Physical Therapist Involved Physical Activity Interventions: A Systematic Review and Meta-Analysis

Abstract

Background and Purpose:

Little is known about the delivery of physical therapy services to promote physical activity (PA) for adults with neurologic conditions. This systematic review aimed to: (1) identify the roles of physical therapists (PTs) in the delivery of PA interventions, (2) describe PA intervention types, and (3) explore the efficacy of physical therapy PA interventions.

Methods:

This systematic review included randomized controlled trials of interventions including physical therapy, PA or exercise, and PA health promotion in adults with neurologic conditions. Database searches included: PubMed MEDLINE, Embase, CINAHL, PsycINFO, and CENTRAL. Meta-analyses were performed for accelerometry-based and self-report measures of PA; meta-regressions explored moderators.

Results:

Sixty articles met the inclusion criteria, representing 54 unique trials with 15,874 participants. Most articles had a moderate quality of reporting (85%) and a high risk of bias (RoB) (90%). PA interventions led to small, but significant effects on direct measures of PA (effect size [ES] = 0.18, CI_95%, 0.01, 0.35). Small-moderate effects were present when articles with high RoB were excluded (n = 6, ES = 0.36, CI_95%, 0.05, 0.67). Meta-regressions suggested that interventions including both PA and behavior change techniques were more positive than either intervention alone (β = 0.43, CI_95%, 0.23, 0.62).

Discussion and Conclusions:

PTs used direct and supervisory roles to deliver effective PA interventions, particularly using interventions combining PA and behavior change techniques. However, heterogeneity of PA measures limits the interpretation of the meta-analyses. Further research should develop consensus on feasible direct measures of PA outcome measures to use in research and clinical practice for adult neurologic populations.

Video Abstract

Video Abstract.

Download video file ^{(45.8MB, mp4)}

INTRODUCTION

Physical activity (PA) supports health and well-being for adults with neurologic conditions.^1,2 Yet, engagement in PA is consistently low in stroke,³ spinal cord injury (SCI),⁴ multiple sclerosis (MS),^5,6 Parkinson’s disease (PD),^7-9 and traumatic brain injury (TBI).^10,11 Barriers to PA include personal factors (eg, time), environmental access, and condition-related mobility limitations (eg, balance).^12-17 Decreased PA leads to secondary conditions that can further restrict participation, leading to poor health outcomes.^18-23

Physical therapy aims to restore and maintain movement and function, which includes promotion of PA.^24-27 Physical therapists (PTs) are well equipped to address barriers to PA in neurologic conditions throughout the continuum of care.^28,29 PTs can address barriers by tailoring exercise, monitoring PA, and using behavioral change techniques (BCTs) such as motivational interviewing, goal setting, and action planning.^2,30-34

Despite this alignment, the use of physical therapy to deliver PA interventions is not routine in neurologic PT practice due to limited evidence and resources in the workplace, community, or health systems.^25,35 A lack of clarity among referrers and clients surrounding PTs’ roles related to delivering PA interventions is also a challenge.^29,36-38 Prior reviews support PA programs that are community-based,³⁹ volunteer-led,⁴⁰ and primary care-led⁴¹ for sedentary adults and people with cardiometabolic conditions, but many neglect inclusion of neurologic conditions or mobility limitations. The purpose of this systematic review was to describe and explore the role and efficacy of PA interventions provided by PTs for adults with neurologic conditions. This review synthesizes the roles assumed by PTs within interdisciplinary teams to develop or deliver PA interventions, the approaches used to engage people with neurologic conditions in PA, and the efficacy of PA interventions provided by PTs.

METHODS

Search Strategy, Article Selection, and Eligibility

The search strategy for this systematic review (PROSPERO record CRD42016033891)⁴² included articles at the intersection of 4 concepts: physical therapy modality or PTs; PA or exercise; nervous system diseases; and behavior change related to PA health promotion, behavioral interventions, and activity coaching (see Supplemental Digital Content Tables S1 to S6, available at: http://links.lww.com/JNPT/A560 for details). The databases searched were PubMed MEDLINE, Embase (embase.com), CINAHL (Ebsco), PsycINFO (Ebsco), and Cochrane Central Register of Controlled Trials (CENTRAL; Wiley) from inception to June 2023. To increase search relevancy, we added search terms and limited the search to randomized controlled trials (RCTs). Two reviewers independently screened articles (EndNote⁴³/Rayyan⁴⁴) using a 3-stage approach (title, abstract, and full text). A third reviewer facilitated consensus when needed. We identified additional articles using the reference lists of included articles and relevant reviews, reporting results using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA).⁴⁵

We included RCTs, adults (≥18 years), and adult-acquired neurologic conditions commonly treated by neurologic PTs, such as stroke (including transient ischemic attack), TBI, SCI, MS, PD, Huntington’s disease (HD), dementia, Guillain–Barré Syndrome, and peripheral neuropathy. Table 1 presents the inclusion criteria operational definitions for: PT involvement; a primary or secondary goal to increase PA beyond therapy; and PA assessment using direct (accelerometry/activity monitors) or self-report measures. We excluded non-English/Spanish articles; non-randomized or quasi-randomized designs; trials with goals other than improving PA using self-report or direct measures; trials limited to research-provided exercise classes, interventions focused on symptom management, or functional restoration or non-PA health promotion (eg, sleep, smoking, nutrition); no comparison between the PA intervention and a control group on PA outcomes; and conditions with unique considerations due to the high prevalence of both central and peripheral sources of dysfunction (eg, vestibular syndromes, low back pain).

Table 1.

Operational Definitions of Key Intervention Inclusion Criteria

Inclusion Criteria	Definition	Examples
PT involvement	PT was involved in	Included:
	intervention development, outcome assessment, intervention direction/supervision, implementation of an intervention, and interpretation of data	PT developed the intervention as described in the methods or indicated by PT authorship. PT assessed participants for appropriateness of intervention and/or outcomes. PT trained and supervised the intervention being delivered by unlicensed personnel. PT directly implemented the intervention. PT was involved as an author, even if not discussed as a part of the intervention, suggesting they had a role in project development or interpretation of data.
Intervention to increase PA	A cognitive and/or behavioral intervention with a stated purpose to increase the participant’s physical activity on their own time, outside of the prescribed intervention.	Included
		Primary purpose: an exercise intervention performed in a clinic setting with instruction and goal setting toward increasing PA behaviors at home. Secondary purpose: a fall prevention program that included a progressive walking component where the participant was trained to track their steps and increase their time or number of steps taken on their own.
		Excluded
		No PA purpose: high-intensity, clinic-based gait training programs with the goal to increase gait speed and no primary or proxy measures of PA. PA purpose interpreted as compliance with the researcher-provided exercise class and not individual PA behavior on participants’ own during or following the intervention
PA was assessed	There was a direct accelerometer-based measure or self-report measure of PA compared to a control group	Included:
		PA questionnaire Exercise/PA diary or log Accelerometer-based activity monitor.
		Excluded:
		Purpose of the PA intervention was to improve mood, and the only self-report outcome measures were depression and stress. Stated PA purpose, but only reported measures of physical function, including 6-min walk test, Modified Rankin Scale, or similar. Assessed PA in a randomized crossover design trial where the PA values were not compared between intervention and control.

