instructions, and exercises were reviewed and updated monthly.

Adherence Measures

The physiotherapist recorded participation in class and supervised home exercise sessions. Participants were asked to record all exercises completed without supervision in a log book, including the number of repetitions completed each session for each prescribed exercise. All participants were prescribed 78 exercise sessions (3 times per week for 26 weeks). Adherence was defined as the percentage of prescribed sessions where at least some (i.e., any) of the prescribed exercise was recorded as being completed. Because some participants undertook more than the prescribed number of sessions, percent adherence was calculated using two methods. For one method, the total number of sessions undertaken was calculated (i.e., uncapped data, allowing >100% adherence) and for the other the number of completed sessions was capped at 78 (i.e., capped data, restricting adherence to a maximum of 100%).

Predictor Variables

Data for potential predictors of exercise adherence were collected on entry to the trial in each participant's home. Participants who were taking antiparkinsonian medication were tested while on, approximately 1 hourr after ingesting their usual dose. Potential predictor variables were assessed in a standardized order and, for the purposes of data analysis, were grouped into the following domains: demographic; disease severity and duration; falls and fear of falling; pain; self-reported health/quality of life; cognition; physical activity level; FOG; functional mobility and balance; and knee extensor strength. Demographic information consisted of age and gender. Disease severity and duration were assessed using the UPDRS motor subsection[18] and by the number of years since diagnosis, respectively. Falls and fear of falling were assessed using the Parkinson's Disease Fall Risk Score,[19] which includes weighted contributions from measures of FOG, balance, and knee extensor strength. Past falls were assessed as the participant's report of the number of falls experienced in the previous 12 months. Fear of falling during daily activities was assessed using the Falls Efficacy Scale-International (FES-I) questionnaire.[20]

Pain was measured using a 6-point scale in response to question 21 from the SF-36.[21] Self-reported health/quality of life was measured using the SF-12v2,[22] from which the physical (physical well-being) and mental (mental well-being) composite scores and the SF-6D (health and well-being) were calculated.[23] Disease-specific health-related quality of life was assessed using the 39-item Parkinson's Disease quality of life Questionnaire (PDQ-39).[24] Positive affect was measured using the positive affect subscale of the Positive and Negative Affect Schedule.[25] Cognition was measured using the Frontal Assessment Battery (FAB).[26] Physical activity level was measured using a questionnaire that was later developed into the Incidental and Planned Exercise Questionnaire.[27] This recorded the amount of planned exercise and incidental physical activity performed, as well as the participant's maximum walking time (up to 1 hour). FOG was assessed using the sum of questions 3 to 6 of the Freezing of Gait Questionnaire.[28]

Functional mobility and balance were assessed using fast and preferred walking speed measured over the middle 4 m of a 6-m walkway, the time taken to stand up and sit down five times quickly and the coordinated stability test of balance.[29] The coordinated stability test measures participants’ ability to control their center of mass while guiding a swaymeter around a path at the limits of their stability and is scored based on the number of errors made. The Short Physical Performance Battery was used as a composite score of mobility and balance.[30] Knee extensor strength of the left and right knee extensor muscles was measured in sitting using a strain gauge[31] and reported as the average of both sides.

Statistical Analysis

Regression analysis was conducted using the adherence variable calculated with the total number of sessions undertaken (i.e., uncapped data, allowing >100% adherence). Associations between potential predictor variables and exercise adherence were explored using uni- and multivariate linear regression models. Predictor variables with P ≤ 0.2 from univariate analyses became candidates for inclusion in the multivariate linear regression model. Where there was more than one candidate variable within one domain, or where two candidate variables from the same domain were highly correlated (> 0.7), the variable with the lowest P value from the univariate analysis was entered into the multivariate model. All statistical analysis was conducted using SPSS software (v19; IBM Corp, Armonk, NY).


One hundred fifteen participants were randomized to the exercise program. Seven (6%) did not return their adherence records, so the present analyses include data from the remaining 108 participants (94%). Average age of these participants was 71.6 years (SD, 8.1), 63 (58%) were male, and 104 (96%) were taking medication containing levodopa. Demographic and clinical characteristics for included participants are shown in Table 1.

Table 1. Demographic and clinical characteristics of the study participants at baseline
Characteristic Participants With Exercise Adherence Data (n = 108)

1Mean (SD) scores or n (%) are presented.

2aHigher score better.

