Safety of disinvestment in mid- to late-term follow-up post primary hip and knee replacement: the UK SAFE evidence synthesis and recommendations

CPRD–HES

The CPRD comprises the entire computerised medical records of a sample of patients attending general practices in the UK. 72 The CPRD contains information on more than 11 million unique patients (i.e. around 7% of the UK population) registered at over 600 general practices in the UK. With 4.4 million active (i.e. alive and currently registered) patients meeting quality criteria, approximately 6.9% of the UK population are included, and patients are broadly representative of the UK general population in terms of age, sex and ethnicity. 73,74 GPs in the UK play a key role in the delivery of health care by providing primary care and referral to specialist hospital services. Patients are registered with one practice that stores medical information from primary care and hospital attendances. The CPRD is administered by the MHRA. The CPRD records contain all clinical and referral events in both primary and secondary care, in addition to comprehensive demographic information, prescription data and hospital admissions data. Data are stored using Read and Oxford Medical Information Systems codes for diseases that are cross-referenced to the International Classification of Diseases, Tenth Edition (ICD-10). Read codes are used as the standard clinical terminology system within UK primary care. Only practices that pass quality control are used as part of the CPRD database. Deleting or encoding personal and clinic identifiers ensures the confidentiality of information in the CPRD.

The CPRD GOLD data were linked to the HES database. CPRD already provide access to the HES data that are held under the CPRD Data Linkage Scheme. CPRD and HES-linked data are available for around 50% of patients in the CPRD database. Previous research by the CPRD team has shown that linked practices/patients are representative of the CPRD GOLD population as a whole. 73

NJR–HES–PROMs

Starting in 2003, the NJR has collected information on all hip and knee replacements performed each year in both public and private hospitals in England, Wales, Northern Ireland and the Isle of Man. Data are entered into the NJR using forms completed by surgeons at the time of surgery and revision operations are linked using unique patient identifiers. Data recorded in the NJR includes prosthesis and operative information (e.g. prosthesis type, approach and thromboprophylaxis use), patient information [e.g. age, sex, BMI and American Society of Anesthesiologists (ASA) grade] and surgical and unit information (e.g. surgeon and unit caseload and public/private status).

The HES database holds information on all patients admitted to NHS hospitals in England, including diagnostic ICD-10 codes providing information about a patient’s illness or condition and OPCS Classification of Interventions and Procedures version 4 (OPCS4) procedural codes for surgery. HES include episodes of care delivered in treatment centres (including those in the independent sector) funded by the NHS, episodes of care in England when patients are resident outside England and privately funded patients treated within NHS England hospitals. HES cover a smaller geographical area than the NJR (excluding patients operated on in Wales and Northern Ireland) and do not include privately funded operations. However, HES provide additional information for every patient (including detailed comorbidity information and deprivation indices) and about every procedure (including LOS and need for blood transfusion or critical care). Additional records contain details of readmissions, reoperations and revisions not recorded in the NJR database. Data for all-cause mortality are provided by the Office for National Statistics (ONS) and are linked to the HES database.

Since April 2009, PROMs data have been collected on hip and knee replacements performed in public hospitals in England. Preoperative and 6-month postoperative quality-of-life questionnaires (i.e. the EQ-5D75) and joint-specific PROMs (i.e. the OHS76 and OKS77) are collected along with patient-reported measures of preoperative disability and postoperative satisfaction. For this analysis, we used NJR records linked to data from the HES and PROMs databases on all hip and knee replacement operations.

Participants

Anonymised records were extracted for all patients aged ≥ 18 years receiving primary hip or knee replacement surgery. For CPRD–HES data, the time span covers the years 1995–2016. For NJR–HES–PROMs data, the time span covers the years 2008–16. Patients were included if they had primary THR or TKR, primary hip resurfacing (metal-on-metal hip resurfacing) or UKR. We excluded patients who had revision surgery and total joint replacement (TJR) of unspecified fixation. The following exclusions were made to remove potential case mix issues: diagnostic codes indicating fracture or cancer of the hip bones, other injuries due to trauma (e.g. transport accidents and falls), non-elective admissions and a diagnosis other than primary hip/knee osteoarthritis. There was overlap between patients in the two data sources (e.g. around 7% of patients receiving knee replacement between 2009 and 2016); however, these anonymised data sets were analysed independently of each other.

Secondary care predictors

Patient-level characteristics available in NJR and HES include age, sex, BMI, area deprivation, rurality, ethnicity, Charlson Comorbidity Index (CCI)78 (calculated from HES using ICD-10 codes at the time of admission for surgery) and ASA grade. Data from the NJR provided additional information on surgical and operative factors, including whether or not a minimally invasive technique was used, annual surgeon volume/caseload, operative time, grade of operating surgeon, surgical approach, patient position, implant fixation, type of mechanical or chemical thromboprophylaxis and unit type (e.g. public, private, independent sector treatment centre). Data from the PROMs database provided additional information on symptoms of pain, function and HRQoL preoperatively and at 6 months post surgery. Pain and function were measured using the OHS and OKS. The EQ-5D consists of five questions (assessing mobility, self-care, ability to conduct usual activities, degree of pain/discomfort and degree of anxiety/depression), ranging from 1 (no problem) to 3 (severe problems). The EQ-5D can be expressed as a preference-based overall index (graded from –0.594 to 1) or as ordinal responses for each category. Preoperatively, patients rate their general health on a five-point Likert scale, from very poor to excellent, and are asked to report if they considered themselves to suffer from a disability (yes or no).

Primary care predictors

The CPRD database includes information on age, sex, BMI, joint replaced (hip/knee), year of joint replacement operation, recorded diagnosis of osteoarthritis (yes/no), fracture pre surgery (yes/no), calcium and vitamin D supplements, use of bisphosphonates, use of selective oestrogen receptor modulators, oral glucocorticosteroid therapy, smoking status and alcohol intake recorded closest to the date of the primary surgery, Index of Multiple Deprivation (IMD), region of UK, comorbid conditions (i.e. asthma, malabsorptive syndromes, inflammatory bowel disease, hypertension, hyperlipidaemia, ischaemic heart disease, stroke, chronic obstructive pulmonary disease, chronic kidney failure, neoplasms and diabetes) registered by the physician and use of drugs that can affect fracture risk (e.g. proton pump inhibitors, antiarrhythmics, anticonvulsants, antidepressants, anti-Parkinson drugs, statins, thiazide diuretics and anxiolytics).

NJR–HES–PROMs-linked data

In the NJR, before personal data and sensitive personal data are recorded, express written patient consent is provided. The NJR records patient consent as either ‘yes’, ‘no’ or ‘not recorded’. With support under Section 251 of the National Health Service Act 2006,80 the Health Research Authority (HRA) Confidentiality Advisory Group (CAG) allows the NJR to collect patient data when consent is indicated as ‘not recorded’. Approval for NJR data was received on 3 October 2016 (NJR internal reference ‘Is it safe to completely disinvest in TJR follow-up or will this expose people to harm?’). CAG Section 251 approval was received on 24 February 2017 (CAG reference 17/CAG/0030). A Data Access Request Service (DARS) application was made to NHS Digital (formally known as the Health and Social Care Information Centre) for HES and PROMs data to be linked to data from the NJR. This was approved, with the data-sharing agreement signed by NHS Digital on 4 July 2018 and by Oxford University Research Services on 31 July 2018.

CPRD–HES linked data

The CPRD has ethical approval from the National Research Ethics Service (NRES) for all anonymised, observational research. The study was approved by the Independent Scientific Advisory Committee (ISAC) for Medicines and Healthcare products Regulatory Agency (MHRA) database research (protocol number 11_050AMnA2RA2).

Data summary

A team of clinicians has supported us in defining code lists for variables based on ICD-10 diagnosis codes, including OPCS4 operation codes, Read codes for GP consultations and product/British National Formulary codes for medications. The team has worked closely with the data manager in resolving queries and checking codes. All data sets were made ready as ‘flat’ files for statistical analysis.

Patient demographics

Older age at the time of primary operation was associated with a lower risk of mid- to late-term revision (Tables 3–6). The effect of age was linear and the association was stronger for knee replacement. For knee replacement, a 1-year increase in age at surgery reduced the risk of outcome by 5%. For THR, a 1-year increase in age at surgery reduced the risk of outcome by 3%. These findings were consistent across the CPRD–HES and NJR–HES–PROMs data sets. There was no effect associated with sex for patients receiving THR. However, for knee replacement, males had a greater risk of mid- to late-term revision than females. This was observed in only the CPRD–HES data, in which males had a 32% increased risk of outcome, but the effect size was weaker and non-significant in the NJR–HES–PROMs data set.

