Setting and location

The registry captures data for all patients who undergo elective hip and knee arthroplasty at St Vincent’s Hospital, a tertiary referral centre for TJA surgery in Melbourne, Australia, which receives referrals from all of Victoria. In 2018, the population of Victoria was estimated at 6.3 million people, with 77% residing in Melbourne.14

Relevant dates

On 1 January 2005, as part of a PhD project, the registry was established. Data collection has been prospective since 1 January 2005, with retrospective data collection for cases from 1998 to 2004. The clinical data of all patients who underwent elective arthroplasty dating back to 1 January 1998 were extracted and added to the registry. This was the earliest available date that data could be reliably extracted from patient medical records. Cases were prospectively added from 1 January 2005 and routine prospective collection of PROMs was introduced from 1 January 2006. Any ad hoc PROM collection related to specific studies that occurred prior to 1 January 2006 were also added to the Registry. Data collection is ongoing and approximately 750 to 800 new cases have been added to the registry in each of the five calendar years 2015–2019. This represents 3% of cases performed in the state of Victoria.15

Eligibility criteria and how patients are recruited

All patients who undergo elective total hip or knee arthroplasty at St. Vincent’s Hospital (Melbourne) are eligible for inclusion on the registry. SMART captures data on primary elective total hip and knee arthroplasty, as well as unicompartmental and patellofemoral knee, and revision hip and knee arthroplasty procedures. Informed consent is currently obtained from participants preadmission (ie, prior to their admission assessment) using an ‘opt-out’ approach, which has been ratified by the St Vincent’s Hospital Melbourne (SVHM) ethics committee. The form is attached as online supplemental file 2. Specifically, when patients are scheduled for a preadmission assessment in preparation for arthroplasty surgery, they are sent an information pack containing the opt-out consent form, which is available in 11 languages: English, Arabic, Croation, Greek, Italian, Macedonian, Simplified Chinese, Traditional Chinese, Spanish, Turkish and Vietnamese. The Registry has 100% capture of elective arthroplasty procedures and <0.1% of patients opt-out of completing PROMs.

Description of data collection methods and follow-up

Data collected includes patient demographics, current diagnoses and history of comorbidities such as cancer resection, and confounders such as smoking history (see online supplemental file 1) for the specific variables collected and the source of each variable). Figure 2 outlines the process of baseline and follow-up data collection. Information is initially obtained by resident doctors in preadmission clinic and via correspondence with other health professionals involved in the patient’s care. For example, comorbidities recorded during the preadmission clinic assessment are cross referenced against anaesthetist assessment recorded on the day of surgery. Preoperative PROM data are collected by a research assistant (RA) prior to surgery. Baseline patient clinical data are extracted from the St Vincent’s digital clinical databases—Medical Records On-line and Patient Administration System—by an RA after the 6-week postoperative follow-up appointment, which they subsequently add to the patient’s Registry file. Since the inception of the Registry, the baseline variables include: age; sex; body mass index (BMI)—measured on the day of surgery by the perioperative nursing team and classified according to WHO definitions as underweight (BMI <18.5 kg/m²), normal weight (BMI 18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), obese class I (BMI 30–34.9 kg/m²), obese class II (BMI 35–39.9 kg/m²) or obese class III (BMI >40 kg/m²); a measure of comorbidity status obtained from the preoperative anaesthetic notes taken on the day of surgery according to the American Society of Anesthesiology (ASA) Classification; patient postcode is used to determine Socioeconomic Index for Area and Accessibility/Remoteness Index Australia classification—the postcode is mapped to these classifications by a trained RA, who enters them into the case report form (CRF) (see online supplemental file 3a,b) and the Registry database; aetiology; radiographic OA severity, graded using the Kellgren-Lawrence classification system applied to the preoperative plain films, taken within 6 months of surgery.16–23 These are collected in line with the goals of the registry as per the rationale described by Gliklich et al.24 Hospital cost data, including surgery, imaging, pharmacy and theatre costs, are currently available for years 2006–2007 and 2012–2016 from the Hospital’s Decision Support Unit. All complications that represent a clinically meaningful alteration to the patient’s postoperative course are recorded, in line with guidelines issued by the Royal Australasian College of Surgeons.25 A clinically meaningful alteration includes, but is not limited to, investigations, interventions, delayed discharge, representation and/or readmission to the hospital, and revision surgery. This broad capture of complications enables classification systems, such as the Clavien-Dindo system, to be applied to the SMART Registry because detailed data are rigorously routinely collected.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Flow chart—baseline data collection and follow-up. MRO, Medical Records On-line; PAS, Patient Administration System; PROM, patient-reported outcome measure; RA, research assistant.