Male gender, n 63 (58)
Age, years 71.6 (8.1)
Height, m 1.7 (0.1)
Weight, kg 76.0 (15.8)
Body mass index, kg/m2 26.3 (4.6)
Time since PD diagnosis, years 7.7 (5.9)
H & Y stage, n
 2 31 (28.7)
 3 74 (68.5)
 4 3 (2.8)
on UPDRS motor examination, 0–108 25.9 (8.9)
MMSE, 0–30a 28.6 (1.5)

Participants performed a wide range of the prescribed 78 exercise sessions (range, 1%–191%), resulting in a mean adherence of 72% (SD, 38%). Twenty-nine (27%) participants completed more than the 78 prescribed exercise sessions. When total adherence was capped at 100% (i.e., 78 sessions), participants undertook a mean of 68% (SD, 32%) of prescribed exercise sessions. Some bodily pain was reported by 93 participants (86%), and in 25 (22%), pain and other conditions interfered with performance of exercises, requiring modification of prescribed exercises. When the program was modified, the total number of prescribed exercise sessions remained unchanged. Twenty-four participants (21%) discontinued the exercise program. Ten participants discontinued because of pain, 9 cited personal and/or other unrelated health problems, 2 preferred to do different exercises to those prescribed in the trial, 1 died, and 2 discontinued for unknown reasons.[8]

Univariate linear regression analyses (Table 2) identified that shorter disease duration, less bodily pain, better self-reported physical well-being (SF-12 physical composite score), and better self-reported health and well-being (SF-6D) were all significantly (P < 0.05) associated with higher levels of adherence to the exercise program. There were also trends for milder disease severity, better scores on the FAB and the Short Physical Performance Battery, longer maximal walking time, and faster walking speed at fast pace to be associated with higher levels of adherence (P ≤ 0.2).

Table 2. Descriptive data for predictor variables and results of the univariate linear regression analysis for predicting adherence to the exercise program (n = 108)
Predictor Variable Mean (SD) P Value Coefficient (95% CI) Standardized β

1aHigher score better.

2IQR, interquartile range; CI, confidence interval.

 Age, years 71.5 (8.1) 0.3 −0.42 (−1.31 to 0.47) −0.09
 Male gender, n (%) 63 (58) 0.9 1.23 (−13.43 to 15.89) 0.02
Disease severity and duration
 on UPDRS motor (score/108) 25.9 (8.9) 0.2 −0.54 (−1.35 to 0.27) −0.13
 Disease duration, years 7.7 (5.9) 0.02 −1.47 (−2.68 to 0.26) −0.23
Falls and fear of falling
 PD Fall Risk Score (0–1) 0.4 (0.3) 0.7 −5.86 (−31.43 to 19.72) −0.04
 Number of falls in past 12 months, median (IQR) 2 (3) 0.3 0.05 (−0.05 to 0.14) 0.10
 FES-I (range, 16–64) 32.0 (9.2) 0.3 −0.42 (−1.20 to 0.37) −0.10
 SF-36 severity of bodily pain (range, 1–6) 3.2 (1.3) 0.004 −8.01 (−13.40 to −2.62) −0.28
Self-reported health/quality of life
 SF-12v2 physical composite score (/100)a 42.0 (7.7) 0.02 1.07 (0.16 to 1.99) 0.22
 SF-12v2 mental composite score (/100)a 51.6 (6.6) 0.7 0.24 (−0.87 to 1.33) 0.04
 SF-6D health and well-being (range 0–1)a 0.6 (0.09) 0.03 89.1 (8.11 to 170.05) 0.21
 PDQ-39 (score/100) 28.5 (14.0) 0.6 −0.13 (−0.65 to 0.39) −0.05
 Positive Affect Scale (range, 10–50)a 31.8 (6.7) 0.8 −0.10 (−1.19 to 0.99) −0.02
 FAB (score/18)a 14.1 (2.5) 0.08 2.60 (−0.30 to 5.51) 0.17
Physcial activity level
 Exercise (h/week) 3.7 (3.5) 0.7 −0.34 (−2.39 to 1.71) −0.03
 Maximum walking time (proportion of 1 hour) 0.6 (0.3) 0.2 14.93 (−8.96 to 38.82) 0.12
 Freezing of Gait Questionnaire (Questions 3–6, score/16) 4.2 (4.0) 0.9 0.11 (−1.72 to 1.93) 0.01
Functional mobility and balance
 4 m walk speed (fast) (m/s)a 1.35 (0.38) 0.2 11.30 (−7.71 to 30.31) 0.11
 4 m walk speed (preferred) (m/s)a 0.97 (0.26) 0.3 14.33 (−13.81 to 42.47) 0.10
 Sit to stand time (stand/s)a 0.35 (0.15) 0.3 27.11 (−21.73 to 75.95) 0.11
 Coordinated stability (error score) 16.6 (13.3) 0.4 −0.23 (−0.77 to 0.32) −0.08
 Short Physical Performance Battery (score/12)a 9.8 (2.3) 0.1 2.37 (−0.75 to 5.48) 0.15
Knee extensor strength
 Knee extensor strength, average (kg)a 28.8 (10.0) 1.0 0.02 (−0.71 to 0.74) 0.004