There was no effect of obesity on outcome for THR. However, for knee replacement, an association was seen in the NJR data set. In the NJR data set, when compared with patients with a normal BMI, underweight patients were at an increased risk of revision and obese patients were at reduced risk of mid- to late-term revision. For THR, there was no effect of IMD deprivation; however, for knee replacement in the NJR data set, patients in the most deprived areas were less likely to undergo mid- to late-term revision. Note that there was no such association with obesity or deprivation for knee replacement observed in the CPRD–HES data set. An association with ethnicity was observed for knee replacement in the NJR data set only, with patients of non-white ethnicity less likely to undergo a mid- to late-term revision than those of a white ethnicity.

Implant factors (NJR data set)

For knee replacement, none of the implant-related factors was associated with an increased mid- to late-term revision risk. For THR, there was an effect of the bearing surface, when, compared with metal-on-polyethylene (MoP), implants with a ceramic-on-ceramic (CoC) or ceramic-on-polyethylene (CoP) bearing surface had a reduced risk of revision. Larger head size appeared to increase revision risk, with the risk being lowest in those receiving implants with the smaller head size (≤ 28 mm) (see Tables 4 and 6).

Preoperative and 6-month postoperative patient-reported outcome measures

There was no association between preoperative PROMs at primary surgery and risk of mid- to late-term revision for THR. However, worse 6-month postoperative pain and function was associated with an increased risk of revision.

For knee replacement, there was a clear linear trend with the preoperative and 6-month postoperative OKS: patients with the most pain and functional limitations at the time of surgery, and 6 months after surgery, were substantially more likely to require mid- to late-term revision. Patients with preoperative anxiety/depression were found to be less likely to undergo a mid- to late-term revision operation (see Tables 4 and 6).

Primary care comorbidities and medication use

Through the CPRD data set, we were able to investigate comorbidities recorded prior to surgery and medication use. The findings observed were, once again, different for the hip and knee joints.

For THR, comorbidities played a role and patients with malabsorption or history of fracture (i.e. pelvis, proximal humerus, wrist/forearm, spine, rib) were more likely to require revision. In addition, those patients with hypertension were less likely to require revision. There was no association between revision risk and preoperative comorbidity in the case of knee replacement. Oral glucocorticoid steroid therapy was associated with a lower risk of revision for knee replacement, whereas the use of antiarrhythmics and anticonvulsants was associated with a higher risk. Antidepressant use was associated with a higher THR revision risk.

For the pain medication use, an increased revision risk was observed in opiate use for knee replacement and steroid injections for THR. When examining effects of medication use in more detail, by looking at DDDs calculated from the 1 year prior to the primary surgery and divided into tertiles, further patterns emerged. For THR, the use of statins was associated with an increased risk of outcome in those with a DDD of < 370 compared with no medication use. An effect of bisphosphonate use was also seen, with those in the highest tertile of > 340 DDD at an increased revision risk. The effect of steroid use was apparent in only the higher dose category of > 55 DDD. For knee replacement, the effect of opioids was only significant in the highest DDD tertile of > 600 DDD (see Tables 3 and 5).

Hospital Episode Statistics

Hospital Episode Statistics93 is a database that is controlled by NHS Digital. 94 HES contains data that are routinely collected in secondary care, relating to emergency departments, admitted patient care and outpatient episodes at hospitals in England.

ResearchOne

ResearchOne95 is a database controlled by TPP. 96 ResearchOne contains data that are routinely collected in primary care by organisations using the SystmOne (The Phoenix Partnership, Leeds, UK) clinical information system. 97

Population

Patients were to be included in the study population if, between 1 April 2000 and 31 March 2015 (i.e. the index period), they had been admitted to a hospital in England for one of the following procedures:98 (1) hip replacement, (2) hip revision, (3) knee replacement or (4) knee revision. 99 To be included, patients also had to be aged ≥ 18 years at the time of the index episode. However, it is important to note that data items may have included those relating to a period when the patient was aged < 18 years. Definitions were based on those provided within the OPCS4 operation codes relevant to procedures recorded on the document published by the NJR. 98 Based on data provided within the NJR annual report,99 800,683 primary hip replacements, 89,023 hip revisions, 875,585 primary knee replacements and 54,278 knee revisions were reported (subject to specific inclusion and exclusion criteria) during the period from 1 April 2003 to 31 December 2015. Therefore, for a comparable period within the study, we anticipated that around 900,000 hip replacements and 1,000,000 knee replacements (i.e. 1,900,000 joint replacements in total) would be identified. Patients were to be excluded from the study population if they registered a type 2 objection100 with NHS Digital to prevent their identifiable data from any health and social care settings being released. 101 Although NHS Digital does not routinely apply type 2 objections (now national data opt-outs) to pseudonymised data requests, we requested that national data opt-outs be applied in the application for work package 2a (ResearchOne–HES).

NHS Digital was to select the study population using inpatient admissions recorded as admitted patient care episodes within HES and based on the above criteria.

Data items

Data items to be included for each patient in the study population were determined by the research team based on those that were necessary and sufficient for analysis. No patient identifiable data102 were required by the research team.

For patients included in the study population, the following data items were selected for inclusion from HES for the period from 1 April 2000 to 31 March 2015:

• Episodes from HES [accident and emergency (A&E)]103 that contain a diagnosis of dislocation/fracture/joint injury/amputation for an anatomical area of one of the following: (1) hip, (2) groin, (3) thigh, (4) knee or (5) lower leg. For each episode, data items relating to the following were to be included: (1) attendance, (2) diagnoses, (3) investigations, (4) treatments, (5) socioeconomic status, (6) provider and (7) demographics.

• Episodes from HES (admitted patient care)104 that contain a treatment specialty of trauma and orthopaedics or rheumatology. For each episode, data items relating to the following were to be requested: (1) admission, (2) cause, (3) diagnoses, (4) procedures, (5) specialty, (6) discharge, (7) spell, (8) provider, (9) waiting list, (10) demographics and (11) socioeconomic status.

• Episodes from HES (outpatient)105 that contain a treatment specialty of trauma and orthopaedics or rheumatology. For each episode, data items relating to the following were to be requested: (1) appointment, (2) waiting list, (3) referral, (4) diagnoses, (5) procedures, (6) specialty, (7) provider, (8) waiting list, (9) demographics and (10) socioeconomic status.

For patients included in the study population whose data relating to general practice is included in ResearchOne,106 the following additional data items were selected for inclusion from ResearchOne for the period prior to 1 April 2015:

• patient demographics and socioeconomic status

• selected general practice events, including (1) coded diagnoses/observations that relate to selected comorbidities,107 hip/knee replacement/revision and pain, and referrals to trauma/orthopaedics and rheumatology; (2) prescriptions and repeat prescriptions for opioid analgesics, non-steroidal anti-inflammatory drugs, non-opioid analgesics and compound analgesics, rubefacients, topical non-steroidal anti-inflammatory drugs, capsaicin and poultices, drugs that suppress the rheumatic disease process and corticosteroids; (3) non-coded referrals; and (4) practice registrations.

A detailed data specification was included as an appendix to the research protocol, which was provided as part of the approval processes (see Appendix 3, Tables 39–47). Data relating to general practice are included in ResearchOne for patients if (1) the patient is registered to a general practice that uses the SystmOne clinical information system, (2) the general practice has opted into ResearchOne and (3) the patient has not individually opted out of ResearchOne. Further information regarding the consent model used for ResearchOne can be found in the database protocol. Definitions of comorbidities were based on those provided in the Quality Outcomes Framework business rules previously published by Primary Care Commissioning and now published by NHS Digital,107 and included the following comorbidities: asthma, atrial fibrillation, high blood pressure, cancer, coronary heart disease, cardiovascular disease, chronic kidney disease, chronic obstructive pulmonary disease, dementia, depression, diabetes, epilepsy, heart failure, hypertension, learning disability, obesity, osteoporosis, psychosis, schizophrenia or bipolar affective disorder, peripheral arterial disease, palliative care, rheumatoid arthritis, smoking, stroke and stroke (transient ischaemic attack). All other clinical definitions were determined by an expert clinician.