Data Extraction

Two reviewers extracted the data for each article, including participant number, age, diagnosis, acuity-level, PT role, intervention type and setting, dose, follow-up, comparators, PA outcome measures, and results. PT roles were coded as author only; delivery of the PA intervention; and interdisciplinary teams in which PTs trained trainers (non-licensed exercise professionals) to deliver the intervention. Interventions were categorized as physical PA only, BCT only, and combined PA and BCT. Physical PA intervention subtypes included exercise participation, general exercise instruction/encouragement, and task-specific training (eg, balance or gait training). We defined BCT interventions as those focused on changing PA behaviors outside of therapy. These included motivational interviewing, health coaching, goal setting, addressing personal barriers and facilitators, self-efficacy, action planning and applying specific behavioral change theories.^46-52 Control or comparator groups were categorized using Cochrane criteria as active control (eg, different education or intervention) or inactive control (eg, usual care).⁵³

Risk of Bias and Reporting Quality

Risk of bias (RoB) for each article’s primary PA outcome was assessed using the Cochrane Risk of Bias 2.0 by one reviewer and discussed with a second.^54,55 RoB 2.0 evaluates bias arising from 5 domains: the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported result. Each domain is judged as low, some concern (moderate), or high.

Two reviewers independently assessed reporting quality using the Physiotherapy Evidence Database (PEDro) Scale, a validated tool for bias assessment.^56-59 A third reviewer resolved scoring discrepancies. PEDro items 4 (baseline group similarity) and 8 (attrition <15%) were scored specifically in relation to PA outcome measures. PEDro scores below 4 reflect poor quality, 4 to 5 fair, 6 to 8 moderate or good quality, and 9 to 10 high or excellent quality.^60-62

Synthesis and Analysis

We used descriptive statistics and narrative synthesis to summarize PT roles, then completed separate meta-analyses of accelerometer-based direct measures of PA and self-report measures of PA. If multiple PA outcomes of the same type were reported, we prioritized the measure most common in our dataset (eg, steps/day rather than MVPA minutes) to reduce heterogeneity.⁶³ Effect sizes (ESs) reflect mean change scores and standard deviations from baseline to immediately post-intervention; when the standard deviation of change from baseline was not reported, we estimated the standard deviation assuming a correlation of 0.7, based on the data that were reported.⁶⁴ We excluded follow-up assessments due to variability in trial timelines. We estimated the mean difference between PA treatment and control groups using a random-effects meta-analysis model and DerSimonian–Laird’s method to estimate between-trial variance (τ²).⁶⁵ DerSimonian–Laird is a deviation on the inverse-variation method based on the assumption that trials estimate related intervention effects. It takes into account the variation between intervention effects by using the standard errors of trials.⁶⁶ Sensitivity analyses assessed robustness of the pooled estimates by (1) restricting to low/some-concern RoB, (2) restricting to moderate to high quality reporting (PEDro ≥ 6), and (3) restricting to true inactive controls (eg, best medical care with no therapy or additional education). The 95% confidence intervals are presented to reflect the precision of the pooled mean effect, while the 95% predictive intervals are presented to quantify between-study heterogeneity. Predictive intervals are reported for the main and sensitivity meta-analyses to demonstrate the potential effects of a future trial.⁶⁷ When a predictive interval crosses 0, a future trial may result in a positive or negative effect. Funnel plots are reported for primary meta-analyses with ≥10 articles, as recommended for reliable interpretation.^68,69

We completed meta-regressions to identify sources of heterogeneity by adding article-level fixed effects to the model described above. Seven potential univariable moderators were identified a priori based on consensus of the authors on clinical variables of interest, and 2 additional variables were added (see Table 2 for variables and categories).^68,70,71 Across the 2 outcome measure types, 18 meta-regressions were conducted; thus, α was Bonferroni-corrected from .05 to .0028.⁷² Analyses used R 4.2.1 (meta, metafor).⁷³

Table 2.

Meta-Analysis Methods: A Priori Moderator and Additional Variables of Interest

A Priori Variable (Unit of Measure)	Description	Categories
Intervention duration (wk)	Short-term time of intervention	<12, 12, and >12 wk
Diagnosis (category)	Neurologic condition(s)	Stroke, cardiovascular accident/transient ischemic attack; neurodegenerative diseases: Parkinson’s disease (PD), dementia; Huntington’s disease (HD); multiple sclerosis (MS); other: mixed neuro; spinal cord injury (SCI)
Acuity (mo)	Time post diagnosis	Acute–subacute, 0-6 mo, and varied acute through chronic; chronic, >6 mo
Impairment (category)	Level of functional ambulation status or disease-specific rating scales	Mild, ambulatory w/o aide/device; mild–moderate, may need aide/device but independent; moderate–severe, may need assistive device/assistance or severely limited ambulation with assistance to nonambulatory
PA as outcome measure (category)	Degree of outcome measure focus on PA	Primary; secondary
Intervention type (category)	Intervention focus to increase PA	PA only, only PA or exercise focused intervention; BCT only, only behavior change technique(s) focused; combined PA and BCT, both interventions in some capacity
Direct measures of PA (category)	Accelerometer based	Wrist worn accelerometer total number of threshold activity counts per day (Activity_Counts_Day); average % of time stepping per day (Time_Step); number of steps per day (Steps_Per_Day); time (normally reported as minutes)/duration spent in MVPA per day or per week (MVPA); total vector magnitude PA threshold counts per day (VM_Counts); and total wheeled wheelchair minutes per day (WC_Min_Day)
Self-report measures of PA (category)	Self-report questionnaire	MET minutes per week (MET_Min_Week); Physical Activity Scale 2.1 (PAS2); Physical Activity Scale for the Elderly Questionnaire (PASE); Physical Activity Scale for Individuals with Physical Disability (PASIPD); Human Activity Profile—Adjusted Activity Score (HAP-AAS); Physical Activity and Disability Scale (PADS); Longitudinal Ageing Study Amsterdam Physical Activity Questionnaire (LAPAQ); International Physical Activity Questionnaire (IPAQ); Moderate intensity exercise reported in IPAQ (Mod_IPAQ); Godin Leisure-Time Exercise Questionnaire (GLTEQ); and Leisure Time Physical Activity Questionnaire for SCI (LTPAQ)
Settinga (category)	Intervention delivery setting	Clinical: Intervention was initiated in 1 or more of the following settings: hospital (inpatient), inpatient rehabilitation, outpatient physical therapy, home health, or across multiple phases of the rehabilitation continuum
		Community: Intervention was delivered through 1 or more of the following settings: a fitness center, a home-based program, or a community-based research initiative
Control group typea (category)	Inactive control 1, 2, and active control 3, 4 groups	(1) Best medical care, includes waitlist control and physician recommendations only; (2) usual care PT or standard PT, referred to as inactive control but with PT clinical intervention; (3) education, structured provision of information which may include PA or other health management benefits (eg, pamphlets, group information sessions) without therapeutic engagement. May be labeled as inactive control in some trials; (4) active PA or exercise, alternative PA or exercise intervention

