Background

According to Global Burden of Disease studies, musculoskeletal disorders have been the leading cause of disability for 30 years [1,2,3,4,5], and, due to population aging, are expected to rise significantly over the coming decades [6]. Patients with musculoskeletal disorders comprise up to 30% of consultations in primary care [7,8,9,10,11], making this the second-highest reason for consulting a general medical practitioner (GP) [10]. In most countries, it is standard practice for GPs to refer to a hospital-based orthopaedic surgeon (OS) for advice, investigations, and/or interventions, [12,13,14] although it is believed that fewer than 40% of patients [12, 15,16,17] require an orthopaedic intervention. Thus the urgent challenge facing current health services internationally is to improve access to timely, high quality orthopaedic consultations for the people who need them most, whilst offering equally effective and cost effective alternative care pathways for people who may not require orthopaedic surgeon intervention [18].

Consequently, during the latest two decades, alternative models of care, such as physiotherapist-led (PT-led) orthopaedic triage, have been explored, predominantly in Australia [19], Canada [20, 21], Ireland [22], UK [23] and Sweden [24]. This model of care involves a physiotherapist (PT) assessing, diagnosing and managing patients referred for orthopaedic consultation; a procedure normally done by an OS [25, 26]. This model could potentially be cost effective. Morris et al. [27] reported that appointments with a PT are significantly less expensive than appointments with an OS, and data from four countries suggest that OSs cost the health system approximately twice as much as advanced practice PTs [28,29,30,31,32,33,34,35,36].

When implementing alternative models of care, evaluating effects on patients’ health outcomes is essential. Musculoskeletal disorders burden the individual, the family and society through pain, and disability [37], often limiting participation in daily life activities [37,38,39] and negatively influencing health-related quality of life [40,41,42]. Additionally, patients’ perceptions should be considered, particularly when interventions differ from traditional scope of practice [43, 44]. Patients’ reports of their experiences are increasingly recognised as a valid measure of quality in health care, as well as clinical effectiveness and safety [45, 46].

Several systematic reviews have been published comparing outcomes of advanced or extended scope of PT-led orthopaedic triage with standard care [47,48,49,50], however, these reviews differ in scope, comparators, and outcomes, include studies on specific populations, and were published in 2014 and 2015. Furthermore, the reviews concluded that the generally poor methodological quality of included studies, and low certainty of evidence, did not provide evidence for benefits of advanced or extended scope PT-led orthopaedic triage.

The fields of advanced and extended scopes of are continually developing, underpinned by an expanding body of evidence. An up-to-date review was therefore required to collate the current body of evidence for PT-led orthopaedic triage for musculoskeletal conditions on health and process outcomes. Objectives for this study were to establish the current evidence body on the impact of PT-led orthopaedic triage on health, quality, and service outcomes for patients referred for orthopaedic consultation, compared with standard (orthopaedic surgeon) care.

Methods

Protocol registration

The review protocol was registered on PROSPERO on 30-06-2017 (CRD42017070950). The protocol was modified to include prospective observational studies during the review process, due to the paucity of randomised controlled trials (RCTs) identified in the searches.

Quality framework and reporting standard

The review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [51], and was undertaken in accordance with Cochrane guidelines [52].

Search strategy

The electronic library databases Medline, EMBASE, Scopus and CINAHL were searched from inception to 7 May 2018. An updated search was performed on 24 April 2020. Reference lists of studies meeting the eligibility criteria were screened for additional relevant studies. Furthermore, a search in the database ClinicalTrials.gov yielded 34 hits, none of which were relevant for our review. A comprehensive search strategy was developed with support from a medical librarian, using a combination of keywords and MeSH/thesaurus terms. An example of the search terms is outlined in Table 1. No limitations were placed on language or publication status.

Table 1 Search terms
Full size table

Eligibility criteria

Eligibility criteria were defined using a PICOS model (Population, Intervention, Comparator, Outcome and Study design) [53,54,55] reported in Table 2.

Table 2 PICOS for the study
Full size table

Exclusion criteria

Studies were excluded if triage was conducted by health professionals other than PTs; or the PTs conducting the triage were working in advanced or extended scope in specialties outside musculoskeletal / orthopaedics.

Study selection

Two reviewers (KS and SB) independently screened titles and abstracts for eligibility using the online screening tool Rayyan [56]. Discrepancies were resolved by discussion. The full texts of potentially relevant papers were then retrieved and independently reviewed for eligibility, and differences were discussed when necessary. A third reviewer (ML) was consulted when the two reviewers could not reach consensus.

