Assessment of Outcomes of Inpatient or Clinic-Based vs Home-Based Rehabilitation After Total Knee Arthroplasty

Key Points Question Is inpatient or clinic-based rehabilitation associated with superior outcomes after total knee arthroplasty compared with home programs? Findings This systematic review and meta-analysis included 5 unique studies involving 752 unique participants comparing clinic- and home-based rehabilitation and 1 study comparing inpatient rehabilitation with a home-based program. Based on low- to moderate-quality evidence, no associations between settings, no clinically important differences for mobility or patient-reported pain and function at 10 and 52 postoperative weeks, and no significant differences in quality of life or range of motion were found. Meaning For adults who underwent total knee arthroplasty, clinic or inpatient vs home-based rehabilitation appeared to offer no clinically important advantages.


Introduction
Total knee arthroplasty (TKA) was the most frequently performed inpatient operating room procedure in the United States in 2012. 1 From 2003 to 2012, the incidence of TKA increased from 145.4 to 223.0 per 100 000 population (a 4.9% mean annual increase), with the total number performed in the United States projected to increase from 711 000 in 2011 to 3.48 million by 2030. 2 Similarly in Australia, the incidence increased from 108.3 per 100 000 population in 2003 to 222.3 per 100 000 population in 2017, with more than 54 000 TKAs performed in 2017. 3e increased volume of surgery constitutes a significant burden on the acute health care budget, but because the surgery is typically followed by a protracted rehabilitation period, the latter can add significantly to the cost of care.Several studies [4][5][6] describe a significant cost differential among rehabilitation pathways involving inpatient rehabilitation after TKA, ranging from a 5-fold to a 26-fold cost differential between a rehabilitation pathway that included inpatient therapy and one that did not despite no differences in outcomes between groups.Concern about the total episodeof-care costs for arthroplasty, including the rehabilitation period, has led to the introduction of bundled payments in the United States, consisting of a single bundled payment to health care organizations for all services related to the TKA to 90 days after surgery. 7,8This payment approach has subsequently driven health care providers to reconsider the use of the more expensive inpatient rehabilitation pathways. 7tside inpatient rehabilitation, the setting, cost, and modes of provision vary greatly when rehabilitation is delivered in the community. 6,90][11][12][13] Previous systematic reviews of randomized clinical trials 14,15 have concluded that no single setting-clinic-or homebased, in water or on land-appears to be associated with better recovery across a range of outcomes.Despite this finding, to date, no evidence-based clinical practice guideline exists to promote the use of home-based programs after uncomplicated TKA.Trials published since the aforementioned reviews, [16][17][18] however, have included new comparisons (inpatient and 3-arm trials) and constitute the largest TKA rehabilitation trials to date.Thus, a more contemporary review is warranted, potentially as a precursor to development of a much-needed clinical practice guideline.
The aim of this systematic review and meta-analysis was to investigate the importance of the rehabilitation setting on outcomes for adults after elective, primary, unilateral TKA.Specifically, we aimed to determine whether inpatient or clinic-based rehabilitation is associated with superior function and pain outcomes after TKA compared with any home-based physiotherapy program (monitored or unmonitored, or domiciliary [physiotherapy home visitation]).Superiority was defined as a change considered to be clinically important for each outcome assessed.

Methods
This systematic review of randomized clinical trials follows the methods described in the Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0 19and is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline. 20The protocol was updated to include the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) component for assessing quality of evidence. 21

Types of Studies
Published randomized clinical trials were eligible for inclusion.We excluded studies reported only as abstracts if adequate data could not be obtained from the authors, studies in which TKA data could not be separated from other procedures (eg, total hip arthroplasty), and studies for which we were unable to obtain potentially relevant data from the authors on request.No language restrictions were applied.

Types of Participants
We included studies of adults (age Ն18 years) who had undergone a primary unilateral TKA and commenced rehabilitation within 3 months of surgery.We excluded studies of unicompartmental surgery, revision TKA, or TKA secondary to trauma.

