Patient-Relevant Outcomes Following First Revision Total Knee Arthroplasty, by Diagnosis

Background: The purpose of this study was to investigate patient-relevant outcomes following first revision total knee arthroplasties (rTKAs) performed for different indications. Methods: This population-based cohort study utilized data from the United Kingdom National Joint Registry, Hospital Episode Statistics Admitted Patient Care, National Health Service Patient-Reported Outcome Measures, and the Civil Registrations of Death. Patients undergoing a first rTKA between January 1, 2009, and June 30, 2019, were included in our data set. Patient-relevant outcomes included implant survivorship (up to 11 years postoperatively), mortality and serious medical complications (up to 90 days postoperatively), and patient-reported outcome measures (at 6 months postoperatively). Results: A total of 24,540 first rTKAs were analyzed. The patient population was 54% female and 62% White, with a mean age at the first rTKA of 69 years. At 2 years postoperatively, the cumulative incidence of re-revision surgery ranged from 2.7% (95% confidence interval [CI], 1.9% to 3.4%) following rTKA for progressive arthritis to 16.3% (95% CI, 15.2% to 17.4%) following rTKA for infection. The mortality rate at 90 days was highest following rTKA for fracture (3.6% [95% CI, 2.5% to 5.1%]) and for infection (1.8% [95% CI, 1.5% to 2.2%]) but was <0.5% for other indications. The rate of serious medical complications requiring hospital admission within 90 days was highest for patients treated for fracture (21.8% [95% CI, 17.9% to 26.3%]) or infection (12.5% [95% CI, 11.2% to 13.9%]) and was lowest for those treated for progressive arthritis (4.3% [95% CI, 3.3% to 5.5%]). Patients who underwent rTKA for stiffness or unexplained pain had some of the poorest postoperative joint function (mean Oxford Knee Score, 24 and 25 points, respectively) and had the lowest proportion of responders (48% and 55%, respectively). Conclusions: This study found large differences in patient-relevant outcomes among different indications for first rTKA. The rate of complications was highest following rTKA for fracture or infection. Although rTKA resulted in large improvements in joint function for most patients, those who underwent surgery for stiffness and unexplained pain had worse outcomes. Level of Evidence: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.

pTKAs are revised within 10 years 1 and that the need for revision total knee arthroplasty (rTKA) is steadily increasing 5,6 .This trend reflects aging populations and an increasing number of joint replacements among younger patients 4 .
Research has demonstrated that rTKA produces less improvement in joint function and health-related quality of life than that produced by pTKA 7,8 .Previous work has highlighted clinically important differences in patient outcomes associated with the indication for surgery 9,10 .rTKA for infection or other urgent indications has been shown to incur greater costs and higher rates of complications than elective rTKA performed for other indications [10][11][12][13] .Researchers have also reported that fewer than half of patients who underwent rTKA for stiffness were satisfied with the outcome, compared with two-thirds of patients who underwent surgery for a worn component 9 .
We previously identified the domains of outcome that are important to patients undergoing joint replacement: implant survival, joint function, health-related quality of life, medical complications, and the impact of hospital admission 14 .Information on these outcomes in the context of different indications for rTKA is lacking and is needed to support shared decision-making.
The aim of the present study was to investigate patientrelevant outcomes following first rTKA performed for different indications.Utilizing routinely collected national data from the United Kingdom (U.K.), we investigated the rate of repeat revision surgery (re-revision) at 2 and 5 years postoperatively, the rate of mortality and serious medical complications up to 90 days postoperatively, patient-reported outcome measures at 6 months postoperatively, and the length of the hospital admission for the index procedure.

Materials and Methods
T his study was reported according to the RECORD (REporting of studies Conducted using Observational Routinely-collected health Data) checklist 15 .Ethical approval was obtained.

Study Data Sets
Routinely collected national data were analyzed from the National Joint Registry (NJR) for England, Wales, Northern Ireland, the Isle of Man, and Guernsey; Hospital Episode Statistics Admitted Patient Care (HES APC); National Health Service (NHS) Patient-Reported Outcome Measures (PROMs) 16 ; and the Civil Registrations of Death.

NJR
The NJR is a prospective register of rTKA procedures.Submission is mandatory for public and private health-care providers 17 .The NJR originally defined rTKA as "an operation performed to remove and replace one or more components of a total joint prosthesis, for whatever reason." 18The instructions on procedures to be reported changed over the study period, with the addition of secondary patellar resurfacing procedures on December 1, 2013, and the addition of debridement, antibiotics, and implant retention procedures with or without modular exchange on June 25, 2018 19 .