The final multivariate linear regression model (Table 3) included disease duration, bodily pain severity, the SF-12 physical composite score, the FAB, the Short Physical Performance Battery and maximum waking time, and explained 9% of the variance in adherence to the exercise program (P = 0.02). Shorter disease duration and less bodily pain had a small-to-moderate association with exercise adherence and were the only predictors to remain significantly associated with adherence in the multivariate model (standardized β = −0.2; P = 0.04 for both predictors).

Table 3. Results of the multivariate linear regression analysis for predicting adherence to the exercise program (n = 108)
Predictor Variable Coefficient (95% CI) P Value Standardized β

1Overall model adjusted r2 = 0.09.

2aHigher score better.

3CI, confidence interval.

Disease duration, years −1.28 (−2.48 to −0.07) 0.04 −0.20
SF-36 severity of bodily pain (range, 1–6) −6.39 (−12.48 to −0.31) 0.04 −0.22
SF-12v2 physical composite (score/100)a 0.35 (−0.81 to 1.51) 0.55 0.07
FAB (score/18)a 1.31 (−1.64 to 4.27) 0.38 0.09
Short Physical Performance Battery (score/12)a 0.39 (−3.14 to 3.92) 0.83 0.02
Maximum walking time (proportion of 1 hour) 1.52 (−23.90 to 26.94) 0.91 0.01


On average, participants undertook 72% of prescribed exercise sessions (68% if adherence capped to 100%), with a wide range of individual adherence. A multivariate model explained only 9% of the variance in adherence. Adherence to the program was higher in people with PD who had shorter disease duration, less bodily pain, better perceived physical well-being, or better perceived overall health and well-being. Of these, shorter disease duration and less pain were the strongest predictors of exercise adherence.

Our findings relating to overall adherence, the large variation in adherence levels, and the predictors of adherence from this 6-month trial are comparable to those reported in a 6-week trial of a semisupervised home-based exercise intervention for people with PD.[16, 32]Pickering et al.[16] reported that participants completed 79% of all prescribed repetitions, whereas in the present trial 72% of prescribed exercise sessions were undertaken (uncapped data for both trials). The lower adherence in the present trial is likely to be influenced by the longer program duration and lower levels of supervision, with 13% of sessions supervised in the present trial compared with 18% in the 6-week trial.[32] Results from the 6-week trial indicated that higher adherence was associated with lower age, reduced disease severity, better balance, and better perceived health and disability, including less pain.[16] Whereas age did not influence adherence in the present trial, the other predictive factors were similar, suggesting that these factors continue to influence exercise participation in the longer term.

Adherence is likely to be influenced by many factors, including the level of supervision, the setting for the exercise (e.g., group vs. individual or home vs. facility), and the type of exercise performed. A fully supervised, 24-week group exercise program where participants undertook tai chi, strength training, or stretching exercises reported an overall adherence of 77%,[7] which is higher than the 68% adherence (capped data) in the present trial. High levels of supervision may increase adherence owing to increased accountability to attend the program.[12]However, in older adults, evidence suggests that whereas center-based programs may enhance adherence in the short term, there may be better long-term adherence to home-based programs.[33] Taken together, this information suggests that adherence and factors influencing it are likely to be complex and multifaceted. It also seems likely that individual preferences may influence adherence, and factors that may improve adherence in one individual may reduce adherence in another. This complexity and potential for individual variation may have contributed (in part) to the low level of variance in adherence explained in the present study.

In the current study, shorter disease duration and less bodily pain were found to be independent predictors of adherence in the multivariate model. Given that the average disease duration of participants in this study was 7.7 years, this suggests that minimally supervised exercise is more likely to be adhered to by people with PD in the first few years after diagnosis. Furthermore, the RCT[8] from which these data came showed that the exercise program was effective in reducing falls in people with milder disease, but not more-severe disease. Taken together, these results suggest that minimally supervised exercise may be an effective, sustainable model for falls prevention in the early stages of PD, whereas those with longer-standing disease may require a different, more closely supervised approach to exercise.