Linkage

Linkage of data items from HES and ResearchOne was required to enable data items relating to the same patient across data sources to be determined. No persistent link exists between HES and ResearchOne and, to our knowledge, no linkage methodology had been previously enacted between the two sources. Prior to the design of work package 2a (ResearchOne–HES), another research project at the University of Leeds (LP-MAESTRO91,92) had proposed a methodology for linkage of data items between these two sources. LP-MAESTRO91,92 was progressing through approval processes on the basis of this methodology and establishing precedent. Therefore, the same linkage methodology (Figure 6) was adopted in the design of work package 2a (ResearchOne–HES).

FIGURE 6.

A summary of the key data flows and processing activities of the methodology.

Linkage was undertaken by NHS Digital in accordance with recommendations provided in the Caldicott Review: Information Governance in the Health and Care System108 document published by the National Data Guardian. Each data source generated two unique references for each patient: (1) a pseudonym that was generated by applying a one-way cryptographic hash function (SHA-512) to an input that comprises a cryptographic salt and the patient’s NHS number and (2) a source-specific identifier. For a patient with a given NHS number, each data source generated the same pseudonym but a different source-specific identifier. Both the pseudonym and source-specific identifier generated for each patient were specific to the study. Pseudonyms were used by NHS Digital to (1) communicate to TPP those patients for whom data are required from ResearchOne and (2) generate mappings between different source-specific identifiers for each patient. 91 NHS Digital and TPP provided the required data items to the research team, including only the source-specific identifier as the unique reference for each patient. The mappings generated by NHS Digital were then provided to the research team and used to enable data items relating to the same patient across data sources to be determined.

A detailed description of the data linkage methodology was included as an appendix to the research protocol, which was provided as part of the approval processes (see Appendix 3, Tables 39–47).

Ethics, consent and permissions

Ethics approval was received from the North West Haydock REC Yorkshire (reference 17/NW/0469) and all participants provided written informed consent. Protocol amendments are documented in Appendix 4, Table 48.

Data collection

Data were captured from participants during their inpatient stay post revision surgery and from their medical records. Data collected included independent living and carer status, working status, details of orthopaedic follow-up pathway and pathway to current revision surgery (see Report Supplementary Material 1 for the participant questionnaire). Participant data were corroborated and supplemented with data collected from medical notes, including demographics, GP and orthopaedic appointments, medication related to recent problems with joint replacement, medical history, primary joint replacement history, reasons for scheduling revision at this time and details of revision surgery (see Report Supplementary Material 2 for the case report form).

Hospital effect

The ICC was 0.081, indicating that 8% of the follow-up pathway can be explained by between-hospital differences. The likelihood ratio statistic was 101.73 (1 degree of freedom), providing strong evidence that the between-hospital variance was non-zero.

Participant characteristics

In the hip revision group, a time to revision > 10 years (OR 3.804, 95% CI 1.353 to 10.694; p = 0.011), PPF (OR 20.309, 95% CI 4.574 to 90.179; p < 0.001) and dislocation (OR 12.953, 95% CI 4.014 to 41.794; p < 0.001) were associated with UR (Table 14). In the knee revision group, there were no associations with UR.

Interview process

Professional leads and service managers were identified from site contacts and were invited via e-mail by the study project manager to take part in a telephone interview. A number of contacted individuals did not respond to the initial contact and alternative contacts were sought when possible. It was not always possible to identify other suitable candidates for interview. For those candidates who agreed, informed consent was requested along with permission to record the interview. Appointments were made to undertake the interviews, although a number of participants (n = 4) were unavailable at the time arranged for the interview. One follow-up e-mail was sent as a reminder but no other contact was attempted.

The interviewers used a semistructured topic guide (developed from the available literature by the study team) to help guide discussions to understand sites’ current follow-up care pathway and to explore any changes that had taken place. The interviews were audio-recorded and transcribed verbatim when permission was given.

The participants

The 16 included participants (n = 13 male; n = 3 female) were drawn from a range of geographical areas and types of NHS trust). They included two arthroplasty specialist nurses, one specialist physiotherapist, one manager of orthopaedic service and 12 orthopaedic consultants (all male).

Data management and analysis

The data were managed based on principles of information governance at the University of Leeds. The data from the interviews were analysed using a framework approach, allowing a structured exploration of the participants’ perspectives and a method to compare and contrast different follow-up pathways. 137 This approach was used to ensure the collection of a large amount of detailed information about the pathway in use, geographical location, the acceptability of pathways, resources (including staff), how patients accessed further care if needed, budgets and input from commissioners. The interview transcripts were used to identify key information for each service and key themes about the pathway adopted in each centre to establish some of the reasons for their implementation (Figure 10 and Box 1).

FIGURE 10.

Early thematic development: UK SAFE qualitative thematic map – health-care professionals. Bold, main and subordinate categories; arrows indicate how emergent themes associated with the main and subordinate categories.

No longer-term follow-up

Staff at a number of centres reported that they do not follow up patients beyond the immediate post-surgical period (see Appendix 5, Table 49). All of the interviewees explained their follow-up at the point of discharge from the ward:

. . . then they’re seen 8 weeks, or around 8 weeks post their surgery, in a consultant-led sort of traditional follow-up clinic. And at that point the vast majority, unless there are any specific problems, they are discharged and they’re not, they’re not routinely followed up any further than that.

Interviewee 8, large tertiary

One interviewee from a major surgical centre reported that any follow-up would be carried out at the patient’s local hospital:

I would say across the patch everybody is discharged by a year. And the great majority are now discharged between 3 and 6 months. There is no routine follow-up.

Interviewee 10, specialist centre

Interestingly, one interviewee reported a change in their hospital’s established practice, from not seeing any patients after the postoperative visit to bringing back all patients after the postoperative visit. This was not without its problems and was perhaps not sustainable in the longer term:

. . . the reason we went from none to following up everyone is we had a large cohort of hips more than knees although I think we have revised some knees as well that came very late to revision in fact we had a high periprosthetic fracture rate.

Interviewee 4, district general hospital

This change had made the clinic difficult to manage, with increased patient numbers. The follow-up patients are currently seen by the nurses in the clinic, who have been given additional training to examine the radiographs and collect PROM scores. If the nurses identify any potential problems then they bring in the consultant. Three out of the five nurses who are able to do this will retire in the near future but there has been little interest in succession planning by the managers despite a reported reduction in their PPF rates, as it is difficult to attribute this to the follow-up clinic.

Pressure for changing practice

A number of interviewees explained some of the pressures for changing their practice. For example, bringing people back into clinic when there was no obvious clinical indication caused logistic problems:

. . . oh yes! The clinic’s tiny. So it looked like Beirut during most clinic sessions that we had, probably capacity for about 20 patients sat, and then probably capacity for another 10 to stand with them. Um and obviously with four or five clinics running it was just carnage.

Interviewee 12, district general hospital

The reality is, I work in the NHS, we’re virtually bankrupt – the NHS, not this hospital – and the system is falling apart. So we’ve had to do something to work around um what is in my opinion an appalling state of affairs. But we’ve tried to do it in a manner that’s safe, and so it’s not my first choice, but I’ve been a huge advocate for this over the last 10 years, and we’ve made huge improvements. Of necessity really.

Interviewee 12, district general hospital

. . . there was some issue regarding the logistics of getting so many patients into the follow-up clinics as the numbers are limited and the trust was finding it very difficult to erm to arrange these patients into the follow-up clinics. There was a need for extra . . . extra clinics and that was another reason.

Interviewee 13, district general hospital

Caveats to the no follow-up clinic: fidelity to protocols

The interviewer explored with the participants if the described care pathway was adopted widely across their department and whether or not there were exceptions to their stated follow-up pathway. In some centres, there were no exceptions and clinical staff were discouraged from bringing patients back into clinic in a routine manner. There was some evidence that consultants who had been in post for some time were more likely to bring patients back beyond that recommended by the hospital protocol:

There are a few caveats em there are a couple of consultants that like to, for sort of young patients, uncemented, with uncemented hips, a couple of the consultants will see them again maybe at 6 months or a year to get another X-ray [radiograph] but they’re generally discharged at that point but I’d say that’s probably only . . . 5–10% of the patients at most really.

Interviewee 8, large tertiary

. . . but what we do is have a patient outcome database so that every patient who gets a joint replacement at X will get questionnaires sent to them by our outcomes department and that goes on forever. So we have really fantastic patient-reported outcome measures reported at 97% completion which is way ahead of anyone else in the NHS.

