Ethics approval and consent to participate

With support under Section 251 of the NHS Act 2006, the Ethics and Confidentiality Committee (ECC) (now the Health Research Authority Confidentiality Advisory Group) allows the NJR to collect patient data where consent is indicated as “Not Recorded.” 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.” The NJR research committee approved this analysis, and the NHS Health Research Authority tool guidance dictates that the secondary use of such data for research does not require approval from a research ethics committee.

Data sources

This is an observational study using prospectively collected anonymised data for patients in the NJR. The NJR registers mandatory data from surgeons and their hospitals regarding all joint replacement activity for hips, knees, shoulders, elbows and ankles whether they were funded by the NHS or independently in England, Wales, Northern Ireland, the Isle of Man and Guernsey. The NJR passes details of patients with primary or revision joint replacement procedures to the NHS Personal Demographics Service which provides dates of death from the Office for National Statistics if the NHS number is traceable. Information on the area of residence of patients, as defined by the 2011 census Lower Layer Super Output Areas (LSOAs) were collected. The relative level of deprivation of patient’s area of residence was obtained using the Index of Multiple Deprivation (IMD) version 2015 [17]. Each LSOA was linked to an IMD deprivation score and ranked according to deprivation deciles. IMD deciles were calculated by ranking the 32,844 small areas in England from most deprived to least deprived and splitting them into 10 equal-sized groups [17]. The exposure was IMD quintiles derived from the IMD deciles, with deciles 1 and 2 producing quintile 1 (Q1)—associated with the 20% most deprived LSOAs nationally—up to deciles 9 and 10 creating quintile 5 (Q5), representing the 20% of least deprived LSOAs.

Patients eligible for inclusion received a primary THR between 1 January 2007 and 31 December 2017 for the sole indication of osteoarthritis (which was the most common operative indication 87.9%) [18], and were aged 50 years or more, operated on and living at the time of surgery in England as indicated by LSOA, and had follow-up data to 2023 for outcomes. Patients who had metal-on-metal implants or simultaneous (performed on the same day) bilateral THRs were excluded. Those included had a first unilateral THR procedure only, either (i) a unilateral THR or (ii) a non-synchronous bilateral THR performed on different dates (in which case only the outcomes for the first replacement were examined).

The NJR data were linked to routinely collected Hospital Episode Statistics (HES) Admitted Patient Care and NHS England PROMs databases. NJR records could be linked to HES or PROMs if they took place in an NHS Trust or were funded by the NHS in England. The unit of analysis in this cohort was the individual.

Outcomes

Outcomes examined following THR surgery were:

1. (i) Cumulative 90-day all-cause mortality: Date of death minus date of primary THR was calculated to determine mortality status at 90-days following this operation.
2. (ii) Cumulative 5-year hip revision: Date of first revision minus date of primary THR for the same joint were used to determine all-cause revision status at 5 years. The three most common reasons for revision were examined: any indication which included dislocation, infection, or aseptic loosening.
3. (iii) Cumulative health complications recorded by 6 months: Health complications at 6 months post-primary THR were determined using selected International Classification of Diseases (ICD10) codes recorded in 20 HES diagnosis fields for each admitted patient care episode. Complications were: (a) health-related: urinary tract infection; respiratory tract infection (including pneumonia); acute renal failure; acute myocardial infarction; stroke (excluding mini stroke); (b) surgery-related: complication of prosthesis; pulmonary embolism or deep vein thrombosis; surgical site infection; wound disruption; fracture after implant; and neurovascular injury (S1 Table). In addition, blood transfusion was included as a surgery-related complication and was determined using selected Classification of Interventions and Procedures (OPCS-4) codes recorded in 24 HES operation fields (S1 Table).
4. (iv) Cumulative rehospitalisation and reoperation for an orthopaedic indication by 1 year: Codes from existing publications were extracted and indicated on a comprehensive ICD10/OPCS-4 code list, these lists were reviewed and classified by an orthopaedic surgeon (JF). Rehospitalisation was determined by searching diagnosis fields for specific ICD10 codes (S2 Table) recorded in the year post-primary THR. Similarly, reoperation was determined by searching operation fields for specific OPCS-4 codes (S3 Table) in the year post-primary. Primary or revision hip admissions or operations recorded in the NJR database were excluded.
5. (v) PROMs using the Oxford Hip Score (OHS): OHS [19] for pain and function related to the affected hip was measured up to 6 months before (pre) and up to 12 months after primary THR (post) [20]. OHS was derived from 12 questions, related to hip pain and function domains, added to form an overall total, with subscale scores related to each domain (each utilising 6 questions) [19]. Each question is scored from 0 to 4 with 4 being the best outcome, with a maximum of 48 for the full score and 24 for the subscales [21]. The difference in the scores was used to determine the change in score from pre- to post-THR. The Minimal Clinically Important Difference (MCID) for the full score was set at five points [22]. MCID was dichotomised (0/1) as ‘Improved ≥5 points’ (coded as 0) versus ‘Not improved or worse <5 points’ (coded as 1), respectively, to allow presentation of findings with other dichotomous outcomes using a clinically relevant measure.