Types of Interventions
We included studies investigating rehabilitation after TKA in which patients who had received postacute inpatient or clinic-based rehabilitation were compared with others who had received a monitored or an unmonitored home-based or domiciliary program after discharge from the acutecare facility.We excluded telerehabilitation because other reviews in progress during conduct of our study and subsequently published [22][23][24] have investigated this option.

Outcomes
The goal of physiotherapy-based rehabilitation after TKA is to improve physical function, including walking, activities of daily living, and knee mobility.We grouped outcomes into the following categories that broadly reflect these goals: physical performance test results (6-minute walk test [6MWT], 25 measured as laps walked on a flat surface in 6 minutes; walking speed, stair ascent and descent tests, and chair rise test), patient-reported pain and function (Oxford knee score 26 [OKS; range, 0-48, with higher scores indicating best outcomes], Knee Injury and Osteoarthritis Outcome Score 27 [range, 0-100, with higher scores indicating worse outcomes], and Western Ontario and McMaster Universities Osteoarthritis Index 28 [5 items for pain, 2 for stiffness, and 17 for functional limitation, with higher scores indicating worse outcomes]), generic health-related quality-of-life measures (12-and 36-Item Short Form Health Surveys and EuroQol-5D 29 ), and knee range of motion (ROM), expressed as active or passive ROM, extension, and/or flexion.

Primary and Secondary Outcomes
Primary outcomes were mobility (6MWT) and patient-reported pain and function (OKS) measured at 10 to 12 postoperative weeks.Secondary outcomes included mobility and patient-reported pain and function, knee ROM, postoperative complications, and health-related quality of life measured at 10, 26, and/or 52 weeks.

Identification and Selection of Studies
We searched Embase, PubMED, MEDLINE, and CINAHL from inception to June 19, 2018, using search terms that included knee arthroplasty, randomized controlled trial, physiotherapy and related terms, and rehabilitation (eMethods 1 in the Supplement).We later scanned references of all included studies.
Two reviewers (M.A.B. and J.M.N.) independently screened titles and abstracts of the search output to identify studies suitable for further scrutiny.They discussed inconsistencies in the screening process before a decision was made to review the full text.The same reviewers then

JAMA Network Open | Physical Medicine and Rehabilitation
Inpatient or Clinic-Based vs Home-Based Rehabilitation After Total Knee Arthroplasty screened full-text articles to determine inclusion in the review.Discrepancies in the final list were discussed, and a consensus was reached for all articles.

Assessment of Study Quality, Risk of Bias, and GRADE Assessment
Two reviewers (M.A.B. and J.M.N.) independently assessed study quality using the Cochrane Collaboration Risk of Bias tool, 30 which includes the following variables: random allocation sequence, allocation concealment, blinding (of patients, therapists, and outcome assessors), attrition (loss to follow-up and intention-to-treat analysis), and selective outcome reporting.Disagreements in risk of bias were resolved by discussion, or, when necessary, a third person arbitrated.Included studies were also assessed using the Physiotherapy Evidence Database scale, 31 used to identify trials that are more likely to be valid and to contain sufficient information to guide clinical practice.
Two reviewers (S.A. and A.L.) independently used the GRADE component to categorize the quality and strength of the evidence as high, moderate, low, and very low for the 6MWT and patientreported pain and function at 10 to 12 postoperative weeks (the primary outcomes) and at 52 weeks. 32Disagreements were resolved by consensus between the 2 reviewers.To ensure reproducibility and consistency, the reviewers used a checklist to rate each component of the GRADE assessment. 33We used GRADEpro software to create summary of findings tables. 34Because 4 of the investigators (M.A.B., J.M.N., I.A.H., and W.X.) were involved in randomized clinical trials relevant to this review, the GRADE assessment was undertaken by reviewers not involved in any of the included studies.