Abbreviations: BCT, behavior change technique(s); MET, metabolic equivalent; MVPA, moderate to vigorous physical activity; PA, physical activity.

^aVariables added post hoc.

RESULTS

Article Selection

The search identified 14,350 records, with 146 added from reference lists and review articles (Figure 1 PRISMA). After de-duplication, 9,469 remained; 398 full texts were reviewed; and 60 articles met inclusion criteria, representing 54 unique RCTs (hereafter referred to as trials, as some RCTs had multiple publications from the same dataset).^63,74-132 Forty-one articles contributed to meta-analyses.^{63,75-78,80,83,85,87-96,100,102,107-110,112,114-116,118-120,122,123,125-132}

Figure 1.

PRISMA flow diagram with reasons for exclusions.

Participants, Physical Therapist Roles, Trial, and Intervention Characteristics

Across trials, 7815 participants with neurologic health conditions engaged in PA interventions and 8059 were controls (total = 15,874). Trial sizes ranged from 14 to 6024 participants; median group sizes were 33.0 (intervention) and 31.0 (control). Across diagnoses, participant totals were: 13 408 participants with stroke; 816 with PD; 661 with MS; 326 with SCI; 100 with dementia; 74 with TBI; 46 with HD; and 443 with mixed neurologic conditions. Table 3 summarizes data extracted from each trial with emphasis on PT roles, intervention types, categorized BCTs, and outcomes (additional detail in Supplemental Digital Content Table S7, available at: http://links.lww.com/JNPT/A569).

Table 3:

Data Extraction Table Summarizing PT Roles, Interventions, and Outcomes of 54 Trials