Data extraction and analysis

Data extraction from included papers was performed by one reviewer (KS) and checked for accuracy by another reviewer (SB). Disagreements were resolved by discussion. Information was extracted on study characteristics (e.g. publication year, country, setting); patient characteristics (e.g. number of participants, age, gender); description of the triage model and any control interventions; follow-up; types of outcomes assessed (patient reported outcomes and experiences; care processes; costs); and the outcome data reported. Main findings of each study were summarized and presented in tables. Mean differences with 95% CI were calculated and presented in a summary of findings Table. A synthesis of the findings is presented in narrative format. Because of heterogeneity between studies regarding study designs and outcome measures used, meta-analysis was not conducted.

Assessment of methodological quality

All authors were involved in assessing methodological quality of included studies working in randomly-assigned pairs to assess 3–4 studies each. Authors rotated through pairs to ensure consistency of decision-making. Differences in quality assessment scores were discussed by all authors and resolved using consensus. Quality assessment of the trial by Samsson et al. [24, 57, 58] was done by KG and JM, who had not been involved in that study. Assessment was made using a slightly modified version of the Downs and Black checklist (see Additional file 1) [59]. This checklist is appropriate for assessing study quality for RCTs and studies of other designs, and has better reliability and validity than other tools for studies of varied design [60, 61]. As previously reported [62] item 27 (study power) was modified for our review. The modified Downs and Black score ranges were assigned corresponding quality levels according to previously reported cut-offs [63]: excellent (26–28), good (20–25) fair (14–19) and poor (</ = 13). To reduce the risk of bias due to poor methodological quality, only studies with fair to excellent quality (i.e. score above 13) were included in the synthesis [64].

Assessment of certainty of evidence

To assess confidence in the combined estimates of effect, the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach was applied for each outcome, using the following criteria: risk of bias, consistency, directness, precision, and reporting bias [65]. Two reviewers (KS and SB) performed assessments of certainty of evidence. A third reviewer (ML) was consulted if there was disagreement.

Results

Search results

The search process and results are reported in a PRISMA flowchart [51], modified to account for the two literature searches (Fig. 1). The database searches yielded a total of 1593 citations, with six additional papers identified through reference list screening. After removing duplicates, 1312 papers remained, and after screening titles and abstracts as well as full text articles when necessary, 15 relevant studies were included, reported in 17 papers. After methodological quality assessment, three poor quality papers were excluded (see Additional file 2), leaving 12 relevant studies reported in 14 papers. Of the 12 included studies, five studies (reported in seven papers) [24, 57, 58, 66,67,68,69] had not been included in previous systematic reviews.

Fig. 1
figure 1

Flow diagram of selection process and search results. Adapted from Liberati et al. [51]

Full size image

Characteristics of included studies

Table 3 reports on the included studies; two RCTs (395 participants) and ten cohort studies (783 participants). These studies were published between 1999 and 2020, and were conducted in the United Kingdom [23, 70], Canada [20, 66, 69, 71, 72], Ireland [73] Australia [19, 67, 68], and Sweden [24, 57, 58]. Nine studies were conducted in orthopaedic hospital/outpatient departments/clinics, one in a tertiary care center, one in a knee screening clinic, and one in primary care.

Table 3 Characteristics of included studies
Full size table

Participant numbers varied from 30 to 383 (age range 15–85 years, mean age (where available) ranging from 46.2 to 67 years). The majority of studies included patients with one or two problem areas; knee [70, 73], hip and knee [20, 68, 71], shoulder [67, 69, 72], shoulder or knee [66], shoulder, knee or spinal [19], with two studies having a broader inclusion of musculoskeletal or orthopaedic conditions [24, 57, 58, 67].

Quality appraisal

Methodological quality was fair to good (see Table 4), with the two included RCTs (reported in four papers) assessed as having good methodological quality (range 19–22), and the ten cohort studies assessed as having fair to good methodological quality (range 14–23). Certainty of evidence was largely constrained. Two RCTs and three cohort studies had low to unclear risk of bias [23, 24, 57, 58, 67,68,69], and seven cohort studies had unclear risk of bias [19, 20, 66, 70,71,72,73]. One study [19] provided inadequate information and reported incomplete data. None of the included studies blinded healthcare professionals nor outcome assessors. Only one study [67] blinded patients to the profession of their assessors i.e., the patients did not know which clinician was the PT and which was the OS, and were unaware of their clinical decisions until all assessments had been completed. Almost all studies explicitly stated that they blinded assessors to each other’s findings [20, 66,67,68,69,70,71,72,73].