Data Extraction
Four reviewers (M.A.B., J.M.N., S.A., and A.L.) independently extracted data.We collected data related to participants (diagnosis, age, sex, and body mass index); country; study dates; inclusion and exclusion criteria; setting, timing, duration, and intensity of the intervention and comparison (control) conditions; duration of follow-up; losses to follow-up and reasons; and outcomes.Means (SDs) were extracted for outcomes reported as continuous variables.Proportions were extracted for categorical outcomes.Appropriate conversions were applied when outcomes were reported as medians and interquartile ranges or means and 95% CIs. 19r studies with incomplete data, we attempted to contact the corresponding author.We also asked whether any outcomes not reported in their publications had been collected.When authors of included studies were unable to provide additional data, all available data were included in the review.If data had been provided by authors to other reviewers in published reviews, these were included in the analyses in the case of failure to retrieve data from the primary source and acknowledged appropriately.Authors of included studies were also contacted when there was incomplete reporting of data.Where possible, we used data from intention-to-treat analyses in our calculations to determine between-group differences.
For studies with 3 randomized arms, we adopted a strategy described by Higgins and Green 19 of including each comparison separately but with the shared intervention group divided evenly among the comparisons.For continuous outcomes, only the total number of participants was divided, with the means (SDs) left unchanged.

Measures of Treatment Effect
Data were analyzed from June 1, 2015, through June 4, 2018.We used the mean differences (MDs) and 95% CIs for continuous outcomes with the same units (eg, 6MWT).We presented continuous outcomes with different units as standardized MDs and 95% CIs.Categorical outcomes were expressed as a number with percentage.

JAMA Network Open | Physical Medicine and Rehabilitation
Inpatient or Clinic-Based vs Home-Based Rehabilitation After Total Knee Arthroplasty

Data Synthesis
The 2 main comparisons (clinic-vs home-based and inpatient vs home-based) were considered separately.Where possible, we pooled data using random-effects meta-analysis. 35Because the standardized MD can be artifactually affected by correlation between baseline and follow-up measurements when including the SD of change along with the SD of absolute values, we used the baseline SD for change score values when combining change scores and absolute values.For dichotomous outcomes, we used a pooled odds ratio.We used the I 2 statistic to assess statistical heterogeneity among included studies.We planned to explore publication bias using funnel plots if we had a minimum of 10 included studies, but the number found did not reach this.We used RevMan software (version 5.3) 36 to compile data and perform statistical analyses.

Results of the Search
The search strategy yielded 2286 references.After duplicates were removed using the duplicate removal program within EndNote commercial reference management software (Clarivate Analytics) and titles and abstracts were screened, we retrieved 15 studies for evaluation, of which 9 studies were excluded (eMethods 2 in the Supplement).[39][40]

Included Studies
Five unique studies 17,[37][38][39][40] with a total of 752 unique participants (451 [60%] female; mean [SD] age, 68.3 [8.5] years) included in the meta-analysis compared outpatient rehabilitation (individual and/or group) with home-based rehabilitation (monitored or unmonitored).The sixth study 16 with 165 patients (112 [68%] female; mean [SD] age, 66.9 [8.0] years) compared inpatient rehabilitation with home-based rehabilitation monitored by a health care professional.This study was included in a qualitative synthesis only.In all studies, rehabilitation commenced within 3 months of surgery, and participants were followed up for 26 to 52 weeks.Among studies that reported diagnostic data, the most common diagnosis was osteoarthritis.The mean (SD) age of study patients ranged from 66.2 (8.2) to 70.9 (SD not provided) years among studies reporting age.All studies excluded patients with complications in the acute postoperative period.Four studies reported patient adherence to the program (88% 16 ; 77% 17 ; 96% 37 ; and 61% 38 ) (Table 1).One study 17 included in this meta-analysis had 3 randomized arms.We included each comparison separately (in this case, outpatient group-based vs home-based rehabilitation and outpatient one-to-one therapy vs home-based rehabilitation) but with the shared intervention group (ie, the home-based treatment arm) divided evenly among the comparisons as previously described in the data extraction section. 19

Risk of Bias in Included Studies
As shown in Figure 2, 5 studies 16,17,[38][39][40] used adequate methods for generating the randomization sequence, with the sixth study 37 not providing this information.Four studies 16,17,38,39 described the use of adequate methods to conceal allocation.Blinding of participants and therapists was not possible in any of the studies owing to the nature of the intervention, but all included studies blinded assessors of objective outcomes to group allocation.Two studies 16,17 were free of selective outcome reporting.Risk of bias was present in 2 studies 37,38 owing to uneven losses to follow up.
Physiotherapy Evidence Database scores assessing study quality ranged from 5 to 8 (maximum of 10) (eTable 1 in the Supplement).We did not use funnel plots to explore publication bias owing to the small number of included studies.