HES APC
HES APC includes NHS-funded secondary care episodes in England.Data are entered by trained clerks.Submission is mandatory in order for health-care providers to receive financial remuneration.Diagnoses are coded using World Health Organization International Classification of Diseases, 10th revision, codes 20 .Procedures are coded using the Office of Population Censuses and Surveys' Classification of Surgical Operations and Procedures, version 4, codes 21 .

PROMs
Patients completed a questionnaire preoperatively and at 6 months postoperatively 16 (see the section entitled "Outcome Measures," below).

Civil Registrations of Death
Date and cause of death are recorded from death certificates.

Inclusion and Exclusion Criteria
Patients undergoing a first rTKA between January 1, 2009, and June 30, 2019, were included.Cases were censored on December 31, 2019 (or earlier if an event was observed).We excluded early NJR data (for poor compliance 22,23 ) and data during the COVID-19 pandemic (the data were not representative of usual practice).Adult patients ( ‡18 years old) were eligible for inclusion following a first rTKA recorded in the NJR.rTKAs with no linked pTKA were excluded because we could not determine the sequence of events.For patients who underwent rTKA for both knees, only the first procedure was included to fulfill the assumption of independence of observations.Partial knee arthroplasty procedures (for example, unicompartmental and patellofemoral arthroplasties) were excluded.For the analysis of medical complications and the length of hospital stay (LOS), first-linked rTKAs were joined to corresponding episodes on HES APC.For the analysis of PROMs, preoperative PROM questionnaires needed to be returned within 90 days prior to rTKA and postoperative questionnaires needed to be returned 180 days to 1 year following rTKA.

Data Cleaning and Linkage
The statistical code is provided on GitHub 24 .The preparation of HES APC data followed the principles of HES Pipeline R 25 .Records in each dataset were de-duplicated, and variables were checked for out-of-range, missing, and invalid entries.
The NJR was used as the master data set for record linkage.The NNNid field was assigned a study_id and represented 1 patient.Patient identifiers stored within the NJR (NHS number, family name, given name, gender, date of birth, and postcode) were supplied to NHS Digital to create an HES patient cohort.NJR records were joined to the Civil Registrations of Death on study_id where available.For the analysis of complications and that of LOS, the index NJR procedures were joined to HES APC on study_id and date of surgery (±7 days).PROM data were linked indirectly to NJR on epikey with use of HES APC 16 .

Groups
Patients were grouped by the indication for revision surgery.Each procedure was assigned a single, dominant diagnosis: infection, malalignment, aseptic loosening, instability, fracture, progressive arthritis (in the case of secondary patellar resurfacing), stiffness, unexplained pain, or other 25,26 (see Appendix A).

Implant Survivorship
The cumulative incidence of re-revision surgery at up to 11 years was analyzed postoperatively.We have reported survival estimates at 2 and 5 years.

Mortality and Serious Medical Complications in the Hospital within 90 Days After Surgery
Deaths that occurred within 90 days following rTKA were identified from the Civil Registrations of Death.The following serious medical complications were recorded from HES APC: acute kidney injury, lower respiratory tract infection, myocardial infarction, deep-vein thrombosis, pulmonary embolism, stroke, and urinary tract infection.

PROMs
A total of 5 domains were measured: postoperative joint function, responder analysis, postoperative health-related quality of life, satisfaction, and perceived success.
1. Postoperative joint function: This outcome was measured with use of the Oxford Knee Score (OKS) 27 .The OKS is a 12-item Likert instrument with good measurement properties for the assessment of pain and function following rTKA 28,29 .For each item, the best response is assigned a score of 4 points, whereas the worst response is assigned a score of 0 points.The total possible score ranges from 0 points (worst) to 48 points (best) 30  was measured with use of the EuroQol-5 Dimensions-3 Levels (EQ-5D-3L) 31 .Each response set was converted to a utility score scaled for the U.K. population from 20.594 (worst) to 1 (best).A utility score of 0 represented a health state equivalent to being dead.4. Satisfaction: Patients were asked, "How would you describe the results of your operation?"The answer options were "Excellent," "Very good," "Good," "Fair," or "Poor."We considered patients who responded "Good" or better as satisfied.5. Perceived success: Patients were asked, "Overall, how are the problems now in the hip/knee on which you had surgery, compared to before your operation?"The answer options were "Much better," "A little better," "About the same," "A little worse," or "Much worse."We considered patients who responded "Much better" or "A little better" to have considered the surgery a success.