First Author, Year Diagnosis	PT Role	Intervention		PA Outcome Measures *Significant difference between groups
		Physical Activity and/or Education & Encouragement	Behavior Change Techniques: Theory-Informed, Coaching, MI, Self-efficacy, Goals, Planning, Barrier & Facilitators, Feedback, Accountability
Stroke
Ashizawa, 2021	PT Author	EE	Goals, Planning, Barrier & Facilitators	Total PA METS, MVPA METS, Sedentary time
Acute Stroke
Askim, 2018	PT Author	PA	Coaching, MI, Goals, Planning, Barrier & Facilitators	IPAQ*, Training diary*, PT-reported meeting PA goals
Subacute Stroke	PT delivered PA and BCT intervention
Batchelor, 2012	PT Author	PA, EE		HAP-AAS, Diary of Home Exercise Program adherence
Subacute Stroke	PT delivered PA intervention
Boysen, 2009	PT Author	PA, EE	Barrier & Facilitators, Accountability	PASE
Subacute Stroke	PT delivered PA and BCT intervention
Brauer, 2022	PT Author	PA	Theory (HAPA), Self-efficacy, Goals, Planning, Barrier & Facilitators	Steps/day*
				PASIPD
Acute Stroke	PT delivered PA and BCT intervention
Danks, 2016	PT Author	PA, EE	MI, Goals, Barrier & Facilitators, Feedback, Accountability	Steps/day, Time spent walking
Chronic Stroke	PT delivered PA (gait training) and BCT
Dean, 2012	PT Author	PA, EE		Steps/day
Chronic Stroke	PT performed eligibility and PA intervention
Dorsch, 2015	PT delivered PA intervention and feedback	PA, EE	MI, Goals, Feedback	Time spent walking
Acute/Subacute stroke
Duncan, 2020	PT Author	EE	Coaching, Goals	Meeting PA goals
Acute Stroke	PT assisted with COMPASS development, not delivery
Duncan, 2011	PT Author	PA		Steps/day
Acute/Subacute Stroke	PT delivered PA (gait) intervention
Harrington, 2010	PT Author	PA, EE	Goals, Accountability	FAI
	PT supported exercise instructors
Chronic Stroke
Kanai, 2019	PT Author	PA, EE	Self-efficacy, Goals, Feedback	Steps/day*, MVPA
(Kanai, 2018)	PT delivered therapy + BCT
Acute Stroke
Kirk, 2014	PT Author/funding	PA, EE		PA Self-Assessment Questionnaire
Acute Stroke
	PT usual cardiac rehab care
Kono, 2013	PT delivered PA and BCT intervention	PA, EE	Planning, Feedback	Steps/day*
				Low, moderate*, and high PA min/day
Acute Stroke
Kottink, 2007	PT Author	PA		Time spent walking
Chronic Stroke	PT recruitment
Lawrie, 2018	PT Author	PA, EE		Smart watch accelerometer
Acute/ Subacute Stroke	Unclear who delivered
Mansfield, 2015	PT Author	PA, EE	Goals, Feedback	Steps/day, Time spent walking
Subacute Stroke	PT delivered BCT feedback
Mansfield, 2018	PT Authors	PA		PASIPD
Chronic Stroke	PT delivered PA intervention
Meester, 2019	PT Author	PA		PASE, Steps/day
Subacute- Chronic Stroke
Moren, 2016	PT Author	EE	MI, Planning	MVPA Steps/day
Acute Stroke	PT developed and delivered PA intervention
Mudge, 2009	PT Author	PA		Steps/day, Steps/min, time inactive, PADS
Chronic Stroke	PT & students lead PA intervention
Olney, 2006	PT Author	PA		HAP-AAS*
Acute- Chronic Stroke
Pandian, 2023	PT Author	EE	MI	MET min/week
Acute/ Subacute Stroke	PT developed intervention educational materials
Pang, 2005	PT Author	PA		PASIPD, MET hour/day
Chronic Stroke	PT delivered PA intervention
Sivertsen, 2022	PT Author	PA		Moderate PA min/day
				Vigorous PA min/day
Acute/ Subacute Stroke	PT delivered PA intervention			Steps/day
Steen Krawcyk, 2019	PT Author	PA, EE		Steps/day, PAS2*, PA diary
	PT designed, assessment, but not delivery
Acute Stroke
Thompson, 2024	PT Author	PA	MI, Goals, Feedback	Steps/day*
	PT delivered intervention
Chronic Stroke
Vanroy, 2019	PT Author	PA, EE	Theory (TTM, SDT), Coaching, MI, Goals, Barrier & Facilitators, Accountability	Steps/day, Energy, Diary, PASIPD, Baecke
Acute-Subacute Stroke	PT delivered PA (physical therapy) + BCT
Spinal Cord Injury
Froehlich-Grobe, 2022	PT Author	PA, EE	Theory (SCT, RP), Coaching, MI, Self-efficacy, Goals, Feedback	Walk/wheel time (min/week), Moderate PA min/week, Vigorous PA min/week, IPAQ
Subacute to Chronic SCI
Kooijmans, 2017	No PT Authors	EE	Theory (TTM, TPB), MI, Self-efficacy, Barrier & Facilitators	Wrist activity, PASIPD
	PT trained provided BCT
Chronic SCI
Nooijen, 2016	PT Author	PA, EE	Theory (TTM, TPB), Coaching, MI, Self-efficacy, Goals, Planning, Barrier & Facilitators	Wheeled PA min/day*, PASIPD*
Subacute SCI	PT BCT coaching
Ma, 2019	Intervention co-developed by PTs, but delivered by trainer	PA, EE	Theory (HAPA), Goals, Planning, Feedback, Support/ Account	LTPAQ*, MVPA Wrist activity (vector magnitude counts)*
Chronic SCI
Thomas, 2011	PT Authors	PA, EE	Theory (TTM), MI, Goals	Min/week, PA log, Days/week with Moderate PA
(Wise, 2009)	PT delivered PA and BCT intervention
Chronic SCI
Multiple Sclerosis and Parkinson’s Disease
Carter, 2014	PT Author	PA, EE	Theory (TTM), Self-efficacy, Goals, Accountability	Steps/day*, GLTEQ*
Chronic MS	PT supervised exercise physiologist
Coote, 2017	PT Authors	PA, EE	Theory (SCT), Coaching, Self-efficacy, Goals, Barrier & Facilitators, Feedback, Accountability	Steps/day, GLTEQ, Energy, Exercise logs
(Hayes, 2017)	PT delivered PA and BCT intervention
Chronic MS
Hugos, 2019	PT Authors	EE	Goals, Barrier & Facilitators	RAPA
Chronic MS	Groups delivered by “MS professionals” as facilitators
Learmonth, 2017	PT Author	PA, EE	Theory (SCT), Coaching, Self-efficacy, Goals, Barrier & Facilitators	MVPA, GLTEQ*, Exercise log, goals attained*
	Delivered by “behavioral coaching”
Chronic MS
Martini, 2018	PT Authors	PA, EE		IPAQ
Chronic MS	PT delivered PA intervention
McGibbon, 2018	PT Authors	PA		Steps/day, Daily activity logs
	“Therapists” fit and trained participants with PA device
Chronic MS
Plow, 2009	No PT Author	PA, EE	Goals	HPLP-II-subscale physical activity
(Plow, 2008)	PTs delivered PA and BCT intervention
Chronic MS
Rice, 2015	PT Author	PA, EE	Theory (SCT), Self-efficacy, Goals, Planning, Barrier & Facilitators	Wrist activity counts
Chronic MS	PT delivered PA (wheelchair training) and BCT
Ryan, 2019	PT Authors	PA, EE	Theory (Michie’s Taxonomy), Goals, Planning, Barrier & Facilitators, Feedback	Step time*, Step counts, MVPA, sedentary, PA Diary
Chronic MS	PT (trained) led BCT intervention
Thomas, 2017	PT Author	PA, EE	Theory (SCT, SDT), MI, Self-efficacy, Goals, Planning, Barrier & Facilitators	Accelerometer, GLETQ*
Chronic MS	PT delivered BCT intervention
Collett, 2017	PT Author	PA, EE	Theory (SDT), Barrier & Facilitators	MVPA, PASE, PA Diary
(Coe, 2018)	Delivered by “Professional experienced in clinical exercise”
Chronic PD
Ellis, 2019	PT Authors	PA, EE	Goals, Planning, Feedback	Steps/day, MVPA, PA log
Chronic PD	PT delivered PA and BCT intervention
van Nimwegen, 2013	PT Author	PA, EE	Theory (TTM, SCT), Coaching, Goals, Barrier & Facilitators, Feedback, Accountability	LAPAQ, Energy*, PA Diary *
	PT delivered PA and BCT monthly coaching
(van der Kolk, 2014)
Chronic PD
White, 2009	PT AuthorPT delivered PA in interdisciplinary team + students	PA, EE		Time spent walking, MVPA
Chronic PD
Other Neurologic Conditions
Busse, 2017	PT Author	PA, EE	Theory (SDT), Coaching, Goals, Planning, Barrier & Facilitators	IPAQ
Chronic HD	PT delivered BCT
Suttanon, 2013	PT Author	PA, EE		HAP-AAS
Chronic AD	PT delivered PA intervention and supervised program
Goldberg, 2019	No PT Author	PA, EE	Theory (SDT), Goals, Barrier & Facilitators, Support/ Account	Total # of steps, IPAQ*
Chronic Mild Dementia	PT delivered PA + BCT with interdisciplinary team
Brenner, 2012	PT Author	EE	Theory (TTM, SCT), Self-efficacy, Goals, Barrier & Facilitators	HPLP-II- subscale of physical activity
Chronic TBI	PT part of interdisciplinary trained BCT facilitators team
Elsworth, 2011	PT Author	PA, EE		Activity monitor, PASE
Chronic Mixed Neuro	PT at initial PA intervention start, and as support
Hassett, 2020	PT Author	PA, EE	Coaching, Goals, Feedback	Steps/day, Time spent stepping, upright, IPEQ*
Mixed Neuro	PT delivered PA intervention
Haworth, 2009	PT Author	PA, EE	Self-efficacy	HAP-AAS, HAP-Maximum Activity Score
Chronic Mixed Neuro	PTs recruitment, assessment, and PA intervention, not BCT

General abbreviations: PA, physical activity; PT, physical therapist; BCT, behavior change techniques; MET, metabolic equivalent of task; MI, motivational interviewing.

Diagnoses: HD, Huntington’s disease; MS, multiple sclerosis; PD, Parkinson’s disease; SCI, spinal cord injury; TBI, traumatic brain injury.