We achieve approximately 98% follow-up of PROMs at 1 year, 95% at 2 years, 85% at 5 years and 75% at 7 years. Lost to follow-up is almost exclusively due to patient death, with <1% at any timepoint due to other reasons.

A range of surgery-related and prosthesis-related variables are recorded including surgical approach, alignment referencing technique, type of anaesthesia, wound drain output and surgeon level of training. Prosthesis-related variables include implant brand, type of cement (if used), size of components, number of bone screws used (if applicable), femoral offset (hips) and degree of constraint (knees). Full details of the variables captured are available in the data dictionary (online supplemental file 1).

Revision arthroplasty is defined in line with the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR).26 Any further intervention required for the prosthesis, such as liner exchange, stem replacement are included as a revision procedure, which is recorded both as a complication related to the primary procedure and as a procedure of its own such that details specific to the revision surgery are captured comprehensively with the same rigour as for primary procedures. Washout and debridement for infection are recorded elsewhere as complication a complication rather than revision. The following measures are undertaken to maximise the capture of revision surgeries:

1. Patients are contacted for PROM surveys, at which time they are asked about any additional procedures related to their arthroplasty.

2. The operating theatre list is scrutinised each week in order to identify patients booked for revision surgery.

3. Data on revision procedures that are performed elsewhere are captured because the treating hospital makes a request to our institution for details of the components used during the index procedure as part of their planning for the revision procedure.

4. While individual patient data are not available from the AOANJRR, aggregate data are available which facilitates comparison between SMART Registry revision rates and national revision rates in order to flag whether there is the potential for substantial under-reporting of revision procedures.

Over time, the nature of the research questions able to be answered using the data has developed as a result of changes and improvements in Registry data capture. Figure 1 presents measures in the registry for which the capture has changed over time in order to broaden the scope of research that can be conducted using the data. Routine collection of baseline quality of life and condition-specific questionnaires for knee (n=4722) and hip (n=4393) arthroplasty, commenced from 1 January 2006 onward. Since then, one condition-specific and one quality of life measure has been routinely collected pre-surgery and at the follow-up time points (see online supplemental file 1). The Knee Society Score (KSS) contains both patient-reported and surgeon-reported components.27 The scoring system is both clinician derived and patient derived and consists of a knee subscale completed by the surgeon, and satisfaction, functional activities and expectation subscales which are completed by the patient, with higher scores indicating better outcomes.28 The Harris Hip Score (HHS) was developed to assess the results of hip surgery and contains 10 items covering the domains of pain (one item, 0–44 points), function (seven items, 0–47 points), absence of deformity (one item, 4 points) and range of motion (two items, 5 points).29 The maximum score is 100 points, with higher scores indicating better results. Commencing on 1 January 2012, the HHS and KSS condition-specific questionnaires were superseded by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) for two main reasons: (1) to solely capture pain and function from the patient perspective as opposed to using questionnaires that give weight to the clinician perspective and (2) the use of the WOMAC has been endorsed by peak international bodies (Osteoarthritis Research Society International, Outcome Measures in Rheumatology as a valid measure for use in OA and TJA in multiple populations and is available in >80 languages.30 The WOMAC is a valid 24-item measure of pain and function for people with knee and hip osteoarthritis (OA) and for TJA.31–36 The WOMAC consists of three subscales that measure pain (0–20), joint stiffness (0–8) and physical function (0–68), with a lower score representing a better outcome. The Veterans RAND 12-item Health Survey (VR-12), a health-related quality of life instrument which measures well-being with a Physical Component Score (PCS) and Mental Component Score (MCS), is also administered.34 37 38 The VR-12 PCS comprises four scales: physical functioning, role-physical, bodily pain and general health; the MCS also comprises four scales: vitality, social functioning, role emotional and mental health. Higher scores on the VR-12 indicate better quality of life. An overall utility score can also be calculated.39

Introduced on 1 January 2006, the Charlson Comorbidity Index (CCI) is calculated from the data obtained from the preoperative anaesthetic notes taken on the day of surgery, and provides an accurate prediction of mortality based on weightings of certain comorbidities.21–23 40 41 Anaesthetists independently assess patients through history and examination as well as medical record review of presurgery assessments performed by the resident medical officer in preparation for surgery, for presence of active comorbidities, and previous cancer diagnoses.