Table 4 Modified Downs and Blacks score
Full size table

A summary of findings of the included studies is presented in Table 5, and outcomes are discussed individually below.

Table 5 Summary of findings
Full size table

Patient-reported outcomes

Pain, disability, health state, psychological status, and health-related quality of life were investigated in both RCTs [23, 58]. There were no significant differences in any outcome, between PT-led orthopaedic triage and standard care at three, six or 12 months’ follow-up (Table 6). For all outcomes, certainty was downgraded one level due to study limitations (mainly lack of blinding), and an additional level for serious imprecision (large 95% CIs that included possible unfavourable effects). For health state, certainty was further downgraded one level due to serious inconsistency (effects in opposite directions). Certainty of evidence was thus very low for health state, and low for all other patient-reported outcomes.

Patient-reported experiences

These were reported in both RCTs [23, 57], as patients’ perceptions of the quality of care received, and their satisfaction with it (Table 6). Samsson et al. [57] reported significantly higher perceived quality of care for PT-led orthopaedic triage than for standard care, for all reported elements (e.g. caregivers’ medical technical competence, identity-oriented approach). Daker-White et al. [23] reported that patients were more satisfied after PT-led orthopaedic triage compared with standard care. Certainty of evidence for this outcome was moderate. The RCTs were downgraded one level due to study limitations, mainly lack of blinding. Four of the cohort studies also reported significant differences regarding patient satisfaction, favouring PT-led orthopaedic triage [19, 20, 66, 72] and one reporting no difference between groups [69].

Table 6 Patient-reported outcomes and experiences
Full size table

Sick leave

(Days off work) was investigated in one RCT [58], which reported no differences between study arms (Table 6). Certainty of evidence was assessed as low; this having been downgraded one level due to study limitations (mainly due to lack of blinding) and an additional level due to serious imprecision (large 95% CIs that possibly included unfavorable effects).

Process outcomes

Inter-rater reliability of decision-making, care processes and costs are reported in Table 7.

Table 7 Outcomes related to care processes and cost effectiveness
Full size table

Surgery conversion rate (SCR) refers to the number of patients who proceeded to surgery following assessment, with the decision of the OS considered the gold standard. Surgery conversion rate was assessed in one RCT [24] and three observational studies [66, 68, 73]. Certainty of evidence was moderate, as the RCTs were downgraded one level due to study limitations (mainly lack of blinding). Samsson et al. [24] reported a higher SCR after PT-led orthopaedic triage (55%, compared with 25% in standard care) (p = 0.002). The observational studies reported higher SCR rates: Napier et al. [66] reported 91% after PT-led orthopaedic triage compared with 22% for standard care; Jovic et al. [68] 78% compared with 38%, and Ashmore et al. [73] reported a conversion to surgery rate of 84% with PT led orthopaedic triage.

Good agreement between PTs and other medical professionals regarding treatment approach (conservative or surgical) is relevant to the safety of PT-led orthopaedic triage, as it is important that PTs do not offer a lesser standard of assessment and decision-making than OS. Comparison between treatment approach made by PTs and OS was reported in seven cohort studies [19, 20, 67,68,69, 71, 72], indicating overall strong percentage agreement, ranging from 70 to 94%, and inter-rater agreement (κ-values ranging from 0.38 to 0.77 and AC1 = 0.93). However, the certainty of evidence was low due to the study design.

Referral for investigations was evaluated in the two RCTs [23, 24] and in two cohort studies [20, 69] with mean percentage difference in referral proportions between PTs and OS ranging from − 27.6 to 32.8%. Samsson and Larsson [24] reported a significantly lower proportion of patients referred for investigations for PT-led OT compared with standard care (17% vs 29%; difference − 12% [95% CI − 23 to 0.6%], p = 0.039). Daker-White et al. [23] reported that a higher proportion of PTs ordered no investigations at all (48% vs 15%; p < 0.001) and fewer plain X-rays (14% vs 41%; p < 0.001) compared with standard care. Certainty of evidence was moderate. The RCTs were downgraded one level due to study limitations, mainly lack of blinding. The cohort studies reported equal rate of investigations between PTs and OSs [20, 69].

Agreement on referral for investigations was reported in five cohort studies [20, 67,68,69, 72], overall strong percentage agreement (70 to 97%) and inter-rater agreement (κ-values ranging from 0.42 to 0.93 and AC1 = 0.87). However, certainty of evidence was low due to study designs.