Adjustment factors

Other variables included: sex (female, male); age group (50–54, 55–64, 65–74, 75+ years old); body mass index (BMI < 18.5 underweight, 18.5–24.9 normal, 25–29.9 overweight, ≥30 kg/m² obese, unknown BMI); American Society of Anesthesiologists’ classification of pre-operative physical status (ASA grade P1-Fit and healthy, P2-Mild disease not incapacitating, P3-Incapacitating systemic disease, P4-Life threatening disease/P5-Expected to die within 24 h) [23]; and Royal College of Surgeon’s Charlson score (No comorbid conditions, Mild-one comorbid condition, Moderate-two comorbid conditions, Severe-three or more comorbid conditions) [24]. The Charlson score was calculated based on the presence of several chronic conditions, identified using ICD-10 codes at the hip replacement admission and all admissions in the preceding 3 years.

Statistical analysis

Prevalence rates were calculated as percentage or per 10,000 persons for IMD groups, year of joint replacement and other patient characteristics. Modified Poisson regression models [25] with a log link function and robust standard errors to account for clustering of patients within hospitals were used to compare rates and estimate rate ratios (RRs), as well as to calculate absolute differences for risk differences (RDs). The number needed to harm (NNTH) were calculated as the reciprocal of the RD. The models were adjusted for patient characteristics measured at the primary operation: patient sex, age group, BMI, ASA grade and Charlson score as categorised above. As the association between age and many of the outcomes was non-linear, age was modelled using age groups to account for this non-linearity. Linear regression models were used with the continuous post-op OHS as the outcome variable, adjusting for the pre-operative OHS, IMD deprivation score as the primary exposure variable, and the patient characteristics of sex, age group, BMI, ASA grade and Charlson score.

Sensitivity analysis calculated adjusted RRs for: (1) the larger cohort of those that did not have linked HES data but for whom NJR outcomes of mortality and revision were available (adjusted for sex, age group, BMI and ASA grade at primary operation, but not Charlson score as this is derived from HES data); the main outcomes were additionally adjusted (2a) for pre-THR OHS for those patients who had PROMs scores reported; and (2b) for calendar year. Analyses were conducted using Stata version 18.5 statistical software (StataCorp, College Station, Texas). The RECORD (REporting of studies Conducted using Observational Routinely-collected health Data) guideline was followed to report the study (S1 Checklist) [26].

Patient involvement

Patient representatives sit on the committee structure of the NJR. The research priorities of the NJR are identified and approved by patient representatives as part of this committee. One of the NJR patient representatives (RB), was involved in setting the research question and outcome measures, contributed to designing this work, discussing findings, and interpreting the results in a meaningful way for a lay audience. We are unable to disseminate results of this study directly to study participants due to the anonymous nature of the data. We plan to disseminate our findings to the NJR, via their communications team, to consultations relevant to the provision of joint replacement and to the general population through the local and national press.