Interviewee 10, specialist centre

I think our arthroplasty team is spot on in doing the follow-ups and everything, there’s a set protocol everybody follows that and that’s very good.

Interviewee 7, district general hospital

. . . we have several knee and several hip consultants, um, the consultants on the other side of the trust, their registrars may well do the odd random patient for a 2-year follow-up. But that wouldn’t be the, that’s not the majority.

Interviewee 3, district general hospital

One participant outlined how others in their department approached follow-up that was potentially too short, but the participant acknowledged that they had understood their own skills and weaknesses as a surgeon. New consultants at this centre were also encouraged to opt out of the follow-up usually provided by the specialist nurses to better understand their surgical practice and patient outcomes:

I mean one of my colleagues started out by just discharging everyone at 6 weeks [laughs]. Why? Because he was confident in his results. And you can only do that when you’ve been around the block a couple of times as a consultant. One of the things that I didn’t say at the start, which perhaps I should of done, is that we said to the junior consultants, and we’ve got a couple who’ve just started in the last 2 or 3 years, is that we did not feel that they needed to feed their patients into this system. They needed to see how their patients were behaving at 6 weeks, in terms of their recovery and rehabilitation needs, so that they understood what their practice was about.

Interviewee 12, district general hospital

The protocols were often fairly well controlled within the no long-term follow-up group. Those colleagues who arrange to see patients for no obvious clinical reason could affect colleagues who may wish to see a patient back in clinic for a specific reason:

So the commissioners certainly do look at our new-to-follow-up ratio. They look at our total number of appointments per pathway. I think some of my colleagues see people a lot and that makes it harder for those who don’t, like myself which means that when I do want to see someone I’m, I’m feeling guilty about it I suppose.

Interviewee 15

Follow-up beyond 1 year

When an interviewee was asked about the number of patients from their follow-up clinics with problems indicating intervention and potential revision, the response was slightly surprising in that the interviewee observed that they had seen an increase in the number of patients coming in acutely with a PPF:

Not really. Um, one observation we have made is that we seem to be doing fewer elective revisions, but we’re certainly doing much, much more periprosthetic revisions now.

Interviewee 5, tertiary centre

A partial explanation for the increase in the number of patients (i.e. one patient per week) needing an operation was the hospital’s geographical location. Situated on a border between counties, the hospital received referrals from many smaller periphery hospitals.

Patients were discharged after 1 year’s follow-up if they were aged ≥ 80 years. However, at the time of the interview, of the patients currently on the ward, there was a patient with a PPF and aged 82 years. The interviewee was concerned about patients like this who are not followed up but can still present with very serious complications:

[Catchment] borders on a different health-care trust, where we often pick up patients that haven’t been followed up for many, many years. It’s always at the periphery of, we’ve got hospitals at the periphery of the county, it’s always at the peripheries that you pick up these individuals with weird and wonderful implants in that haven’t been followed up for 10, 15 years.

Interviewee 5, tertiary centre

Rationale for the follow-up pathway

How patients get back into the system

There was reliance on patients identifying when they had issues with their implant. Most of the interviewees understood that patients would make their way back into the system via their GP. Some interviewees had explained that patients sometimes contacted the consultant via their secretary or made telephone calls to the ward where they had their original surgery.

However, there was some lingering concern that patients with problems might not seek the appropriate advice:

I think . . . this is all sort of anecdotal. We are, we are obviously concerned that there are patients with sort of silently failing hips out in the community that are going to present with, you know, catastrophic failure and periprosthetic factures that we could have perhaps done something about had it been, had the loosening been picked up earlier. But anecdotally we haven’t seen people coming through the system presenting in that way. Generally speaking, if they have any problems then they still present . . . relatively early.

Interviewee 8, large tertiary

I’ve never seem an asymptomatic failure, and you have to caveat that by saying it starts out asymptomatic and then becomes symptomatic, so I’ve never seen anyone who’s come along with a loose hip who didn’t have some symptoms beforehand . . . of those people that come back into the system what is it that we could have spotted beforehand?

Interviewee 6, district general hospital

I guess the issue is reminding people in later life about that conversation we had when we discharged them at 1 year. Which is, if your symptoms come back, take it seriously, have an X-ray [radiography], get your orthopaedic surgeon to look at it, or your GP to review the report. And I think most people forget that. And maybe it’s time for a very different approach to this.

Interviewee 12, district general hospital

Other interviewees were confident that patients would find their way back into the system:

Via their GP. We import a lot of patients from upcountry, because we’re a retirement area . . . So if there’s a problem they’ll come through their GP. So the answer is they nearly always come through their GP.

Interviewee 12, district general hospital

Patient experience and adding value to the consultation

There were many negative expressions about the patient experience of attending an outpatient appointment for follow-up. Some interviewees were of the opinion that this added little value to the patient or their outcomes. However, one interviewee, when describing their early care pathway, noted that some patients with perhaps long-term experience of attending orthopaedic clinics expected to be seen on several occasions postoperatively, and that these patients put pressure on the registrars to make repeat appointments:

If they are good at 6 months, again they will be discharged but there are some patients who come back and back because they in effect bully the registrars erm to be seen again.

Interviewee 16, teaching hospital trust

Some interviewees expressed the opinion that patients, particularly older patients, decide that, even if a problem with loosening has been detected, they might not necessarily want to proceed to revision surgery when asymptomatic. This was reflected across a number of interviewees and this comment is typical of this view:

. . . another thing you need to note is that even if you are following them up and you find that there is some kind of loosening of the prosthesis and they’re asymptomatic they’re very, very reluctant to come and have a procedure. So even if you catch them early you’re not guaranteed that they would proceed with a big operation.

Interviewee 13, teaching hospital trust

A number of interviewees were very interested in moving towards a virtual clinic or other specialist practitioners, which would alleviate some of the problems that patients experience when attending clinics:

. . . well a virtual clinic what I would do it would probably be a nurse-led virtual clinic but obviously they need a review of their X-rays [radiographs] that would be a service that I would definitely quite like to do I think if you sit in an orthopaedic clinic you see that patients sit around waiting for ages err and most of the time they are perfectly fine and really they could be assessed by way of a [tele]phone call and an X-ray [radiography].

Interviewee 16

We have trained nurses who are waiting in the wings, they come and see the patients in the clinic but what we want to do, from us it goes to them, they do more detailed survey, more detailed looks, do all the checks, all the measuring scores like we have Oxford Knee Score, Harris score and all scoring but that would also the functional outcome of the patients which we don’t usually check when – in the follow-up clinic it’s just whizz and go I mean they come ‘are you fine?’ ‘X-ray [radiograph] is fine no concerns, OK off you go’

Interviewee 7, district general hospital

So they do have sort of a semistructured approach to this 6-week review. It’s not just going along and saying, ‘how are you’. So that was me really tryna stop rubbish outpatient appointments, and to improve the quality or the consultation . . . our nurses I know, seem to be happier sort of following this protocolised approach to discussing things with the patients. Whereas registrars tend to do things randomly, in my opinion . . . It’s very difficult, it’s like herding cats with registrars.

Interviewee 12, district general hospital

Commissioners, funding and future plans

Almost all the interviewees had some form of experience of preparing business plans for managers or having discussions with commissioners when they have wanted to initiate changes or to be innovative in their practice. These relationships often had mixed results, with some being supported and others not:

. . . there were some discussions at higher levels with commissioners. What has been decided early on was that the trust would follow them at 12 months or individual case needed and there is a plan to put in practice a virtual clinic at the end of 5 years . . .

Interviewee 13, teaching hospital trust

. . . we’ve had discussions with the commissioners because we need to fund the virtual follow-up. Because obviously if they’re not coming into clinic and getting a clinic tag then the £150 follow-up fee or whatever it is and then we’re then taking that on our chin by having the surgical care practitioner do it from their office, then we the hospital are going to be subsidising the commissioners. So discussions have been had.

Interviewee 12, district general hospital

With the primary care commissioners and we had to . . . because the X-ray [radiography] departments were actually managed by a different trust for the satellite services – so for the community hospitals they were run by North Devon NHS Trust, so we had to have a, um, we actually had to have an agreement with them to actually be able to do this service. So it took years, literally, to actually get everything . . . so from that first conversation in 2009 it probably took 5 years to get the thing properly off the ground.