JAMA Network Open | Physical Medicine and Rehabilitation
Inpatient or Clinic-Based vs Home-Based Rehabilitation After Total Knee Arthroplasty

Intervention Outcomes and Comparisons
The results for all outcomes and comparisons of clinic-based rehabilitation with monitored or unmonitored home-based programs are summarized in eTable 2 in the Supplement.Because only 1 study assessed inpatient rehabilitation, 16 meta-analysis was not possible, and a brief narrative summary is provided.

Mobility
Two studies 17,37 reported the 6MWT at 1 or more follow-up points (eTable 2 and eFigure 1 in the  low-quality evidence suggests that there may be no clinically important difference between clinicand home-based programs for mobility at 10 and 52 weeks (Table 2).Minimal heterogeneity was found across studies reporting mobility outcomes.

Patient-Reported Pain and Function
Three studies 17,38,39 reported a pain and function outcome at 1 or more follow-up points using the OKS (absolute values or change from baseline).Based on the GRADE component, moderate-quality evidence suggests little or no difference between clinic-and home-based programs for patientreported pain and function in 457 patients at 10 weeks (MD, −0.15; 95% CI, −0.35 to 0.05) and in 388 patients at 52 weeks (MD, 0.10; 95% CI, −0.14 to 0.34) (Table 2 and eFigure 2 in the Supplement).

Secondary Outcomes
Patient-Reported Quality of Life Two studies 17,38 reported quality-of-life outcomes at 1 or more follow-up points using the 12-or 36-Item Short Form Health Survey.No superiority of outcomes was found for patients receiving

Figure 2. Cochrane Risk of Bias Table
Buhagiar et al, 2017 16 Ko et al, 2013 17 Kramer et al, 2003 37 Madsen et al, 2013 38 Mockford et al, 2008 39 Rajan et al, 2004 Study quality was assessed using the Cochrane Collaboration Risk of Bias Tool. 30o studies included in this review considered whether outcomes of post-TKA rehabilitation delivered in the domiciliary setting differed from those in other rehabilitation settings.One study comparing inpatient with domiciliary rehabilitation 44 combined data from recipients of total hip and knee arthroplasty and concluded that the combined cohort had no difference in pain, functional outcomes, or patient satisfaction between the 2 treatment groups and that inpatient rehabilitation was not cost-effective.We were not able to obtain individual joint data from the authors, so were not able to include these data in our meta-analysis.
We were also unable to include the largest randomized clinical trial conducted to date concerning rehabilitation after TKA (n = 390). 18This study compared usual care with a home-based exercise program; however, usual care consisted of any combination of clinic-or inpatient-based programs, and many in the home-based program also accessed clinic-based care.Thus, we were unable to assign their participants to exclusively home-based or facility-based care.

Comparison With Other Reviews
A systematic review and meta-analysis 14 published in 2015 examined the effectiveness of physiotherapy exercise after TKA and found no differences for outpatient compared with homebased physiotherapy exercise for physical function or pain outcomes.A short-term benefit that favored home-based physiotherapy exercise for ROM flexion was not clinically important.These findings are consistent with those of our review.

Quality of the Evidence
The risk of bias in the 5 studies included in the review was variable.The primary source of potential bias was from uneven losses to follow-up in 2 studies. 37,38Another potential source of bias, because of the nature of the intervention, was that participants could not be blinded to their treatment.