Ninety-day all-cause mortality

Of 1,695 (0.4%) patients that died within 90 days of primary THR (S5 Table), the median time to death was 29 days, interquartile range 10–58 days. The adjusted RR of 90-day mortality reveals that there was higher mortality in the most deprived group compared to the least deprived group (Fig 2); (adjusted RR: 1.25 (95% confidence interval, CI [1.07, 1.46])). However, the absolute RD was 11 per 10,000, with an adjusted RD of 9 (95% CI [2, 16]) per 10,000 (Fig 2) and adjusted NNTH 1,105 (95% CI [640, 4,044]) (S5 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Rate ratios and risk differences for outcomes by Index of Multiple Deprivation (IMD) group adjusted for sex, age group, BMI, ASA grade and Charlson score at primary operation.

MCID, Minimal Clinically Important Difference for the full Oxford Hip Score set at a five-point increase from pre- to post-total hip replacement operation (coded: 0’Improved ≥5 points’; 1’Not improved <5 points’) [21,22]. Note: All outcomes (except Oxford Hip Score) N = 448,184: Q5 Least deprived (reference) n = 106,374; Q4 n = 108,614; Q3 n = 100,676; Q2 n = 76,521 and Q1 Most deprived n = 55,999. For Oxford Hip Score N = 200,522: Q5 Least deprived (reference) n = 50,741; Q4 n = 50,174; Q3 n = 45,259; Q2 n = 32,612 and Q1 Most deprived N = 21,736. Abbreviations: ASA, American Society of Anesthesiologists’; BMI, Body Mass Index; CI, Confidence Interval; IMD, Index of Multiple Deprivation; N, number; MCID, Minimal Clinically Important Difference; Q, quintile.

https://doi.org/10.1371/journal.pmed.1004870.g002

Five-year revisions

Among the 7,023 (1.6%) of patients who underwent revision surgery within the first 5 years after their primary THR (S5 Table), the median time to revision was 1.4 years, interquartile range 0.3–2.9 years. The adjusted RRs for revision, in the most deprived group compared to the least deprived group, were similar at 1.02 (95% CI [0.94, 1.10]) (Fig 2). The absolute RD was 0.12% or 12 per 10,000. The adjusted RD was 0.02% (95% CI [−0.10, 0.15]) (Fig 2) and adjusted NNTH of 4,123 with wide CIs which included number needed to benefit (NNTB) (S5 Table) [27]. Three common indications for revisions: dislocation, infection or aseptic loosening, produced similar findings to all-cause revisions.

Complications at 6 months

Of 21,527 (4.8%) patients that had one or more complications within 6 months of primary THR, with 4.3% having complications in the least deprived group compared to 5.7% in the most deprived (S5 Table). Complications were more likely among those from the most deprived group compared to the least deprived group (the reference) (adjusted RR: 1.26 (95% CI [1.21, 1.32])) (Fig 2). The most common complications overall were complication of prosthesis, urinary tract infections and respiratory infections with rates of 111, 100 and 98 per 10,000 patients, respectively (S6 Table). The difference in rates of complication between the most and least deprived group were greatest for respiratory infection by 66 (95% CI [55, 78]) per 10,000 patients (adjusted RR 1.67 (95% CI [1.52, 1.84]); adjusted RD 54 (95% CI [43, 65]) per 10,000 patients and adjusted NNTH 185 (95% CI [155, 231])). In contrast, the differences were smaller for surgical related complications with prosthesis related complications being 20 (95% CI [9, 31]) per 10,000 more for the most deprived group (adjusted RR 1.15 (95% CI [1.05, 1.26]); adjusted RD 16 (95% CI [5, 27]) per 10,000 patients and adjusted NNTH 615 (95% CI [366, 1,936])) (S6 Table and Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Rate ratios and risk differences for complications by Index of Multiple Deprivation (IMD) group adjusted for sex, age group, BMI, ASA grade and Charlson score at primary operation (N = 448,184).