Interviewee 14, tertiary centre

Evidence for change

The question of evidence to support change has arisen throughout the interview and analysis process. The interviewees expressed mixed views about the state of the current evidence base for continuing with long-term follow-up, moving to a virtual system or withdrawing all follow-ups after the immediate postoperative period:

I mean I did have to point out that when the patients came back at 1 year, I think one particular year when we [still] saw the patients at their 1 year, you know there was no one that needed reviewing and everyone was just, every single patient had been fine and discharged. Every single one of those appointments had just been so the consultant could know how the patient was doing, and not for the patient’s benefit at all.

Interviewee 3, district general hospital

So certainly we want to standardise it, we need to know what we’re doing is appropriate. It’s quite expensive what we’re doing, and I think we’re probably . . . there’s lots who don’t follow-up like we do, but we think it’s important until proven otherwise. But I don’t think we will be proven otherwise.

Interviewee 5, tertiary centre

I think there is a paucity of evidence at the moment. But our concern was that we’d be following up a huge number of patients at . . . and that very small group where they’d potentially have a violently failing hip that you know, but we’re not entirely clear that seeing them regularly would have necessarily changed their . . . final outcome.

Interviewee 8, large tertiary

. . . have an X-ray [radiography] after 10 years and if you look at the National Joint Registry survival rates now we are talking about 20–25 years and 95–97% survival. So we are talking about an extreme minority of patients at 2 or 3% or maybe 5% at the most who will have some sort of problems at 10 years or even earlier.

Interviewee 13, district general hospital

The final consensus meeting

The results of the study were presented at the final consensus meeting to discuss the results with a group of invited experts. At this time, there was a feeling expressed by attendees that, contrary to our findings within the qualitative interviews, a significant number of centres were not following up patients beyond 1 year.

We were asked to look at how participants in the main study (see Chapter 5) had entered the system for revision (i.e. PR or UR) at each of the sites involved in the qualitative interviews (Table 16). There were incomplete data on some of the sites participating in the interview study that may not have recruited any analysable participants into the main study.

Modelling method

To answer our research question, we planned to develop a health economic decision-analytic model, representing the natural course of knee and hip replacement. Different options are available for developing decision-analytic models to perform economic evaluations. Many previous decision-analytic modelling studies have used cohort Markov models to perform economic evaluations for joint replacements. 142–146 Markov models are appropriate for handling disease progression of chronic conditions in which the decision problem can be represented in terms of health states and they are flexible in handling a longer time horizon with multiple health states. 147 The specific process to choose the most appropriate modelling method is described in Appendix 5.

Model structure

The structure of the cohort Markov model used in this study was discussed and agreed between clinicians and researchers, and it is presented in Figure 11. The cycle length employed in the model was 1 year, which was assumed appropriate to capture the transitions between the health states included in the model. A lifetime horizon was employed, in which the model runs until all patients are in ‘death’ state or reach 100 years of age. Crucially, the model’s starting point is at 6 years after primary joint replacement, which was chosen because the analysis was focused on the impact of long-term follow-up, as discussed above. 140,141

FIGURE 11.

Cohort Markov model for the health-care costs and QALYs associated with long-term follow-up and no follow-up of patients after 6 years following their primary hip or knee replacement. Separate models with the same structure were developed for knee and hip replacements.

In the model, patients started at the ‘no revision after primary joint replacement’ state and could stay there or move to ‘revision after primary joint replacement’. Patients could only remain at the ‘revision after primary joint replacement’ state for a single cycle (i.e. 1 year) and then they move to either ‘no re-revision after revision’ or to ‘re-revision after revision’. From ‘no re-revision after revision’, patients could either remain in that state or move to ‘re-revision after revision’. Finally, patients could only stay at ‘re-revision after revision’ for a single cycle and then they had to move to ‘no further revision’, where they would stay until the end of the simulated time. Therefore, the model was limited to two revision surgeries, which was deemed acceptable given that very few patients undergo more than a primary and two revisions on a knee or a hip. Patients from any state could move to ‘death’, which was an ‘absorbing’ state (see Figure 11).

Although the structure was the same (as agreed by the consulted expert clinicians), models were populated and analysed separately for knee and hip replacements.

Study groups

To answer our research question regarding the impact of long-term follow-up on health-care costs and QALYs, we populated and compared model results between having follow-up and not having it. To do this, we created two groups of patients: (1) patients with at least one long-term follow-up visit to the outpatient trauma and orthopaedics service (i.e. the ‘follow-up group’) and (2) patients without any follow-up visit (i.e. the ‘no follow-up group’). Only attended visits accounted as follow-up and, therefore, missed appointments were not considered follow-up.

We set a group of rules to assign patients in the follow-up group. First, the patient must have at least one HES Outpatient record with the code ‘110 = Trauma and Orthopaedics’, describing the specialised service within which the patient was treated, at least 5 years after primary surgery. 149 Second, when follow-up visits occurred within the waiting period (i.e. the time between the date of the decision to have revision surgery and the date of revision surgery), then they did not count as follow-up, as it was assumed that those visits were a consequence of the decision to have the operation (e.g. prepare for surgery). Finally, when patients had only a single follow-up visit within 6 months before the date of the decision to have revision surgery, then the visit did not count as follow-up because it was considered that a single visit is highly likely to be for the purpose of conducting tests to inform the decision for revision surgery (e.g. blood tests, magnetic resonance imaging, check-ups for cardiovascular problems) instead of routine long-term follow-up. If patients were not included in the follow-up group, then they were assigned to the no follow-up group.

As the follow-up visits in the orthopaedic department can be identified not only through HES Outpatient but also through the recorded referrals in the CPRD database, we explored the most appropriate choice of data set for our study. To do this, we applied the criteria, as described in the previous paragraph, to identify CPRD referrals to hospital outpatient using the corresponding ‘medcodes’ and compared our findings to the group selection via HES APC. For knee replacements, 21% of the entire knee cohort had at least one visit identified in HES Outpatient, but no corresponding referrals identified in CPRD. In addition, 23% of the entire knee cohort had at least one visit identified in both HES Outpatient and CPRD. Finally, only 2% of the entire knee cohort had at least one visit in CPRD and no visits identified in HES Outpatient. Similarly, 21% of the entire hip cohort had at least one visit identified in HES Outpatient but no such referrals in CPRD. Twenty-five per cent of the entire hip cohort had at least one visit identified in both HES Outpatient and CPRD. Finally, again, only 2% of the entire hip cohort had at least one visit in CPRD and no visits identified in HES Outpatient. Overall, HES Outpatient appeared to capture more outpatient hospital visits and to report more information about them than referral records in CPRD and, therefore, we used the former to operationalise the follow-up and no follow-up groups for our study.

Age-related subgroups

As previous studies150 have shown that the outcomes of primary joint replacement, in terms of both health-care costs and health effects, may depend on the age of patients, we separated the cohort of patients into two age groups. One age group included patients aged < 70 years at their primary operation and the other age group included patients aged ≥ 70 years at their primary operation. Therefore, our analyses involved four different patients groups: (1) follow-up group with patients aged < 70 years, (2) follow-up group with patients aged ≥ 70 years, (3) no follow-up group with patients aged < 70 years and (4) no follow-up group with patients aged ≥ 70 years. Each patient group was characterised by their lifetime costs and QALYs.

Other included variables

We describe here additional variables used in different stages of our analyses. To identify the age at time of surgery, we used the date of a procedure recorded in HES APC combined with the patient’s year of birth from CPRD. The patients’ sex and ethnicity (i.e. white or other) were recorded in CPRD. We calculated the CCI score using ICD-10 diagnosis codes in HES APC. CCI score is a measure of comorbidity that can take the values 0, 1, 2 or 3 +. 151

Index date

As patients could have a follow-up visit any time between 5 years after primary operation and the end of exposure time (i.e. 10 years since primary surgery), a time-dependent covariate approach was taken when examining the risk of revision. This is the recommended approach to avoid introducing ‘immortal time bias’, which refers to a cohort exposure period during which death (or other outcomes that indicate the end of exposure time) cannot occur. 152 Using this approach, all patients were included in the no follow-up group until their first long-term follow-up visit occurred. At that point, their membership of the no follow-up group ended and that of the follow-up group started. Patients without any follow-up visit remained, therefore, in the no follow-up group. The index date was 6 years after primary surgery for the no follow-up group and for the follow-up group it was the date of the first follow-up visit (which could only be at least 6 years after the primary replacement).