Physical activity outcome measures: Baecke, Baecke Questionnaire of Habitual Physical Activity; FAI, Frenchay Activities Index; GLTEQ, Godin Leisure-Time Exercise Questionnaire; HAP-AAS, Human Activity Profile-Adjusted Activity Score; HPLP-II, Health-Promoting Lifestyle Profile-II; IPAQ, International Physical Activity Questionnaire; IPEQ, Incidental and Planned Exercise Questionnaire; LAPAQ, Longitudinal Ageing Study; Amsterdam Physical Activity Questionnaire; LTPAQ, Leisure Time Physical Activity Questionnaire; PADS, Physical Activity and Disability Scale; PAS2, Physical Activity Scale 2.1; PASE, Physical Activity Scale for the Elderly Questionnaire; PASIPD, Physical Activity Scale for Individuals with Physical Disabilities; RAPA, Rapid Assessment of Physical Activity

Theories: SCT, Social Cognitive Theory; SDT, Self-Determination Theory; Michie’s Taxonomy, Michie’s Taxonomy of Behavior Change; TTM, Transtheoretical Mo HAPA, Health Action Process Approach; TPB, Theory of Planned Behavior; RP, Relapse Prevention Model.

*Note: We summarized two articles from the same trial in a single row when both articles met the inclusion criteria with novel PA outcome measures (e.g., Collett et al. and Coe et al.).

Of the 54 unique trials, PTs were authors in 48 (89%), although PT roles beyond authorship were unclear in 11 (20%). PTs delivered at least part of the PA intervention in 43 trials (80%). An interdisciplinary team-based approach (PTs and other professionals) was used in 14 trials (26%).^{81,82,92,96,99,102,110,114,115,117-119,122,124} In 8 trials (15%), PTs supervised or trained others to deliver the intervention.^{81,90,96,99,114,115,118,119}

PA interventions were delivered in varied settings, structures, and doses. They occurred in homes, communities, clinics, and research settings; 18 (33%) were delivered in part or entirely via phone/video conferencing ^{74,76,94,97,100,102,105,107,109, 115,116,120,122,123,125,127,131,133}; and 17 (32%) used group formats (in-person or video).^{78,81,82,90,94,99,100,105,110,114,117,119,124,127-130} PA intervention dosage varied in frequency (1 session/4 months to 5 sessions/week), intensity (15-120 minutes), and duration (1 week to 24 months), with variable follow-up periods (2 weeks to 24 months).

Intervention Types and Behavioral Change Technique Use

Table 3 shows that 29 of 54 trials (54%) combined PA and BCT interventions; 18 (33%) reported the effectiveness of PA interventions only; and 7 (13%) used BCT interventions only.

The interventions are categorized based on the use of PA interventions or explicit behavior change components. Among the 36 trials using BCTs, 18 (50%) used explicit named behavior change theories (Table 3 footnotes). Most BCTs drew on Social Cognitive Theory—self-efficacy, barriers/facilitators, outcomes expectations, and PA cost/benefits. Other BCTs included components of the Transtheoretical Model with the stages of change, decisional balance, processes of change, or self-efficacy; health coaching; Motivational Interviewing; Self-Determination Theory; or Health Action Process Approach.^96,126 Goal setting was used in 29 BCT interventions (81%). Education or encouragement alone was not included as a BCT without additional BCT strategies in 9 trials.

Overall, the included articles reported 25 distinct PA outcome measures with varied methods of operationalization. Direct measures of PA included accelerometers or activity monitors (eg, steps per day). Sixteen self-report measures of PA were used across trials. Exercise diaries or logs were reported in 12 trials.^{63,74,75,93,97,100,102,104,108,110,112,125} The Physical Activity Scale for Individuals with Physical Disabilities (PASIPD) was the most widely used scale in 6 trials.^88,92-95,126

Risk of Bias and Quality of Reporting

The Cochrane RoB and PEDro quality assessments provide information on potential sources of bias (additional detail in Supplemental Digital Content Table S8, available at: http://links.lww.com/JNPT/A567). Most of the 60 articles had high RoB; only 11 (18%) had low-moderate RoB.^{63,82,103,118,126,128} The PEDro scores indicated that, on average, the quality of reporting across the 60 articles was moderately good,^60,61,135 with a mean PEDro score of 6.3 ± 1.4 (range 3-8). None were rated excellent or high quality (≥9)^60-62 and 9 (15%) were low quality (≤4).

Across articles, high Cochrane RoB concerns clustered in the randomization process,^{74,80,83,89,94,96-98,101,104-107,111,114,115,117,120,122} deviation from intended intervention,^{75,79,83-89,91-93,95-98,100,102,105-107,114,116,117,122,124,125,129} or missing outcome data.^{75-77,83,95,98,100-102,105-107,110,111,114-116,120-122,127,129,130,132} Reporting quality concerns from PEDro included lack of blinded interventionists and participants, unclear allocation concealment, between-group differences in PA outcome measure at baseline, and high attrition in PA outcome measures (additional detail in Supplemental Digital Content Table S8, available at: http://links.lww.com/JNPT/A567). Additionally, 12 trials had ≤40 participants. Only 8 of 54 trials reported harm data.^{74-76,78,80,110,115,126} Adverse events, including falls, changes in health status, and deaths, occurred in both intervention and control groups, with no direct indication that these events were caused by the intervention(s).

Efficacy Synthesis

Without considering covariates, the meta-analyses indicated that PA interventions involving PTs yielded small but significant effects on direct measures of PA (ES = 0.18, CI_95%, 0.01-0.35, 95% PI: −0.70 to 1.06, τ² = 0.16, SE = 0.07, I² = 74.3%, P = .04) but not on self-report measures of PA (ES = 0.10, CI_95%, −0.05 to 0.26, 95% PI: −0.56 to 0.77, τ² = 0.09, SE = 0.06, I² = 73.9%, P = .19; Figure 2). Visual inspection of funnel plots for the primary meta-analyses indicated approximate symmetry, suggesting no clear evidence of small-study effects or publication bias (Supplemental Digital Content Figure S1, available at: http://links.lww.com/JNPT/A559).

Figure 2.

Forest plots of meta-analysis results for direct measures and self-report measures of PA.