Future Considerations
We identified a number of ongoing randomized trials comparing rehabilitation settings after TKA for future consideration.One, identified via a search in ClinicalTrials.gov, 45plans to evaluate unsupervised home exercise with and without a web-based recovery platform compared with traditional outpatient physiotherapy after TKA.Another, with a published protocol 46 and feasibility study, 47 will compare clinic-based group physiotherapy with usual (home-based) care.However, the method proposed suggests possible crossover between settings in the latter arm because some patients were referred to physiotherapy services on an individual basis at the discretion of the hospital physiotherapist, orthopedic team, or general physician, with no indication of how many such referrals were made.We also updated our search to November 2018, with no new eligible trials identified.

Strengths and Limitations
This review has specific strengths.We included only studies in which treatment assignments were randomized, enhancing the strength of the conclusions that could be drawn from the findings.Our review was also comprehensive because we included non-English-language articles in the search strategy, although none were found to be suitable for inclusion.
This review also has several limitations.First, the failure to identify all relevant studies is a common source of bias in systematic reviews.We conducted thorough searches of research databases as well as clinical trial registries, including studies in all languages, using reference list searches of included studies and forward citation tracking, and corresponded with authors of included studies.We identified 1 study that was reported only in a conference proceeding 48 Supplement).Participants who received clinic-based rehabilitation (n = 231) had walked an MD in 6MWT of −11.89 m (95% CI, −35.94 to 12.16 m; P = .33)compared with those who received a homebased program (n = 142) at 10 to 12 weeks and an MD in 6MWT of −3.05 m (95% CI, −29.75 to 23.66 m; P = .82)compared with those receiving a home-based program in both studies with 243 participants at 26 weeks.At 52 weeks, participants who had undergone clinic-based rehabilitation had walked an MD in 6MWT of −25.37 m (95% CI, −47.41 to −3.32 m; P = .02)compared with participants who had undergone clinic-based rehabilitation.Based on GRADE assessment,

Figure 1 .Abstracts screened 15 Full-text articles assessed for eligibility 6 included in review 404 Duplicates removed 1867 Abstracts excluded 9 Full-text articles excluded a 2
Figure 1.Flow of Studies Through the Review 2286 Records identified through database search and other sources
Those authors concluded that a home-based exercise program was not inferior to usual care for a range of patientreported and objectively measured outcomes, including the Western Ontario and McMaster Universities Osteoarthritis Index, walking speed, and knee ROM.
and compared clinic-based rehabilitation with an unmonitored home-based program.However, that 44.Mahomed NN, Davis AM, Hawker G, et al.Inpatient compared with home-based rehabilitation following primary unilateral total hip or knee replacement: a randomized controlled trial.J Bone Joint Surg Am. 2008;90(8): 1673-1680.45.ClinicalTrials.gov.Unsupervised home exercise with and without a web-based recovery platform as compared to traditional outpatient physiotherapy after total knee arthroplasty: a prospective, randomized controlled trial.NCT02911389.https://clinicaltrials.gov/ct2/show/NCT02911389. Accessed February 27, 2019.46.Wylde V, Artz N, Marques E, et al.Effectiveness and cost-effectiveness of outpatient physiotherapy after knee replacement for osteoarthritis: study protocol for a randomised controlled trial.Trials.2016;17(1):289.47.Artz N, Dixon S, Wylde V, et al.Comparison of group-based outpatient physiotherapy with usual care after total knee replacement: a feasibility study for a randomized controlled trial.Clin Rehabil.2017;31(4):487-499.48.Hensman-Crook A. The effectiveness of physiotherapy intervention with home exercise programme versus patient directed home exercise programme following total knee replacement.Intern Med J. 2011;41(suppl 1):39.Search Strings for Systematic Review and Meta-analysis eMethods 2. Studies Excluded After Full-Text Review eTable 1. PEDro Scores of Included Studies eTable 2. Meta-analysis Data eFigure 1. Mobility eFigure 2. Patient-Reported Pain and Function eFigure 3. Patient-Reported Quality of Life eFigure 4. Active Range of Motion eFigure 5. Passive Range of Motion Change From Baseline