Transition probabilities for knee replacement

Costs for knee replacement

Health-care costs for the patients included in the study were derived from three sources: (1) costs for revision surgery were estimated from HES APC records, (2) costs for outpatient follow-up visits from HES Outpatient and (3) primary care costs from CPRD.

Quality-adjusted life-years for knee replacement

Preoperative and postoperative EQ-5D-3L questionnaire responses were available in the HES PROMs database and the responses were used to estimate health utility scores by applying the UK value set tariff, which incorporates the preferences of the general population. 157 The scores are a preference-based measure of HRQoL, ranging between ‘–0.59’ (worst state) and ‘1’ (perfect health), with death being anchored at ‘0’. QALYs were estimated using the area under the curve method.

We linked records in the PROMs database to records for a primary knee replacement for the corresponding patient in HES APC if the date of preoperative EQ-5D completion was between 1 year prior and 31 days after the date of the primary surgery. When more than one PROM record was found to satisfy this condition, we chose the record with a completion date closest to the date of surgery. Age- and group-specific mean postoperative utility scores were used as estimates of yearly QALYs for patients in the ‘no revision after primary knee replacement’ state.

The same procedure was followed to estimate health utility scores for patients having a revision. The mean revision preoperative utility score was used as an estimate for QALYs associated with the single year patients would spend on the ‘revision after primary knee replacement’ and ‘re-revision after first revision’ states. As only a few patients had a revision, we were able to estimate age- but not group-specific health utility scores and, therefore, scores at revision were the same for both the follow-up and the no follow-up groups. The mean revision postoperative health utility estimate was used as the yearly QALYs for patients in the ‘no revision after revision’ and ‘no further revision’ states.

Based on evidence from the literature,158 we assumed that the HRQoL of patients was slightly reduced over time because of health problems not necessarily related to the health condition under study (i.e. knee replacement). For this reason, we added an age-specific decrement in QALYs gained over time, which was estimated separately for the two age groups. For patients in the < 70 years age group, a yearly decrement of ‘0.003’ per year was used. For patients in the ≥ 70 years age group, a decrement of ‘0.007’ was used. 158

Modelling assumptions

As with most decision-analytic models, a number of assumptions were made. First, the structure of the model represented a simplified version of the health-care pathway of patients with knee replacement. Based on the chosen structure, patients could have up to two revisions. It is possible that, through their lifetime, some patients may need more than two revisions and some indeed do. However, we expect that the proportion of patients with knee replacement that would require more than two revisions is very low and the need of three or more revisions is unlikely to be influenced by having or not having follow-up visits.

Second, as there was no information in HES Outpatient regarding the content of each visit or whether or not visits were related to the replaced knee, to increase our confidence that the visits were indeed a follow-up for the replaced knee we set a number of rules to define follow-up visits (see Study groups). Therefore, we conducted our analysis assuming that each selected visit was very likely related to the knee replacement.

Finally, based on the classification of patients in the follow-up and no follow-up groups that we performed, we assumed that patients in the no follow-up group would not have a follow-up visit after 10 years since their primary knee surgery. However, it is possible that some patients who do not have any follow-up for the first 10 years following their primary will have follow-up later on. We did not have enough data to explore this assumption. Some additional parameter assumptions are discussed in the Appendix 6.

Model inputs for hip replacement

The modelling of health-care costs and QALYs associated with follow-up and no follow-up of patients with primary hip replacement followed the same methodology as that applied to the modelling of patients with knee replacement. The specific differences between the data feeding the hip model compared with the knee model are detailed below.

The different AIC of the fitted survival models can be found in Appendix 6, Table 51. The incidence of death at 4 years from index date was 14% for patients in the follow-up group and 21% for those in the no follow-up group, but this difference was not statistically significant based on the adjusted regression model (HR 0.91, 95% CI 0.81 to 1.02).

Model validation

A number of different validation steps were taken to ensure that there were no errors in the developed models. First, we checked if the proportion of patients moving into the ‘death’ state was the same for follow-up and no follow-up, as we used the same transition probabilities. Second, we assigned the same transition probabilities to all groups to examine whether or not all health states have equal number of patients after running the model. Finally, we used extreme values for some parameters, such as ‘0’ and ‘1’ for transition probabilities and health utilities estimates, and ‘0’ or ‘100,000’ for costs in each health state.

Revision rates

To compare the number of revisions between the follow-up and no follow-up groups, the cumulative incidence was estimated, which provided the probability of experiencing a revision within a given period and before the occurrence of the competing risk of death. Revision rates were higher for the follow-up group than for the no follow-up group in both age groups. For patients in the follow-up group with patients aged < 70 years the revision rate was 5.2%, whereas for patients in the follow-up group with patients aged ≥ 70 years it was 2.0%. For patients in the no follow-up group with patients aged < 70 years the revision rate was 1.0%, whereas for patients in the no follow-up group with patients aged ≥ 70 years the rate was 0.4% (Figure 12).

FIGURE 12.

Cumulative incidence of revision following knee replacement. (a) Patients aged < 70 years; and (b) patients aged ≥ 70 years, stratified by follow-up and no follow-up groups, accounting for time-varying exposure. The index date was 6 years from primary surgery for the no follow-up group and the date of first follow-up visit in the orthopaedics department was at least 6 years after primary surgery for the follow-up group.

Utility scores

The mean preoperative and postoperative health utility scores for patients with primary knee replacement are shown in Figure 13. All four cohorts (i.e. by follow-up status and age) reported significant improvements in health utility from pre to post primary surgery, generally going from between 0.34–0.45 pre operatively and 0.70–0.77 post operatively. For revision surgery, the mean preoperative and postoperative health utility scores increased from 0.42 to 0.60 for patients aged < 70 years and from 0.32 to 0.64 for patients aged ≥ 70 years (see Appendix 6, Figure 22).

FIGURE 13.

Mean preoperative and postoperative utility score of patients with primary knee replacement separately for each subgroup.

Costs

The unit costs of revision surgery for each subgroup varied slightly by group. Being in the follow-up group was associated with slightly lower revision costs for patients aged < 70 years (follow-up group £8779 vs. no follow-up group £9568), but slightly higher costs for patients aged ≥ 70 years (follow-up group £10,279 vs. no follow-up group £9076).

For the follow-up group, mean outpatient costs over 4 years since index date are shown in Figure 13 for the two age groups. Regardless of age, outpatient costs were significantly higher (£332–75 per year) in the first year of the follow-up period of interest (i.e. year 6 after the primary surgery) than in the 3 years that followed, during which they remained largely stable (between £61 and £98 in the fourth year). Costs were slightly, but consistently, higher for the patient group under 70 years of age (Figure 14).

FIGURE 14.

Age-specific mean outpatient costs for patients with knee replacement in the follow-up group for 4 years after index date.

The mean primary care consultation costs over 4 years since index date and for unrevised patients are shown in Figure 15. Costs slowly, but consistently, decreased over the 4 years for all groups. Being in the older age group and having follow-up, however, increased consultations costs. However, for the no follow-up group, mean costs for patients aged < 70 years decreased from £275 to £243. Likewise, mean costs for patients aged ≥ 70 years decreased from £452 to £345.

FIGURE 15.

Age- and group-specific mean costs for primary care consultations for patients without revisions following knee replacement for 4 years after index date.

The mean costs after revision followed a similar pattern and were slightly higher than for those unrevised. For years 2 through 4, mean consultation costs were £295 for the no follow-up group with patients aged < 70 years and £360 for the follow-up group with patients aged ≥ 70 years.

Economic modelling

The age- and group-specific lifetime costs and QALYs from the deterministic analysis are reported in Table 18. Follow-up was associated with both higher lifetime costs and lower lifetime QALYs than no follow-up for both age groups, with the differences being larger for the younger age group. The mean lifetime costs and QALYs derived from the probabilistic sensitivity analysis for the four groups are reported in Table 18, along with corresponding CIs, and these were very similar to the results of the deterministic analysis.

We explored the sources of cost differences between the groups (see Appendix 6, Figure 23). For the age group < 70 years, 42% of the cost differences between the follow-up and no follow-up groups were due to outpatient visits, 35% were due to revision surgery costs and 23% were due to primary care consultations. For the age group ≥ 70 years, 40% of the cost differences were due to outpatient visits, 22% were due to revision surgery and 38% were due to primary care consultations.

Revision rates

Revision rates were higher for the follow-up group than for the no follow-up group for both age groups. As shown in Figure 16, the revision rate was 4.2% for patients aged < 70 years in the follow-up group and 1.9% for those aged ≥ 70 years in the follow-up group. For patients in the no follow-up group, rates were 1.8% for the younger age group and 1.1% for those aged ≥ 70 years (see also Appendix 6, Figure 24).

