Key findings from the reviewed papers
| Authors | Outcome type | Summary of findings | Study limitations |
| Aiken and McColl33 | Professional competency | Diagnostic concordance between physiotherapist and surgeon: k=80% agreement for knee diagnoses, agreed 21 out of 24 times. | Low subject numbers and only two clinicians so low external validity/generalisability of findings. Diagnostic accuracy and treatment concordance not reported, as no specific knee data. |
| Aiken et al 34 | Waiting time, patient satisfaction | Physiotherapist deemed 36/107 (34%) as non-surgical cases. All 36 were offered conservative treatment and so were 64/71 (90%) that were sent to see surgeon. Referral to consultation waiting time reduced from average of 140 to 40 days. Numbers on surgical waiting list went from 200 to 59. Satisfaction reported as being high or very high (services and skill). Nobody requested a consultation with consultant, 80% response rate. | More than one time point needed to evaluate longer term success. Only one physiotherapist, so low generalisability. Patient satisfaction survey not referenced and no definition of ‘high satisfaction’ given. Potential for reporting bias through unreturned questionnaires Resource implications of adding the physiotherapist to clinic and cost effectiveness were not evaluated. Rest of data about joint replacement care. |
| Damask TEAM26 (primary paper) Damask41 Damask42 | Clinical and cost effectiveness | No significant difference between MRI and orthopaedic groups for changes in diagnosis (p=0.79) or treatment plans (p=0.059). Significant increase in diagnostic (p<0.0001) and therapeutic confidence in the MRI group. Compared with controls, patients in MRI group improved mean SF-36 physical functioning score by 2.81 (95% CI: −0.26 to 5.89) (p=0.072). Patients randomised to MRI improved mean Knee QoL-26 physical functioning scores by 3.65 (95% CI=1.03 to 6.28) (p=0.007). At a cost per quality adjusted life year threshold of £20 000, there is a 0.81 probability that early MRI is a cost-effective use of National Health Service (NHS) resources. | Pragmatic RCT with no allocation concealment or blinding. |
| Decary et al 37 | Professional and patient flow | High diagnostic inter-rater agreement between the physiotherapist and physicians (k=0.89; 95% CI: 0.83 to 0.94). Good inter-rater agreement for triage recommendations of surgical candidates (k=0.73; 95% CI: 0.60 to 0.86). | Variation in experience between the physiotherapists and physicians 1–2 vs 20 years. Only one physiotherapists and four physicians and interperson comparison varied. Low external validity |
| Desmeules et al 35 | Professional competency, resource use and patient satisfaction | Physiotherapist vs consultant: very high diagnostic agreement (88%): k=0.86; 95% CI: 0.80 to 0.93 and triage (surgical vs conservative) agreement k=0.77; 95% CI: 0.65 to 0.88. No difference in imaging requested: physiotherapist vs consultant: X-ray 42% vs 50%, MRI 13% vs 16%, CT 17% vs 20%. Conservative treatments: physiotherapist referred for more advice and education (98% vs 81%); more NSAIDS (47% vs 24%); more joint infiltrations (43% vs 11%); more supervised physio (62% vs 16%) and more home exercises (84% vs cons. 10%) (p<0.05). No difference between clinicians for non-prescription analgesics, walking aids and orthosis. Visit length for physiotherapists significantly longer (13.0 min vs 11.2 min) (p<0.05). Patients significantly more satisfied with physiotherapist 93.2 vs 86.1 (p<0.001). | The design would have been strengthened by repeating data collection over more than one time point to re-evaluate satisfaction after trying out the intervention/course of treatment the patients were referred for. Longer term should collect data on rereferral rates. Cost and cost–benefit of the physiotherapist pathway are not discussed. Of note, there were no referrals for weight management. |
| Dickens et al 27 | Professional competency | Correct diagnosis by surgeon 92% cases, physiotherapists 80%–84% | Potential for bias, all treatment based on consultant assessment only. Bias relating to improvement at arthroscopy assumed to confirm diagnosis made by surgeon. No statistical analysis (kappa). |
| Doerr et al 39 | Time | Reduced waiting time for initial orthopaedic assessment from 10 to 3 months. Improved equity of access through service redesign | Reporting bias as high level of patient satisfaction in terms of wait time for assessment, but no data presented from experience surveys or interviews. Cost of this redesign is not discussed. Most of the findings are not relevant as report on surgical/postsurgical care. |
| Farrar et al 28 | Resource use and time | 1588 referrals over 5 years, 432/1588 hip or knee referrals. 206 referred to orthopaedic clinic and 226 to MCATS. Groups were similar for gender, and joint affected. Orthopaedic patients significantly older (mean of 8 years) (p=0.01). Orthopaedic clinic longer wait for initial consultation (4 days) (p=0.05). MCATS had longer time from referral to diagnosis (11 days) (p<0.001), had more consultations before diagnosis (p<0.001). Those who were reviewed by a surgeon had more consultations than those seen by APP (p=0.03). Cross-sectional imaging used more in MCATS (p=0.04). Surgical management more common for orthopaedic patients (36% vs 16%) and joint injection (19% vs 12%) p<0.001. Higher use of non-surgical treatment in MCATS (67% vs 44%) p<0.001. 5% MCAT patients referred for orthopaedic opinion. | Only one GP surgery. Follow-up over more than one time point would allow evaluation of rereferral rates. Cost effectiveness needs to be evaluated. |