Note: All outcomes N = 448,184: Q5 Least deprived (reference) n = 106,374; Q4 n = 108,614; Q3 n = 100,676; Q2 n = 76,521 and Q1 Most deprived n = 55,999. Abbreviations: ASA, American Society of Anesthesiologists; BMI, Body Mass Index; CI, Confidence Interval; IMD, Index of Multiple Deprivation; N, number; PE/DVT, pulmonary embolism/deep vein thrombosis; UTI, urinary tract infection; Q, quintile.

https://doi.org/10.1371/journal.pmed.1004870.g003

Rehospitalisation and reoperation

Of 448,184 patients with linked HES data, 78,968 (17.6%) were rehospitalised for orthopaedic indications within 1 year of primary THR and 6,637 (1.5%) underwent non-revision reoperations for orthopaedic indications within 1 year of primary THR. Those from the most deprived group were more likely to be rehospitalised (adjusted RR: 1.16 (95% CI [1.14, 1.19])) or reoperated (adjusted RR: 1.23 (95% CI [1.13, 1.33])) in the year following the primary THR compared to the least deprived group (Fig 2). These findings in terms of RD were 2.78% (95% CI [2.39, 3.17]) for rehospitalisation and 0.31% (95% CI [0.19, 0.44]) for reoperation (Fig 2), with adjusted NNTH of 36 (95% CI [31, 42]) and 319 (95% CI [227, 539]), respectively (S5 Table).

Patient reported Oxford Hip Score (OHS)

Of the 200,522 patients (45%) with valid OHS data in terms of completeness and within the time limits, the pre-operative OHS were collected at a median of 18 (interquartile range, IQR 8–37) days prior to THR surgery and post-operative scores were collected at 6.7 (IQR 6.5 to 7.4) months. Table 1 describes those with complete PROMs data in comparison to those with HES-linked data but missing valid PROMs data. The percentage missing scores in each IMD group varied from 52% to 61% (p < 0.001) compared to the 448,184 patients with other outcome data. There is an improvement for (unadjusted) mean pre- to mean post-operative OHS scores with similar improvements for all IMD groups (S1 Fig). Those from the most deprived group had a mean OHS up to 12 months after primary THR of 37.31 (95% CI [37.21, 37.42]) compared to those from the least deprived group (40.28 (95% CI [40.21, 40.35])) adjusted for patient characteristics and pre-operative OHS (Fig 4). The difference of −2.97 (95% CI [−3.10, −2.84]) is lower than the MCID of more than 5 points. Similar findings was seen for the subscales of pain and function (S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Total Oxford Hip Score (OHS) post-operation adjusted for OHS pre-operation, sex, age group, BMI, ASA grade and Charlson score by Index of Multiple Deprivation (IMD) group (N = 200,522).

Note: For Oxford Hip Score: Q5 Least deprived (reference) n = 50,741; Q4 n = 50,174; Q3 n = 45,259; Q2 n = 32,612 and Q1 Most deprived N = 21,736. Abbreviations: ASA, American Society of Anesthesiologists; BMI, Body Mass Index; CI, Confidence Interval; IMD, Index of Multiple Deprivation; N, number; OHS, Oxford Hip Score; Q, quintile.

https://doi.org/10.1371/journal.pmed.1004870.g004

In terms of MCID, as a binary outcome change in score before and after surgery, those from the most deprived group were more likely to have worse outcome of less than five points improvement RR: 1.64 (95% CI [1.55, 1.75]) compared to the reference group of the least deprived group (Fig 2). The most deprived were therefore more likely to not improve (have <5 point difference) than the least deprived.

Sensitivity analysis

(1) Outcomes of 90-day mortality and 5-year revision rates were examined among 550,910 patients, which included those that did not link to HES data but for whom mortality and revision NJR data were available. For these outcomes, the findings were similar: adjusted RR (most compared to least deprived): 1.34 (95% CI [1.15, 1.55]) and 1.01 (95% CI [0.94, 1.09]), respectively. Adjustment was for the same patient characteristics as the main models apart from Charlson score, as this is derived from HES data. (2a) The main outcomes of 90-day mortality, 5-year revision rates, 6-month complications and reoperation and rehospitalisation at 1 year were additionally adjusted for pre-THR OHS where available, this reduced the sample size to N = 260,427 and findings were broadly similar albeit with weaker associations for mortality at 90-days. (2b) As the number of operations increased over the decade, year of operation was included in the adjusted model (S3 Fig) the findings were not substantively different from those presented above.