FIGURE 16.

Cumulative incidence of revision following hip replacement. (a) Patients aged ≤ 70 years; and (b) patients aged ≥ 70 years, stratified by follow-up and no follow-up groups, accounting for time-varying exposure. The index date was 6 years from primary surgery for the no follow-up group and the date of first follow-up visit in the orthopaedics department was at least 6 years after primary surgery for the follow-up group.

Utility scores

The mean preoperative and postoperative health utility scores for patients with primary hip replacement but no revision are shown in Figure 17. As with the knee replacements, all groups reported a significant improvement, with the range of scores increasing from 0.31–0.42 pre operatively to 0.79–0.86 post operatively. The mean scores for patients with revision are reported in Appendix 6, Figure 25, with improvement in mean scores also observed for all groups, although not in the same magnitude as with primaries.

FIGURE 17.

Mean preoperative and postoperative utility score of patients with primary hip replacement separately for each subgroup.

Costs

As with knee replacement, mean unit costs for revision surgery following hip replacement varied only slightly by each subgroup. Costs were lower for patients in the follow-up group than for patients in the no follow-up group regardless of age (patients aged < 70 years: follow-up group £9162 vs. no follow-up group £10,685; patients aged ≥ 70 years: follow-up group £9890 vs. no follow-up group £10,301).

The mean outpatient costs for hip replacement patients over 4 years since index date are shown in Figure 18. These costs apply only to the follow-up group because the no follow-up group did not have any follow-up visits. As with knee replacement patients, there was a significant drop in outpatient costs during the first year of the long-term follow-up period, compared with the years that follow, reducing from approximately £300 in the first year to approximately £100 thereafter. The mean costs were also slightly, but consistently, higher for patients aged < 70 years than for patients aged ≥ 70 years.

FIGURE 18.

Age-specific mean outpatient costs for patients with hip replacement in the follow-up group for 4 years after index date.

Primary care consultation costs for those unrevised after a primary hip replacement followed the same pattern as knee replacement patients. The mean costs decreased over time, but were higher for the older age group and for those patients having follow-up. Details are shown in Figure 19. The drop after the first year was also reported by revised patients, although age and follow-up status did not appear to be the main drivers, as yearly costs for the second and subsequent years after revision were lowest for the follow-up group with patients aged ≥ 70 years (£177) and highest for the no follow-up group with patients aged ≥ 70 years (£307).

FIGURE 19.

Age- and group-specific mean costs for primary care consultations for patients without revisions following hip replacement for 4 years after the index date.

Economic modelling

The lifetime age- and group-specific costs and QALYs from the deterministic analysis are reported in Table 19. As with knee replacements, follow-up was associated with both higher lifetime costs and lower lifetime QALYs than no follow-up for both age groups, with the differences being larger for those aged < 70 years. The mean lifetime costs and QALYs derived from the probabilistic sensitivity analysis for the four groups are reported in Table 19, along with corresponding CIs.

Sources of costs differences between the groups were also similar to those reported by knee replacement patients (i.e. costs differences were mostly due to the cost of outpatient visits; see Appendix 6, Figure 26). Costs for revision were also important contributors to the total cost differences, and this can be explained by the higher revision rates for the follow-up group, which are reflected in the annual incidence of revision plotted in Appendix 6, Figure 27.

Revision

We found that revision rates were higher among patients in the follow-up group than among those in the no follow-up group for both knee and hip replacements. The higher rates of revision for patients in the follow-up group may indicate that orthopaedic follow-up is an effective surveillance tool that helps identify patients in need of revision. Timely identification of deteriorating implants may prevent substantial bone loss and increased revision costs. 24,160 We found that mean unit costs of revision were higher for patients in the no follow-up group in both age groups for hip replacement and in the younger age group for knee replacement, although differences were small and the proportion of patients having revisions was low.

It is possible that patients in the no follow-up group did not attend follow-up visits because they were less likely to require a revision or they might have not had any complaints, such as pain or discomfort, about the replaced joint. A previous study161 indicated that when primary and secondary reasons for not attending follow-up visits were combined, the most common reason given by patients was that they did not have any problems with the replaced joint. Therefore, it is possible that patients who had follow-up visits were different from patients having no follow-up visits in terms of health-care needs and clinical conditions.

However, we found that a few patients in the no follow-up group did have a revision, even though those events were rare. Patient-reported outcomes, such as pain, function and general satisfaction, are useful indicators of problems related to the replaced joint. Nonetheless, radiography, radiographic reviews and image interpretation by the orthopaedic team during a follow-up visit are essential for the assessment of the joint replacement because they can identify degenerative changes that cannot be determined by PROMs alone. 162,163

Finally, previous studies have raised the question of overtreating patients with joint replacements. 164–166 In the context of this study, overtreatment could influence the increased revision rates for the follow-up group that we observed. However, evidence of overtreatment, or undertreatment, of patients with knee or hip replacements in the existing literature164–166 is inconclusive. Future studies should examine the possibility of overtreatment of patients during long-term follow-up, how this could influence the patients’ well-being and the costs incurred for the health-care systems.

Follow-up outpatient visits

We found that less than half of the patients had a long-term follow-up visit, despite clinical guidelines140,141 recommending that follow-up to be offered to all patients with joint replacement. Approximately half of the patients who had follow-up had their first visit between 5 and 6 years since the primary surgery, with a median of three visits between 5 and 10 years since primary surgery. Although clinical guidelines’ recommendations are not always followed, these findings are not surprising when considering the variation in providing follow-up in clinical practice. It has been reported that only 43% of hospital units performing hip replacements offer follow-up after 5 years since primary surgery in the UK. 47 Overall, it remains unclear how the different hospital units decide on their strategy regarding the duration of follow-up, and the number and type of follow-up appointments, that they offer to patients. It is possible that some patients might have been followed up privately; however, we had no insight into this, as HES data include only NHS-funded admissions, which is consistent with the perspective of our analysis.

As expected, follow-up outpatient visits are related to a substantial cost for the health-care system. The cost of outpatient visits was the largest contributor of the cost differences between the follow-up and no follow-up groups in our analysis, although an assumption was made that the cost of outpatient follow-up visits would continue over time at the same level as that of the last year for which we had data. The mean outpatient cost per patient decreased over time in our analysis; however, considering the actual number of patients who are eligible for follow-up after joint replacement, the implications at a population level are significant.

Primary care costs

Primary care costs were higher for patients who were followed up than for patients who were not. Similar to costs from outpatient visits, the mean yearly primary care costs per patient decreased over time, but they remained substantial. Moreover, primary care costs following a revision were higher than the respective costs for patients without revision, and this may be a signal of patients’ worse physical state around the time during which they had a revision. Primary care costs were higher for the older age group (i.e. patients aged ≥ 70 years) for both knee and hip replacement. A recent study167 carried out in the UK has shown that, over the study’s 5-year follow-up window, primary care consultations and GP visits increased for older individuals aged between 85 and 90 years. In addition, older individuals were more likely to consult their GP than other health-care team members. Therefore, in our study, the differences we found in primary care costs between the two age groups may reflect the complex health-care needs of older patients in general.

Quality-adjusted life-years

The follow-up group accumulated fewer lifetime QALYs than the no follow-up group. This is mainly explained by the higher revision rates experienced by the former, as revisions are associated with lower health utility scores remaining unrevised. It is worth noting that both primary joint replacement and revision surgery had a very positive impact on HRQoL for all patient groups, as has been reported in previous studies. 12,13

These recommendations apply to post primary hip and knee replacement follow-up

There was some general discussion about whether or not recommendations should be separated for hip and knee, or whether or not a single set of recommendations should be agreed to cover both hip and knee replacement follow-up together. The consensus was that the evidence supported a single set of recommendations to cover both hip and knee replacement. It was also highlighted that it must be emphasised that these recommendations were for follow-up after primary surgery and that these recommendations did not apply to follow-up after revision surgery.

There was agreement that any recommendations should provide scope for better ways of providing follow-up to be developed and tested, that face-to-face follow-up provision was not always necessary and that innovations, such as virtual clinics and remote monitoring, could be incorporated into follow-up services.

The importance of educating patients to reduce the risk of the introduction of a rapid access service leading to additional/unnecessary costs through inappropriate self-referral was highlighted. Education of both primary and secondary care clinicians was also emphasised.