Agreement on diagnosis between PTs and OS or imaging/surgery findings was investigated in seven cohort studies [20, 67, 69,70,71,72,73] with overall good percentage agreement (ranging from 72 to 98%) and inter-rater agreement (κ-values ranging from 0.68 to 0.94 and AC1 = 0.72. However, certainty of evidence was low due to the study designs.

Waiting time was investigated in one RCT [24], reporting significantly shorter waiting times to PT-led orthopaedic triage; 19 (SD 12) compared with 28 (SD 14) days for standard care (p < 0.001). One cohort study [72] also reported significantly shorter waiting times with PT-led OT (p < 0.001). Certainty of evidence was assessed as moderate. The RCT was downgraded one level due to study limitations, mainly lack of blinding.

Cost effectiveness of PT-led orthopaedic triage compared with standard care was investigated in one RCT [23] . No significant differences were found in terms of direct costs to the patient, or in organisational (NHS primary care) costs. Direct hospital costs were significantly lower in the PT-led orthopaedic triage arm (mean cost per patient £256 vs £498 in the standard care arm (p < 0.001)), as PTs were less likely to order radiographs or refer patients for orthopaedic surgery. Certainty of evidence was moderate (downgraded one level due to study limitations, mainly lack of blinding).

Adverse events were not reported as having occurred in any study.

Discussion

This review updates the current body of secondary evidence on the impact of PT-led orthopaedic triage compared with standard OS care, for people with musculoskeletal disorders. We identified 12 studies, five of which had not been included in previous reviews [47,48,49,50]. Our review found evidence of moderate certainty that PT-led orthopaedic triage results in higher surgery conversion rates, reduces investigation referral rates and waiting times for orthopaedic consultation, and improves quality of care (patient satisfaction). We found low-certainty evidence for moderate to high agreement between PT-led orthopaedic triage and standard care in diagnosis, treatment and investigation approaches. For patient-reported outcomes, there is low-certainty evidence for no difference between PT-led orthopaedic triage and standard care. Taken together, our findings suggest that PTs are equally effective as OSs in assessing and managing patients referred for orthopaedic consultation, although the low number of studies and the heterogeneity in methodological quality limits our ability to draw firm conclusions.

The evidence base for the effectiveness of PT-led orthopaedic triage is growing, particularly for the assessed process outcomes. Our findings build on the findings of the earlier reviews [47,48,49,50], which concluded low certainty of evidence for the positive impact on both patient and process outcomes of PT-led orthopaedic triage. With the addition of the recent studies identified in our review, the certainty of evidence for the effect on process outcomes has increased, whilst the evidence for PT-led orthopaedic triage impact on patient health outcomes and patient satisfaction with quality of care remains low.

Considering that PTs cost the health system approximately half as much as an OS [27], and as they reduced rates of referrals for investigation [23, 24], PT-led triage could be a cost effective way to address growing waiting lists and ensure that patients needing an orthopaedic consultation receive it in a more timely manner. The significantly shortened waiting times using a PT-led triage model [24, 72] could also have a further impact on economic parameters. A recent scoping review [74] suggests that waiting has a negative effect on patient, institution and societal costs, although the body of evidence was scant. Furthermore, Morris et al. [75] suggest that prolonged waiting time for an orthopaedic consultation may impact quality of life.

A higher conversion to surgery rate suggests that the triage process undertaken by PTs provides patients in need of surgery with an expedited path to requisite treatment, not that an increased number of surgeries are performed. This review generally indicates that PTs can provide earlier assessment than an OS, and earlier identification of patients requiring orthopaedic interventions may facilitate these patients being fast-tracked to further investigations and an OS consultation. Furthermore, PT-triage can offer patients that may not require an OS assessment in the first instance alternative (conservative) care options. An additional benefit is that PTs working with triage assessment tend to give patients advice on conservative management and self-care, e.g. home exercise [20, 57], which may contribute to quicker recovery and further reduce the need for additional healthcare visits. Fennely et al. [76] explored patients’ experience after PT-led triage and found that that patients valued that the PT listened to them, and that they could provide more specific advice regarding self-management, a finding also identified by Samsson et al. [57]. Both studies also found that patients had confidence in the PTs’ professional ability.