Data linkage

Baseline and follow-up data have been probabilistically linked with MBS federally subsidised healthcare and PBS medication dispensing records. The data linkage process was carried out by the Australian Institute of Health and Welfare Data Linkage Unit and involved a probabilistic linkage procedure consisting of 19 passes with clerical review for ‘possible’ record links. Each pass consisted of deterministic pairwise matching of records from SMART and Medicare Enrolment File (MEF) on selected blocking variables. The MEF has been described elsewhere, and the version used for this linkage included all Medicare customers registered from the foundation of Medicare (1 February 1984) to 30 December 2018.42 Weights were calculated based on probabilities of agreement and disagreement for the blocking and match variables for each respective match pair in the block. The blocking and match variables were based on surname, given name(s), postcode and day, month and year of birth. An optimal threshold algorithm was used to determine the optimal record pair.

The first of five annual linkages occurred over September 2019, and four additional annual linkages have already been approved. Annual linkages are funded through the National Health & Medical Research Council Centre (NHMRC) of Research Excellence in Total Joint Replacement. The planned date of these future linkages is due in October each year. Access to the linked data has been approved from 8 May 2018 to 31 December 2028, with the opportunity for additional annual linkages within this period depending on available funding.

The MBS records contain information on services that qualify for a benefit under the Health Insurance Act 1973, for which a claim has been processed. Briefly, MBS records include information about type of service provided, benefits paid and patient characteristics (such sex, year of birth and geographical region).43 44

The PBS data collection contains information on prescription medicines that qualify for a benefit under the National Health Act 1953, and for which a claim has been processed. The database comprises information about the PBS scripts and payments, patients (such as a patient identifier, beneficiary status and patient location), prescribers and dispensing pharmacies.45–47

Data management and quality assurance

Data entry and questionnaire follow-up is overseen by a dedicated Registry Coordinator, aided by two RAs. Data are housed on a dedicated orthopaedic research software system, Standardised Orthopaedic Clinical Research and Treatment Evaluation Software (Socrates).48 Registry data are audited annually by the principal investigator. Mortality data are checked against data from the Registrars of Births, Deaths and Marriages via the AOANJRR.26 The SMART Registry is separate to the AOANJRR, but patients undergoing arthroplasty at St Vincent’s are also included in the National Registry. One of the major differences between the National Registry and SMART Registry is the duration of PROMs data collection, which commenced in 2018 for the National Registry with a preoperative and 6-month postoperative PROM survey, while the SMART Registry has been collecting longitudinal PROMs data comprehensively since 2006 (see online supplemental file 1).49 Individual surgeons at SVHM are also able to use peer benchmarking of their surgical, clinical and patient-reported outcomes, by comparing their own data with aggregate, deidentified data from the other SVHM arthroplasty surgeons. This is performed as part of the annual audit and report of the SMART Registry.

Data validity

The data are extracted from the medical record and entered into a CRF (online supplemental file 3a,b) by trained RAs who liaise with the Registry custodian when challenges arise, and seek further detail from the surgeon who operated on the given patient whenever necessary. The CRF contains pre-established options for as many variables as possible, to minimise spelling errors and data entry errors. Prosthetic joint infections are identified and entered on the database using validated criteria.50

Th registry coordinator, who has extensive training and experience, enters data from the completed CRF onto the database. The database contains pre-established options, with minimal ‘free-text’ fields. The registry coordinator’s office is colocated with the registry team and when potentially erroneous data entry items are identified a request for a data check with a second member of the team is initiated.

A registry custodian and principal investigator conduct routine annual audit during which the data are extensively checked and cleaned.