There was some general discussion about how disinvestment in follow-up may affect disadvantaged groups (e.g. those hard to reach and those of low socioeconomic status). It was also highlighted that the current evidence base misses those patients who are symptomatic but do not have appropriate follow-up and, therefore, do not receive revision surgery despite needing it. It was agreed that further work was needed to understand how to reach such groups, and to explore the needs and outcomes in this population.

The 10-year time point in these recommendations is based on a lack of robust evidence beyond 10 years

There was agreement that the lack of available data beyond 10 years of follow-up within the UK databases utilised to inform the evidence base should be noted, and that a recommendation to disinvest in follow-up beyond 10 years could not be supported.

In these recommendations, the term complex cases refers to individual patient and surgical factors that may increase the risk for replacement failure

In addition to discussion around prosthesis rating (see Recommendations) there was agreement that additional factors must be considered when determining whether or not a patient required additional follow-up provision. Age should be relevant in review, with younger patients more likely to have a failing implant and older patients more likely to outlive their prosthesis. Surgical experience may be important. For junior surgeons, follow-up may provide some additional benefit with respect to their own training and development. Additional surgical factors and patient demographics may also increase the risk for replacement failure, and these factors should be considered prior to disinvestment in follow-up for an individual patient.

For ODEP-10A* minimum implants, it is safe to disinvest in routine follow-up from 1 to 10 years post non-complex hip and knee replacement provided that there is rapid access to orthopaedic review

Based on the evidence from UK SAFE and from clinical experience, there was general agreement among surgeons that they would be happy to discharge a routine patient with an ODEP-10A* prosthesis after the 6-week postoperative check and not see the patient again until 10 years. It was agreed that we could not currently state that follow-up ‘is not needed’, but that there was sufficient evidence to state that it was ‘safe to disinvest’. However, there was much emphasis on the need for a rapid access service to orthopaedics to ensure that patients could access support if the need arose. Surgeons also highlighted the importance of ensuring that these recommendations did not enable NHS trusts to completely disinvest from all follow-ups, with no safety net for patients. Abandoning all follow-up facilities for these patients should not be part of the recommendation from this study. Rather than complete disinvestment in all follow-ups up to 10 years, the UK SAFE guidelines should require the provision of different follow-up facilities for these patients. It is likely that the new facilities will be cheaper to provide than the current ones. No changes in follow-up arrangements should be made unless a pathway is available for urgent review, ideally straight into secondary care, for patients with hip or knee replacements who develop new symptoms in their joint. Patients agreed that the key to ensuring that disinvesting in follow-up was safe was to ensure that all patients had access to a robust, simple and safe mechanism for reaccessing orthopaedic support. GPs highlighted that referral into a rapid access system could be initiated by a GP, but that patient-initiated self-referral could also be considered. Further discussion of this can be found in Chapter 9.

The difficulty of setting up an efficient and cost-effective rapid access clinic was highlighted, as this can be difficult to plan to ensure rapid availability to appointments without the risk of leaving empty clinic slots. This work does not support disinvestment in the follow-up service without such a rapid access service available for, and direct access by, the patients. The need for further work to understand how these clinics would work was emphasised.

For ODEP-10A* minimum implants in complex cases, or non-ODEP-10A* minimum implants, periodic follow-up post hip and knee replacement may be required from 1 to 10 years

Surgeons and GPs highlighted that any recommendations must qualify that ‘safe disinvestment of follow-up’ applies to only those prostheses that are endorsed by the existing recommendations, such as those from ODEP, NJR and MHRA. Any implants that do not meet these standards may require additional follow-up and this must be stipulated and considered on a prosthesis-specific basis. One industry representative highlighted some concern that this was more complex for knee than for hip because of ODEP ratings being based on multiple construct factors, which could lead to difficult in classifying patients for follow-up/no follow-up. The use of current NJR data would be essential when identifying combinations of prostheses that require additional follow-up. We recommend that the lowest ODEP rating of all the components of the joint should determine the overall ODEP rating of that joint.

In addition, it was emphasised that, in some cases, it is the need of the patient that should drive follow-up and not just the prosthesis. Likewise, for complex cases or patients with complex needs, more regular follow-up must also be considered.

At 10 years post hip and knee replacement, clinical and radiographic evaluation is recommended

Following on from discussion regarding the lack of current data to support disinvestment beyond 10 years, stakeholders agreed that all patients should be given the opportunity to re-present for review of their joint replacement. There was emphasis from the majority of surgeons that this review must include both clinical and radiological review, as issues such as silent osteolysis, which become more common after 10 years, may be missed by clinical-/patient-reported review alone. There was support for the potential use of virtual clinics for such review, provided that clinical and radiological review were incorporated.

After 10 years post hip and knee replacement, frequency of further follow-up should be based on the 10-year assessment (and ongoing rapid access to orthopaedic review is still required)

There was general agreement that follow-up beyond 10 years should be based on the clinical and radiological review at the 10-year time point, but that continued rapid access to orthopaedic review if necessary should be re-emphasised.

Suggestions as to how such a new follow-up service might work

• Patients should be empowered to share or take control of their own follow-up after hip or knee replacement.

• Patients should be provided with written details of their implant and its ODEP rating. Only ratings of 10A* and above are suitable for the new model of follow-up.

• Patients should be asked if they are willing to provide consent to their data being collected centrally on a national database.

• Patients should be provided with written instructions as to the timing of their next review with radiography. A login and personal password could be provided to a local or national online follow-up joint replacement pathway website.

• GPs should be provided with details of the model of follow-up, when the next radiography or follow-up is due to take place and how to access the rapid access system if required.

• When possible, the patient should have self-referral access to a local virtual clinic, accepting that this may or may not be the secondary care centre where the primary surgery was carried out. Strict screening and triage criteria will need to be in place for this.

• This self-referral may be through an online portal or directly with their local provider.

• Secondary care should develop an approved and accredited radiography follow-up service, which may be virtual, for GP referral or patient self-referral should a patient develop pain in, or problems with, one of their replaced joints. The approved radiography service should have a special interest in joint replacement review with a lead radiologist who has a special interest in joint replacement follow-up.

• If a patient finds themselves in an area without a UK SAFE pathway in place when they develop pain or other problems with their joint replacement, then urgent referral to a secondary hip or knee replacement service should be made. Urgent radiography of the joint should be arranged if an appropriate follow-up appointment is not available or delayed, or if there is concern regarding impending fracture around the implant. The radiograph should be urgently reviewed by the local pathway. If there are impending urgent problems on the radiograph then a local urgent review should be carried out. Depending on the local service, this patient may then be treated locally or referred to the tertiary hub for revision surgery in that region. Patients with systemic symptoms should be referred urgently, without starting any antibiotics.

If the above points are in place, then it is safe to disinvest in routine follow-up before 10 years after hip and knee replacement surgery.

With the planned setting up of UK regional joint replacement revision services (hub and spokes model), the UK SAFE pathway will become an essential part of this new revision service.

Follow-up of hip and knee replacement patients may be virtual, involving patient-reported and radiological reviews. Virtual clinic models have previously been developed and evaluated for hip and knee replacement follow-up, and are already established in some centres. 52,171

Each major regional centre should identify a radiography facility with the appropriate expertise to offer this service. This may be run by the radiology department, the orthopaedic department or a combined multidisciplinary team. The expertise in interpreting joint replacement radiographs is more important than who runs this service. The specialist societies may consider validating/approving these centres. Evidence-based standardised radiology reporting (e.g. reporting methods previously developed by members of the UK SAFE team171) should be considered.

A local (paper or online) information leaflet that is available in multiple languages should detail the service provision for these patients. Further details of how to access the system and red flags for the patients should also be listed.

Integration with recent National Institute for Health and Care Excellence guidance

Recent NICE guidelines on hip, knee and shoulder arthroplasty (NG157)32 stated that the committee were unable to make recommendations on follow-up because of a lack of evidence in this area. The results of the UK SAFE study provide some of the missing evidence, although there is a need for further research, as detailed in the NICE guidelines.

Patient and clinician education

Both patients and clinicians whose follow-up arrangements are changed by these recommendations will require explanation, education and training. The exact methods used may differ from region to region, depending on local facilities available. Education and ownership by patients are the key to success in roll out of these new services. Local patient and public involvement (PPI) groups, interested GPs and secondary care teams should be involved in the planning of these services.