Pain is a common, problematic nonmotor symptom among people with PD. Pain has been reported by up to 85% of people,[34] and, after adjusting for osteoarticular comorbidities, people with PD are twice as likely to experience chronic pain as people without PD.[35] Given the unpleasant nature of pain, and the potential for it to be exacerbated by exercise, it is not surprising that the presence of bodily pain had an impact on exercise adherence. Effective treatment of pain could therefore improve adherence to exercise. However, the relationship between pain and exercise is complex.[35, 36] There is evidence from animal and human studies that exercise may stimulate neuroplastic changes and activate dopaminergic and nondopaminergic pain inhibitory pathways. Exercise could therefore improve the central processing of pain in people with PD, leading to improved pain modulation.[36] Further research is required to understand the effect of exercise on pain and the relationship between pain, pain management, and exercise adherence in people with PD.

The multivariate model in the current study explained only 9% of the variance in exercise adherence. The inclusion of other factors known to influence exercise participation may have improved the predictive value of this model. For example, it has been reported that low outcome expectations and lack of time,[37] as well as nonmotor impairments (e.g., apathy),[38] are barriers to exercise in people with PD, whereas high levels of exercise self-efficacy (i.e., personal belief that barriers to exercise can be overcome) are associated with regular exercise participation.[39] Age, gender, past falls, concern about falling, positive affect, regular physical activity levels, FOG, and knee extensor strength were not associated with adherence in the present trial. This result is surprising, particularly for age, given that age has previously been shown to influence adherence to exercise in both the general older population[40] and in people with PD.[16] Additionally, a positive affect suggests feelings of greater interest and enthusiasm, yet participants with higher scores for positive affect were no more likely to adhere to the exercise than those with lower scores. Further investigation is required to determine the relative importance of these and other potential predictors of adherence, as well as the most effective interventions, to promote long-term exercise among people with PD.

A limitation of the present study was the reliance on exercise logs kept by participants for the home-based component of the program. The log books consisted of a separate page for each exercise, requiring participants to mark a grid for each exercise completed. Though many participants kept thorough records, some reported that they did not always mark off every exercise grid when they performed their exercises, and others may have marked the grid when they had not actually completed the exercise. Our adherence score was calculated as the percentage of sessions where there was report of at least some exercise completed in that session. We therefore may have overestimated the amount of exercise undertaken by some participants. Furthermore, because participation in this trial was voluntary, participants were people who were willing to undertake the 6-month exercise program, which may have also increased exercise adherence. Nonetheless, our use of all available data, including where participants discontinued the exercise program, means that the data presented here are likely to reflect the overall adherence that could be expected from similar people with PD in a clinical setting. Future work is required to compare measures of self-report with activity levels measured by a device (e.g., activity monitor or in-built participation monitoring in exergames) to ascertain the accuracy of self-report records.

The selection of participants for this trial may have influenced the results. Results from the present trial indicate that participants with better cognition had a trend toward better adherence to the exercise program, whereas past falls and fall risk score were not associated with adherence. However, all participants were volunteers without marked cognitive impairment who had fallen in the past 12 months or were judged to be at risk of falling on physical assessment. The possibility that both cognitive status and falls history may influence adherence and factors predicting adherence to exercise in people with PD warrants further investigation.

In conclusion, the present study suggests that people with PD who were recently diagnosed and who experience little or no bodily pain are likely to adhere to a minimally supervised exercise program designed to reduce falls. Most of the variation in adherence to the exercise program remained unexplained; therefore, further work is required to determine the role of other potential predictors of adherence to long-term exercise programs.


Leave a Comment

The following required items were not provided or are in the wrong format. Please provide the required responses and submit again:

Comment Title
Comment: 1000 characters
  [[put error message here]]



We use cookies to give you the best possible experience with our website. These cookies are also used to ensure we show you content that is relevant to you. If you continue without changing your settings, you are agreeing to our use of cookies to improve your user experience. You can click the cookie settings link on our website to change your cookie settings at any time. The MDS site uses multiple domains, including and This cookie policy only covers the primary domain. Please refer to the MDS Privacy Policy for information on how to configure cookies for all other domains on the MDS site.
Cookie PolicyPrivacy Notice