Sensitivity meta-analyses revealed similar results for direct measures of PA and subjective measures of PA. Exclusion of articles with high RoB or missing data also yielded a small-moderate ES (ES = 0.36, CI_95%, 0.05-0.67, 95% PI: −0.67 to 1.39, τ² = 0.10, SE = 0.10, I² = 68.06%, P = .02) for direct measures of PA (Figure 3), but not for self-report measures (ES = −0.09, CI_95%, −0.39 to 0.22, 95% PI: −1.04 to 0.87, τ² = 0.07, SE = 0.09, I² = 87.9%, P = .58). The only other sensitivity analyses that were significant combined articles using a direct measure of PA with only true inactive controls. It demonstrated a small to moderate effect (ES = 0.39, CI_95%, 0.06-0.72, 95% PI: −0.57 to 1.34, τ² = 0.13, SE = 0.12, I² = 61.05%, P = .02). Insignificant sensitivity analyses are displayed in Supplemental Digital Content Figure S2, available at: http://links.lww.com/JNPT/A559.

Figure 3.

Forest plots of Risk of Bias (RoB) sensitivity analysis results for direct measures and self-report measures of PA.

A forest plot illustrating the meta-regression is presented in Figure 4. With the Bonferroni correction, only the intervention type significantly moderated the treatment effect of PA interventions for direct measures of PA (QM_(df=2)= 14.0, β = 0.43, CI_95%, 0.23-0.62, 95% PI: −0.19 to 1.04, P = .001; τ² = 0.09, SE = 0.04; I² = 61%). PA interventions that solely employed PA interventions or solely employed BCT interventions had significantly decreased treatment effects on direct measures of PA compared to a combined approach (Table 4).

Figure 4.

Forest plots of univariable meta-regression results for direct measures and self-report measures of PA.

Table 4.

Results of Univariable Covariate Analyses Using Meta-Regression Random Effects Model for Direct Measures of PA

	Overall Test of Moderators [QM(df)]	Overall P-value	τ2 (SE)	I2	β, 95% CI	P-value
Direct measures of PA overall model (effect size = 0.18, CI95%, 0.01, 0.35, PI95% [−0.70, 1.06], τ2 = 0.16, SE = 0.07, I2 = 74.3%, P = .04)
Intervention duration weeks (<12 wk—REF)	QM(2) = 4.31	.12	0.14 (0.06)	71%	0.30 (−0.01, 0.61)
Intervention duration—12 wk					0.07 (−0.37, 0.52)	.75
Intervention duration—<12 wk					−0.31 (−0.71, 0.08)	.12
Diagnosis (neurodegenerative diseases—REF)	QM(2) = 2.65	.27	0.17 (0.07)	75%	0.07 (−0.32, 0.45)
Diagnosis—spinal cord injury, mixed neuro					0.45 (−0.14, 1.04)	.13
Diagnosis—stroke					0.07 (−0.37, 0.51)	.75
Acuity (acute–subacute REF)	QM(1) = 0.21	.64	0.16 (0.07)	74%	0.14 (−0.11, 0.38)
Acuity—chronic					0.08 (−0.26, 0.42)	.64
Impairment (mild—REF)	QM(2) = 1.44	.49	0.16 (0.07)	74%	0.32 (−0.01, 0.65)
Impairment—mild–moderate					−0.24 (−0.65, 0.17)	.25
Impairment—moderate–severe					−0.09 (−0.59, 0.41)	.72
Setting (clinical—REF)	QM(1) = 2.55	.11	0.13 (0.06)	71%	0.06 (−0.15, 0.28)
Setting—community					0.27 (−0.06, 0.60)	.11
Control group type (inactive control—best medical care—REF)	QM(3) = 3.02	.39	0.15 (0.07)	72%	0.39 (0.05, 0.73)
Inactive control—usual care PT or standard PT					−0.39 (−0.84, 0.07)	.10
Active control—education					−0.34 (−1.00, 0.31)	.30
Active control—active PA or exercise					−0.17 (−0.61, 0.27)	.44
PA as outcome (primary—REF)	QM(1) = 1.24	.27	0.14 (0.06)	71%	0.24 (0.04, 0.44)
PA as outcome—secondary					−0.20 (−0.55, 0.15)	.27
Intervention type (combined PA and BCT—REF)	QM(2) = 14.0	.001 a	0.09 (0.04)	61%	0.43 (0.23, 0.62)
Intervention type—BCT only					−0.51 (−1.06, 0.04)	.07
Intervention type—PA only					−0.55 (−0.85, −0.25)	<.001 a
Outcome measure (% time stepping per day—REF)	QM(5) = 6.14	.29	0.16 (0.07)	76%	−0.38 (−1.54, 0.79)
Outcome measure—# steps per day					0.55 (−0.62, 1.73)	.36
Outcome measure—activity counts per day					0.54 (−1.31, 2.38)	.57
Outcome measure—MVPA min per day/per week					0.28 (−0.98, 1.54)	.66
Outcome measure—WC min per day					1.19 (−0.23, 2.60)	.10
Outcome measure—vector magnitude counts					1.28 (−0.32, 2.88)	.12

General abbreviations: CI, confidence interval; PI, prediction interval; QM, test of moderators; REF, reference group; PA, physical activity; BCT, behavior change techniques.

Direct measures of physical activity abbreviations: % Time stepping per day (Time_Step), mean percent of time stepping over 5 d; # Steps per day (Steps_Per_Day), total number of steps per day via accelerometer; Activity Counts Day (Activity_Counts_Day), wrist worn accelerometer; MVPA, moderate to vigorous physical activity in time/duration normally reported as (min/d or min/wk) measured using accelerometers; WC Min per day (WC_Min_Day), total wheeled wheelchair minutes per day; VM_Counts, vector magnitude counts.

Neurodegenerative diseases include PD and MS.

^aindicates statistical significance after Bonferroni correction (uncorrected P < 0.0028).

None of the covariates significantly moderate the treatment effects measured with self-reported PA measures after Bonferroni correction (Table 5).

Table 5.

Results of Univariable Covariate Analyses Using Meta-Regression Random Effects Model for Self-Report Measures of PA