| Gwynne-Jones et al 15 | Patient flow and resource use | Referrals: 150 (44%) hip OA, 189 (56%) knee OA. 54 patients referred directly for FSA (mean Oxford knee score 13), and 89 after subsequent review. Oxford knee score for those in FSA slightly worse than those managed in joint clinic (p<0.001). Of 143 referred for FSA, 115 triaged to surgical route, 18 recommended surgery but did not meet the prioritisation score, 10 not recommended surgery. Oxford knee score of those managed conservatively improved from 22 to 25 (p=0.0013). | It is assumed that the patients were consecutive |
| Inglis et al 38 | Patient flow and resource use | HIPs: 87 (10%) referrals from surgeons so directly listed for arthroplasty. 393 (47%) accepted for FSA. 66 (8%) declined as insufficient referral info. 134 (16%) not seen as lower priority condition. 158 (19%) not seen due to insufficient capacity for FSA/op KNEES: 113 (13%) surgeon referrals and patients directly listed. 295 (33%) referrals accepted into FSA. 84 (9%) declined due to insufficient referral information. 173 (19%) low priority condition so not give n FSA. 230 (26%) no capacity in service to be given an FSA appointment. | Mainly arthritis population as other types of conditions and traumatic conditions seen via another pathway/healthcare route. Therefore, generalisability less clear. Potential selection bias as unclear who ended up in the no capacity group. |
| Johnson et al 29 | Resource use | Fast-track screening criteria correctly predicted the outcome of the treatment offered at the orthopaedic clinic in 38/52 patients. Of those fulfilling the criteria for the fast-track screening clinic, 23/25 had a THR. Of the 28 who did not fulfil the criteria, 15 had a THR. | Reporting bias, the high conversion in the group who did not fulfil the criteria for the fast-track service is not acknowledged. This new service could create inequality in care. Better design required to evaluate predictive value of screening tool using regression analysis in a larger sample size. |
| MacKay et al 36 | Patient flow and professional competency | Good agreement on recommendation for orthopaedic consultation surgeon 82% vs physiotherapist 86.9%, k=0.69. Good agreement on recommendation for arthroplasty surgeon 43.5%, physio 32.3%, k=0.70. Surgeon and physio agreed on type of knee complaint in 69% of cases. There was variation in the conservative treatment given. Surgeon more frequently referred for rehabilitation services. Physiotherapists more frequently followed up patients and gave exercise and education advice in clinic. | No kappa values for clinical diagnostic accuracy or nonsurgical management recommendations. Testing order in clinic not clear. |
| Parfitt et al 16 | Clinical competency Resource use | Over 2 years APP’s listed 130 patients for THR, 127/130 had a THR. This was compared with traditional route of referral by GP to orthopaedics. Waiting time for surgery APP vs GP route 21.4 weeks vs 24.7 weeks. Potential cost saving of £145 for those directly listed. | Reporting bias, numbers receiving THR from traditional referral route not reported. Unclear if groups matched at baseline. It is not clear how a potential saving of £145 for the APP route was calculated. Not consecutive cases. |
| Pearse et al 30 | Patient flow and resource use | 150 cases, 33% (50) knees, of these 43% (17) were referred to a consultant. The outcome of the consultant review was arthroscopy 11 cases, advice five cases and injection one case. This means that for the knee the APP’s did not met the benchmark of independently assessing 85% of cases. | Most of the data collected is not split by joint, and therefore cannot be reported. Although a referenced protocol for triage was used the referrals were also checked by a consultant as to who should see APP. |
| Rabey et al 31 | Patient flow and resource use | 9% of all new referrals seen by APP’s referred for surgical opinion. Of these, 42% were knee conditions with 84% going on to have surgery. Of patients seen by APPs 36% referred for a knee X-ray and 23% for knee MRI. | Not all data broken down per joint so cannot be reported. No numbers given just percentages. Unsure if patients consecutive so potential selection bias. |
| Robling et al 32 | Patient experience | Four themes identified: 1. Inadequate information. 2. Social and psychosocial cost of waiting. 3. Varying ability to cope, both passive and proactive strategies demonstrated. 4. Reported variation in clinician effectiveness in managing the condition and ability to provide support. A care pathway with improved information provision may help improve patient well-being. | Unclear if there was more than one researcher analysing the manuscripts (selection bias). Good reported, credible data. Low generalisability |
| Smink et al 40 | Resource use | Most commonly used treatment modalities were education, paracetamol, lifestyle advice, exercise therapy and non-steroidal anti-inflammatories. Cumulative percentage of users for each modality increased over time. At 2 years, percentage use of modalities for each step in the pathway was: Step 1: education 242 cases (82%), paracetamol 250 cases (83%), lifestyle advice 214 cases (73%) and glucosamine 95 cases (34%). Step 2: exercise therapy 187 cases (63%), NSAID cases 155 (54%), dietician 27 cases (14%). Step 3: intra-articular injections 65 cases (23%), multidisciplinary care 23 cases (8%). Low use of dietetics In 0–6 months, 67 cases (21%) had surgeon consultation. After 2 years, this increased to 129 cases (45%). In 0–6 months, 16 cases (5%) had surgery and at 18 months, this had increased to 39 cases (14%). | Low uptake by patients. Patients were not recruited consecutively, GPs could select. |
APP, advanced physiotherapy practitioner; FSA, First Specialist Assessment; GP, general practitioner; MCATS, musculoskeletal assessment service; OA, osteoarthritis; RCT, randomised controlled trail; THR, total hip replacement.