However, there are limitations when evaluating process outcomes with this model of care. The lower rate of investigation after PT-led triage which is perceived as a good outcome, could in fact be a limitation, as some pathology might be overlooked. Furthermore, the PTs might be biased towards avoiding both investigation and surgery as an intervention, whereas the OS might be biased towards performing surgery. One of the reasons for implementing PT-led triage has been to reduce waiting times, and the findings in this review of reduced waiting times with PT-led triage is a factor that could change with time, as demand increases. Furthermore, it could be argued that costs might increase, due to many patients being referred for further physiotherapy after PT-led triage [24]. Another issue with this model of care is the PTs’ advanced role, allowing professional development but also entailing greater responsibility [77]. This increased responsibility should be reflected in salaries, and with that, costs might increase.

The varying titles and positions of PTs, such as advanced practice, extended scope or experienced PTs, and the varying level of training and tasks undertaken potentially constrain comparison between studies. In most countries no formal training exists and availability of experienced PTs might be scarce. Furthermore, PTs working in this role have expressed a lack of formal education [78]. In order to standardise this model of care internationally and improve reporting, a framework for competencies and education standards should be used, such as the one provided recently by Fennelly et al. [79].

A limitation when evaluating this model of care is the lack of consistency amongst studies in patient-reported outcomes, or ways in which patient satisfaction was measured. This limited comparison between studies and review conclusions. Fennely et al. [80] also identified the lack of measurement of global improvement, psychological well-being and/or work ability in their recent systematic review on the use of outcome measures in advanced PT practice, and that although patient satisfaction was frequently measured, non-standardised, locally-devised tools were used.

Strengths and limitations of this review

We modified the PROSPERO-registered protocol once the review had commenced to include prospective comparative studies, because of the paucity of RCTs. Although studies published in any language were eligible for inclusion, only English-language articles were identified. Important limitations of the review are that we only identified two RCTs, and that risk of bias to varying extent were seen in all included studies. Lack of blinding was a problem in all studies.

Definition of terms was problematic. The term ‘triage’ may have been interpreted differently by different researchers. The distinction between triage, screening, and assessment is subtle. We included all these terms in our search strategy, but the development and adoption of consistent terminology and operational definitions for these forms of assessment could facilitate improved understanding among researchers, clinicians and policy makers. Whilst standard care appears to be a relatively vague term, in our review as in all studies included in the review, standard care was considered to be referral from a GP to, and assessment by, an OS.

The included studies involved patients with varied musculoskeletal conditions [24, 57, 58, 67] or patients with one or two problem areas; knee [70, 73], shoulder and knee [66], hip and knee [20, 68, 71] and shoulder [67, 69, 72], shoulder, knee or spinal [19] respectively. Thus, the conditions included in this review are heterogeneous, and the significance of differences between PT-led orthopaedic triage and standard care may have been confounded by different care needs of patients with different conditions.

The Downs and Black checklist used in this review is appropriate for assessing study quality for both RCTs and cohort studies, and has better reliability and validity than other tools for studies of varied design [60, 61]. However, we found it to be less applicable to the cohort studies than to the RCTs. In hindsight, the checklist could have been modified so that the items addressing for example follow-up and randomisation could be scored ‘not applicable’ in cohort studies (thus adjusting the denominator). Therefore, methodological quality of the assessed papers in this review might be lower than previously reported. In order to avoid assessor bias, the quality assessment of the trial by Samsson et al. [24, 57, 58] was done by the authors KG and JM, who had not been involved in that study. Lastly, there is a risk for publications bias; mean differences were higher in the smaller studies included, suggesting that our overall estimate is potentially overestimated.

Future research

There is a need for more, and better quality, primary studies, particularly RCTs, so that the evidence base will have less bias and more certainty. Future studies should be innovative about how patients, clinicians and evaluators can be blinded to allocation to intervention arms. This would improve methodological quality scores and increase the believability of study findings.

Conclusions

There is a growing body of evidence, of low to moderate certainty, that PT-led orthopaedic triage and OS make similar diagnosis, treatment and investigation decisions for patients with musculoskeletal disorders, and that patients are equally or more satisfied with quality or care. PT-led orthopaedic triage is safe, less expensive than OS, and effective in triaging patients on orthopaedic waiting lists to ensure that patients are directed to the most appropriate care as quickly as possible. Consistent with previous reviews, the current body of evidence is limited by study volume, design and quality, as well as heterogenous outcome measures. More research of higher quality is required to investigate the impact of PT-led orthopaedic triage on patient reported outcomes and experiences, work ability, sick leave, cost-effectiveness, and length of waiting time. Rigorous RCTs are needed to increase certainty of the evidence.