	Overall Test of Moderators [QM(df)]	Overall P-value	τ2 (SE)	I2	β, 95% CI	Uncorrected P-value
Self-report measures of PA overall model (effect size = 0.10, CI95%, −0.05, 0.26, PI95% [−0.56, 0.77], τ2 = 0.09, SE = 0.06, I2 = 73.9%, P = .19)
Intervention duration weeks (<12 wk—REF)	QM(2) = 8.97	.01	0.07 (0.05)	70%	0.37 (0.05, 0.69)
Intervention duration—12 wk					−0.24 (−0.61, 0.12)	.19
Intervention duration—>12 wk					−0.74 (−1.24, −0.25)	.003
Diagnosis (neurodegenerative diseases—REF)	QM(2) = 1.92	.38	0.08 (0.05)	65%	0.23 (−0.03, 0.50)
Diagnosis—spinal cord injury, mixed neuro					−0.08 (−0.51, 0.35)	.71
Diagnosis—stroke					−0.23 (−0.57, 0.11)	.18
Acuity (acute–subacute—REF)	QM(1) = 5.05	.02	0.05 (0.04)	58%	−0.06 (−0.25, 0.13)
Acuity—chronic					0.31 (0.04, 0.58)	.02
Impairment (mild—REF)	QM(2) = 1.76	.41	0.10 (0.07)	75%	−0.14 (−0.57, 0.30)
Impairment—mild–moderate					0.31 (−0.17, 0.78)	.21
Impairment—moderate–severe					0.17 (−0.43, 0.76)	.58
Setting (clinical—REF)	QM(1) = 0.39	.53	0.10 (0.06)	71%	0.05 (−0.19, 0.29)
Setting—community					0.10 (−0.22, 0.42)	.53
Control group type (inactive control—best medical care—REF)	QM(3) = 3.88	.27	0.07 (0.05)	63%	0.33 (−0.01, 0.67)
Inactive control—usual care PT or standard PT					−0.41 (−0.83, 0.01)	.06
Active control—education					−0.24 (−0.76, 0.28)	.27
Active control—active PA or exercise					−0.17 (−0.59, 0.24)	.42
PA as outcome (primary—REF)	QM(1) = 0.57	.45	0.09 (0.06)	71%	0.03 (−0.22, 0.28)
PA as outcome—secondary					0.12 (−0.20, 0.45)	.45
Intervention type (combined PA and BCT—REF)	QM(2) = 0.96	.62	0.09 (0.05)	63%	0.17 (−0.04, 0.38)
Intervention type—BCT only					−0.23 (−0.76, 0.29)	.38
Intervention type—PA only					−0.11 (−0.44, 0.22)	.52
Outcome measure (PAS_2 Met hours/day—REF)	QM(10) = 9.33	.50	0.10 (0.07)	60%	−0.49 (−1.30, 0.31)
Outcome measure—MET min per week					0.25 (−0.77, 1.28)	.63
Outcome measure—PASE					0.46 (−0.43, 1.35)	.31
Outcome measure—PASIPD					0.55 (−0.33, 1.44)	.22
Outcome measure—HAP-AAS					0.58 (−0.32, 1.48)	.21
Outcome measure—PADS					0.64 (−0.50, 1.78)	.27
Outcome measure—LAPAQ					0.69 (−0.34, 1.72)	.19
Outcome measure—moderate intensity-IPAQ					0.70 (−0.37, 1.78)	.20
Outcome measure—IPAQ					0.83 (−0.15, 1.81)	.10
Outcome measure—GLTEQ					0.92 (−0.03, 1.88)	.06
Outcome measure—LTPAQ					1.52 (0.23, 2.80)	.02

Self-report measures of PA abbreviations: PAS_2 Met hours/day, Physical Activity Scale version 2.1 (MET h/d); MET min week, Metabolic equivalent minutes per week self-report; PASE, Physical Activity Scale for the Elderly Questionnaire; HAP-AAS, Human Activity Profile—Adjusted Activity Score; PADS, Physical Activity and Disability Scale; PASIPD, Physical Activity Scale for Individuals with Physical Disability; LAPAQ, Longitudinal Ageing Study Amsterdam Physical Activity Questionnaire; LTPAQ, Leisure Time Physical Activity Questionnaire; GLTEQ, Godin Leisure-Time Exercise Questionnaire; IPAQ, International Physical Activity Questionnaire; Moderate Intensity-IPAQ, moderate physical activity min/wk from IPAQ.

Neurodegenerative diseases include PD, MS, Dementia and Huntington’s disease. No results are statistically significant after Bonferroni correction (uncorrected P < 0.0028).

DISCUSSION

This review describes the roles of PTs in delivering evidence-based PA interventions for adults with neurologic conditions and explores how they can inform neurologic physical therapy practice. We found that PTs deliver PA interventions for individuals with neurologic conditions alone or as a part of an interdisciplinary team.^112,115 This finding is consistent with a Cochrane review that found that multiple disciplines, including health professionals, exercise specialists, and PA researchers, delivered PA interventions to community-dwelling older adults, with no differences in the effectiveness of the interventions across disciplines.¹³⁶ The PA interventions demonstrated efficacy when measured with direct accelerometry-based measures of PA, but not self-reported measures of PA. This finding was significant in the full meta-analysis and was stronger when articles with only a low or moderate RoB were included. Exploratory meta-regression indicated that interventions including both PA and BCT components were associated with greater increases in direct accelerometer-measured PA compared to either alone. However, readers should interpret these results cautiously due to heterogeneity of interventions, outcomes, patient populations, and possible selective reporting and attrition in measures of PA.

Our exploratory meta-regressions suggest that combining PA interventions with BCTs was associated with positive changes in PA outcomes.^{74,77,93,95,100,108,109,112} However, the predictive interval crosses the null, suggesting that future trials may still show positive or negative effects.¹³⁷ Our positive effects were congruent with the Morris et al finding that PA interventions, including counseling alone or combined with exercise programs, are more effective at increasing PA than exercise alone following stroke.³⁴ If a PT does not have experience providing BCTs, they could consider seeking training or working with other qualified providers.^138-142 A prior review by Alexander and colleagues (2012) described BCTs used by PTs to promote behavior change in people with non-neurological lifestyle-related conditions such as hypertension, diabetes, and obesity.¹⁴³ While they did not complete a meta-analysis to determine a most effective BCT, the review detailed BCT applications that could guide neurologic PTs.¹⁴³ They found that PTs most commonly used the Transtheoretical Model.¹⁴³ In our review, the Transtheoretical Model is explicitly mentioned in only 4 theory-based BCT interventions, but some aspects of the model (stages of change, readiness to change, and self-efficacy) are mentioned in 12 trials. Aspects of the Social Cognitive Theory⁴⁷ (self-efficacy, outcome expectations, barriers, and perceived benefits of PA) are discussed more frequently (31 of the 36 BCT trials). Self-efficacy is a cross-over concept that is important to both models and is mentioned in 12 of the trials with BCT-based PA interventions. Alexander et al also found that PTs often use the 5A’s model of addressing the agenda, assessing, advising, assisting, and arranging follow-up.^143-145 In contrast, none of the articles we reviewed state use of the 5A model explicitly, but all include 1 or more aspects of the 5As.^37,136,143

Gaps in Current Research

Although clinical practice guidelines in neurologic physical therapy endorse implementation of evidence-based PA interventions, including BCTs, they lack specificity for real-world clinical application.^146-150 This review highlights gaps in the literature, including how to deliver effective PA interventions that incorporate BCT in real-world neurologic physical therapy practice, inconsistent PA measurement, lack of emphasis on primary PA interventions, unknown optimal duration of intervention for sustained effects, and a need for greater methodologic rigor with reduced RoB.