LOS
We calculated the duration in days of the single Continuous Inpatient Spell (CIPS) within HES APC that corresponded most closely with a first rTKA record in the NJR.One CIPS refers to 1 uninterrupted period within NHS secondary care.CIPSs were constructed in accordance with published methods 32 .Supplementary analyses were performed for finished consultant episodes and provider spells 33 .

Statistical Analysis
Patient characteristics (age, gender, American Society of Anesthesiologists [ASA] classification, modified Charlson Comorbidity Index [Summary Hospital-level Mortality Indicator Specification] 34 , body mass index [BMI], index of multiple deprivation, ethnicity, year of surgery) and the aforementioned outcome measures were grouped by indication.Continuous variables were described with use of means and standard deviations (SDs) or medians and interquartile ranges (IQRs) after the inspection of data distributions.Binary and categorical data were described with use of counts and percentages.The cumulative incidence of re-revision surgery was calculated from the complement of net implant survival (1 minus Kaplan-Meier).All procedures were censored at the time of a patient's death or at the end of the study (December 31, 2019).Kaplan-Meier survival curves were presented with 95% confidence intervals (CIs) and risk tables.Mortality and serious complications were presented as percentage frequencies with 95% CIs, with a normal approximation to a Poisson distribution having been assumed.The PROM linkage rate was defined as the total of knees with a completed preoperative questionnaire linked to an rTKA divided by the total of rTKAs in the NJR.Missing data were summarized descriptively for each variable.Since the missingness mechanism was unknown, data were assumed to be missing not at random and imputation was not performed (i.e., all analyses were performed with use of available data only).Statistical analyses were performed with use of R version 4.2.1 (R Foundation for Statistical Computing).An R Shiny application is available for readers to interact with the study results (https://shiraz-sabah.shinyapps.io/rKA-app/).

Source of Funding
This study was funded by a National Institute for Health and Care Research doctoral research fellowship.

Results
A total of 24,540 first-linked rTKAs (24,540 patients) were identified.Data cleaning and linkage are illustrated in Figure 1.Baseline patient characteristics are summarized in Table I.The mean age (and SD) at the first rTKA was 69 ± 10 years.More rTKAs were performed in female patients than in male patients overall (54%; 13,332 of 24,540 rTKAs), and the percentages of female patients were higher for all indications except infection (42%; 2,098 of 4,937) and stiffness (47%; 599 of 1,279).Patients who underwent rTKA for fracture or infection had more severe comorbidity than those who underwent rTKA for other diagnoses.A total of 495 (53%) of 938 rTKAs that were performed for fracture and 1,983 (40%) of 4,937 rTKAs that were performed for infection were done in patients who were ASA Class 3 or higher, compared with approximately 25% of rTKAs for the other indications.
Serious medical complications were common, with    The cumulative incidence of re-revision TKA by indication for first-linked rTKA.Risk tables and 95% CIs are provided in the Appendix. 1616