When researchers initiate a research project using registry data, either a custodian or data scientist extract, deidentify and clean the data. Data validity is further checked by a registry biostatistician to assess missingness. Any missingness triggers a process whereby a registry assistant is assigned the task of retrieving missing variables from the patient medical records or from the Hospitals Health Information Service. Finally, if any variable is genuinely missing (ie, data which could not be recovered from the original data collection form), researchers must apply the most appropriate method to address missingness when conducting data analyses. The method of choice for this will depend on the analysis being carried out. The biostatisticians associated with the registry provides advice on the handling of missing data for studies utilising SMART data.

Before annual linkage with AIHW, data checks are run on each variable with the aim to identify both missing and implausible data. These entries are then flagged with the registry custodian who engages the research team to investigate.

Characteristics of study participants and information on exposures and potential confounders

Table 1 outlines a breakdown of procedure types and numbers recorded in the registry as of 30 June 2018.

Online supplemental file 4 depicts baseline demographic and comorbidity characteristics of registry patients who have undergone primary elective total knee (n=6048) or total hip (n=5811) arthroplasty between 1 January 1998 and 30 June 2018. This is presented alongside the AOANJRR and the Australian Bureau of Statistics National Health Survey (ABS NHS) characteristics wherever comparable information is available. Comparison with these national data sources indicates that registry patients are broadly representative of the Australian TJA and OA populations in terms of sociodemographic characteristics (online supplemental file 4), although with a higher proportion of females in the TKA cohort than the Australian population. Of note, a higher proportion of Registry patients live in the major cities than the ABS NHS patients. This is to be expected considering the registry is based in a large urban centre, whereas the ABS NHS sample comprises patients from all of Australia. This could contribute to the other notable difference, namely that Registry patients have a higher average socioeconomic status (SES), given that that SES decreases with increasing rurality.51 It is also important to note that the SMART Registry contains a higher proportion of ASA class 3 patients.

In Australia, public hospitals such as SVHM are funded by the government, predominantly at the state level but with federal contribution also; public patients do not incur out-of-pocket costs for surgery. Private hospitals receive federal government funding, with hospital costs supplemented by the patient’s private insurance, with patients liable for any gap funding.52 Most of the data (83%) in the registry is obtained from patients treated at the public campus of SVHM hospital. The remaining 17% is obtained from patients treated at the private SVHM campus.

Clinical

Dowsey et al carried out survival analysis on registry arthroplasty patients and compared with the results obtained from the same analysis using a large US-based joint replacement registry.53 Beyond 1 year, in the registry data, patients in the overweight, obese class I, II and III were each found to have lower risk of mortality as compared with normal weight patients. These findings were also obtained in the Kaiser Permanente Total Joint Replacement Registry, with the additional finding that patients were found to be at higher risk of mortality compared with normal weight patients, beyond 1 year. Despite many studies suggesting an increased risk of medical complications in obese patients undergoing TJA, even severe obesity is not independently associated with increased risk of mortality. Withholding surgery on the basis of obesity alone may not be justified in light of these findings.

Health behaviours

Bunzli et al adopted inductive thematic to identify concepts related to patient expectations of TKA.54 A preliminary framework of therapeutic misconception captured discrepancies between anticipated and actual experiences reported by a sample of patients 1-year post-TKA. By assessing and addressing patients’ expectations for the TKR journey, including perioperative, postoperative and long-term outcomes, surgeons can help ensure that patients consent for surgery fully informed about what surgery involves and what their likely outcomes will be.

Health costs

Peel et al developed a negative binomial regression model to quantify factors associated with costs in the first 30 days following TJA.3 The findings were extrapolated to the Australian population and it was found that preventable postoperative complications were major cost drivers, with surgical site infection (SSI) adding $A97 million, venous thromboembolism adding $A66 million and delirium adding $A64 million to the total annual cost of TJA in Australia, which already exceeds $A1 billion. Reducing the rate of postoperative complications in TJA represents an opportunity to make a substantial impact on the overall cost burden on the healthcare system.

PROMs

Dowsey et al used Latent Class Growth Analysis to classify groups of patients according to their trajectory of knee pain/function over 1–5 years post-TKA.55 Modifiable predictors of poor response; namely poor function and/or moderate-to-severe pain, were identified: baseline comorbidity, physical and mental well-being and obesity. Clinicians can use these findings to identify patients with a higher likelihood of poor surgical response, and incorporate this information into a shared decision-making process.