Our meta-regression of articles with a low-moderate RoB suggested that BCT interventions should be included with traditional PT. However, it is unclear what training is needed to implement these interventions. While entry-level physical therapy education typically includes psychosocial health and BCT,¹⁵¹ PTs encounter barriers to delivering PA and BCT interventions.^16,152 PTs generally use simpler BCTs such as education, social support, individualized exercise programming, and participative goal setting.¹⁴³ However, advanced BCTs, such as motivational interviewing or health coaching, require additional certifications that could limit implementation. Many trials lacked details on the level of training provided to implement BCTs, making it difficult to replicate the interventions. Routine use of reporting checklists, like the template for intervention description and replication (TIDieR),¹⁵³ consensus for exercise reporting template (CERT),¹⁵⁴ and the workgroup for intervention development and evaluation research (WIDER)¹⁵⁵ would enhance replication. Future research should provide and compare intervention components using these checklists.

Another gap in the literature surrounds the identification of appropriate PA outcome measures in PT practice. Twenty-five different outcome measures were reported in this review, and measurement confusion was common. We excluded some articles because they did not include a PA measure despite an explicit goal to increase PA, or they included a PA measure without an intervention targeting increasing PA.^156-162 The most favorable outcomes were direct measures of PA (eg, steps per day and MVPA time), though these can be difficult for clinicians to collect¹⁶³ as they can require complex digital health technology with proprietary algorithms or potentially inaccurate commercial devices for slower walking individuals (<0.6 m/s).¹⁶⁴ Furthermore, the accuracy or validity of direct sensor-derived measures of PA across diagnoses could limit the comparability of direct measures of PA data.^165-167 The more clinically feasible self-reported PA outcome measures were less responsive to the PT-led interventions in this review and have only moderate reliability and limited validity compared with instrumented measures.^168,169 Consensus has been reached for PA measurement recommendations after stroke (eg, accelerometers for stepping frequency), but there is no consensus for other neurological conditions.¹⁷⁰ Implementation of PA interventions in neurologic physical therapy practice will be enhanced by the adoption of reliable, valid, and clinically feasible PA outcome measures.

A third gap in the literature surrounds the testing of primary versus secondary PA interventions. We included trials with PA as either a primary or secondary focus. The inclusion of secondary measures and interventions introduces a potential source of reporting bias where positive secondary PA results could be reported more often than negative secondary PA results. Inclusion of these articles may have biased the results away from potentially more effective primary PA-focused interventions. Lack of trials that focus on PA goals and measures may stem from traditional PT roles focused on improving physical function rather than on PA promotion. Additionally, our narrow focus on PA interventions delivered or designed by PTs can inform quality improvement research in health care settings, but does not enable us to draw conclusions about the possibly more effective PA interventions that could be implemented in primary health care or the community.^39-41 For example, we excluded trials with compelling PA interventions delivered outside of physical therapy.¹⁷¹ Future research should determine whether PA interventions in neurologic PT can be improved based on the literature in the community PA promotion field.

A fourth gap in the literature is on the duration and sustained effects of interventions. The types and duration of interventions applied by neurologic PTs began in acute care and continued into chronic phases of recovery. People with neurologic disorders have a lifelong need for PA.^12-17 Research in other populations suggests that PA interventions can improve PA for up to 4 years after intervention.³⁹ Intervention durations varied and frequently included different phases with different doses (eg, more intense bursts with less frequent follow-up contacts). Post-intervention follow-up periods in this review were highly variable, so we could not synthesize the results related to sustained effects. Additional investigations are needed to study the long-term effects of PT-led PA interventions.

The high RoB and low quality of reporting of many of the PT-led trials is a final concerning gap in the current research. It is promising that the pooled ES was larger (0.36 compared to 0.18) and somewhat less heterogeneous (I² 74.3%-68.1%) when we analyzed just the articles with low-moderate RoB. However, neurologic PT researchers need to improve methodologic rigor. The overall moderate PEDro scores of the included articles may mask serious flaws. Sources of bias include unclear allocation concealment and high attrition, which restrict our confidence in the results of individual articles. Additionally, we excluded grey literature and languages other than English and Spanish, so we may not fully capture the range of research that has been performed.⁶⁹

Limitations and Future Directions

The intentionally broad inclusion criteria led to limitations of this review. High RoB, as well as heterogeneity among PA interventions and outcome measures, particularly in the overall analyses, limits efficacy synthesis and confounds our understanding of the role of PTs in enhancing PA. There was substantial heterogeneity across populations, intervention types, settings, duration, intervention intensity, follow-up measurement periods, sample sizes, study phases, and how PTs were involved. To mitigate this limitation, we conducted a sensitivity analysis using articles with low-moderate RoB, as well as meta-regression analyses to explore potential sources of variability. Additionally, the sensitivity analysis for direct measures of PA was based on a small number of articles with wide confidence intervals, which increases uncertainty and reduces the precision of these estimates. Future RCTs should have greater methodologic rigor, including reporting the interventions using the TIDieR and WIDER checklists.^153,155

The purpose of this systematic review was broadly focused to inform the future of neurologic PT research and practice. However, we focus on recommendations for research rather than practice because we did not apply the GRADE framework.¹⁷² The heterogeneity and RoB in most of the articles increase the risk of type I error. Our article extraction table (Table 3), combined with our use of both the Cochrane RoB tool and the PEDro scale (additional detail in Supplemental Digital Content Table S8, available at: http://links.lww.com/JNPT/A567), provides substantial information about each article for researchers or clinicians to see what research is available in their field. Future systematic reviews could use this information to focus their research questions further and prospectively apply GRADE to more systematically evaluate certainty across outcomes and make recommendations for clinical practice.¹⁷²

CONCLUSIONS

PTs may play an integral part in PA interventions for patients with adult-acquired neurological conditions. Although methodologic rigor in existing RCTs is low, PA interventions appear to be associated with greater positive effects when they combine PA and BCT approaches. As the physical therapy profession evolves in PA health promotion activities, further research is needed to develop consensus on PA delivery roles, interventions, and outcome measures. Evidence-based interventions, improved methodologic rigor, and common data elements to document change are critical to advance the field of PT-led PA health promotion and wellness with reduced heterogeneity.²⁹

Supplementary Material

jnpt-50-180-s002.docx ^{(2.4MB, docx)}

jnpt-50-180-s003.docx ^{(23.7KB, docx)}

jnpt-50-180-s004.docx ^{(25KB, docx)}

jnpt-50-180-s005.xlsx ^{(53.4KB, xlsx)}