Discussion
P atients undergoing a first rTKA for infection were at a high risk for re-revision surgery (;1 in 6 at 2 years postoperatively; ;1 in 4 at 5 years postoperatively).The rate of re-revision surgery at 2 years was much lower for patients undergoing rTKA for other indications, ranging from 1 in 13 for instability to 1 in 37 for progressive arthritis.The rate of death within 90 days was high for patients undergoing surgery for fracture (;1 in 28) and infection (;1 in 56) but was £1 in 250 for patients treated for any of the other diagnoses.The rate of a serious medical complication requiring admission to a hospital was highest following rTKA for fracture (;1 in 5) and infection (;1 in 8) but was £1 in 14 for the other indications.Acute kidney injury (1 in 37) and lower respiratory tract infection (1 in 40) were the most frequently observed complications.
Deere et al. 35 recently investigated implant survivorship for first and re-revision rTKAs utilizing the NJR.Their reported rates of revision following first rTKAs (;7% at 2 years and ;12% at 5 years) were similar to ours and they found high rates of additional surgery following second and third rTKA procedures.We previously reported that the rates of mortality and serious complications following first rTKAs for elective indications were comparable with those following pTKAs, whereas rTKAs for urgent indications (such as infection and fracture) had poorer outcomes than both elective rTKAs and pTKAs 10 .We reproduced these findings in the present study with use of more contemporary data, describing patient subgroups in greater detail by utilizing surgeon-coded diagnoses recorded in the NJR.We also included additional domains of outcome (joint function, quality of life) in the present study through record linkage to NHS PROMs.This approach revealed differences in outcomes among elective, aseptic indications for rTKA.For example, approximately half of the patients who underwent rTKA for stiffness (48%) and unexplained pain (55%) reported a clinically meaningful improvement in joint function, compared with 73% of those who underwent rTKA for fracture and 72% of those who underwent rTKA for aseptic loosening.
In the present study, we analyzed routinely collected data from the NHS and private providers of rTKA in the U.K. over a 10year study period.The NJR is a mature registry and has been shown to have excellent data quality 22,23 , and HES APC and the Civil Registrations of Death provide near universal coverage of emergency hospital admissions and deaths in the U.K. The present study had limitations common to other observational research, including the limited information available within each data set.We stratified rTKAs by indication for surgery but did not control for confounding by indication.As such, the outcomes observed may be a result of differences in patient characteristics among the groups.The analysis was not exhaustive, and there may be other sources of heterogeneity that could be investigated in the future, such as differences in outcomes related to the type of components revised or between acute and chronic infection.The rates of serious medical complications are an underestimate of the true rates as they represent only patients admitted to secondary care.There may be large discrepancies in complication rates for conditions commonly managed in primary care or in outpatient departments.A gold standard for estimating postoperative complications does not exist, but recent studies have utilized a similar methodology to our own 36,37 .A high record attrition impacted the analysis of PROMs, such that PROM data were available for only ;1 in 6 rTKAs.This attrition was exacerbated by the requirement to link PROMs indirectly to the NJR via HES APC.Less record attrition may be possible in the future when the Master Person Service identifier is available from NHS Digital 38 .The rate of PROM attrition was particularly high for the infection and fracture groups.Although the PROM questionnaires returned by these groups demonstrated that a high proportion of patients responded to treatment and that there were large improvements in healthrelated quality of life, there is clear selection bias: patients who died within 6 months after surgery would not have returned a PROM questionnaire, and those who experienced a serious complication may have been less likely to have returned a PROM questionnaire.
The information from this study can be utilized to support the informed consent of patients undergoing rTKA and to set expectations for those patients.Patients with infection or fracture experienced high rates of adverse events (re-revision surgery, death, and serious medical complications).A priority for future research is to investigate the risk factors that can be modified to improve outcomes.The data analyzed here predate the extension of the Best Practice Tariff in the U.K. to include distal femoral and periprosthetic fractures 39 .Patients with stiffness and unexplained pain accounted for approximately one-fifth of first rTKAs performed during the study period.For approximately half of these patients, the observed improvements in PROM scores were small and clinically unimportant, whereas serious medical complications were comparable to those associated with other elective, aseptic indications.From the perspective of the NHS, it is important to establish the patient groups for whom these procedures are (or are not) cost-effective.Recent evidence suggests that more attention may need to be given to nonoperative treatment.The Support and Treatment After Replacement (STAR) care pathway was found to be more clinically effective and costeffective than usual care for patients with persistent pain at 3 months following pTKA 40 .The natural history of pain following pTKA also appears to be one of improvement over time, with 1 study demonstrating that two-thirds of patients recovered during the first 4 years postoperatively 41 .As such, we caution against early rTKA for unexplained pain.
In conclusion, this study demonstrated large differences in patient-relevant outcomes among different indications for first rTKAs.This information can be used to support the informed consent of, and to help set expectations for, patients undergoing a first rTKA.

Fig. 1 Flowchart 42 .
Fig. 1Flowchart demonstrating the attrition of study records during data preparation and record linkage.Where 2 tables are joined, the attrition of records is reported for the upper left table only (NJR records).Further information on HES APC fields (EPIEND, EPIKEY, EPISTART, FYEAR, PROCODE 3) is available within a data dictionary from NHS Digital 42 .

Fig. 2
Fig. 2 20 d OCTOBER 18, 2023 PAT I E N T-RELEVANT O U T C O M E S FOLLOWING FIRST REVISION TOTA L KNEE A RT H R O P L A S T Y, B Y D I AG N O S I S Postoperative health-related quality of life: This outcome 1,286 of 17,111 patients developing ‡1 complication within 90 days after surgery (complication rate, 7.5% [95% CI, 7.1% to 7.9%]).