BACKGROUND:The objectives were to evaluate whether the EQ-5D-3L questionnaire demonstrated a ceiling effect after total hip arthroplasty (THA), identify which patients were more likely to reach the ceiling score, and determine whether this limited assessment of potential improvement, relative to baseline, in their outcomes. METHODS:A cohort of 7,871 primary THAs was identified from an institutional arthroplasty database. Patient demographics, American Society of Anesthesiologists grade, socioeconomic status, Oxford Hip Score (OHS), and EQ-5D-3L were prospectively collected preoperatively and one and two years postoperatively. Regression analyses were performed to determine which preoperative factors were independently associated with achieving a ceiling score. RESULTS:The median preoperative EQ-5D-3L was 0.516, improving significantly at one year to 0.850 (P < 0.001), with nonsignificant change at two years (median, 0.883; P = 0.075). A ceiling effect was observed in 47.5% (n = 2,961) at one year, increasing to 48.6% (n = 2,853) at two years. Women, younger age, lower socioeconomic deprivation, and higher preoperative OHS, EQ-5D, and EQ-VAS were independent predictors of achieving a ceiling EQ-5D. Receiver operating characteristic curve analyses indicated that a preoperative EQ-5D score (≥ 0.534) predicted postoperative ceiling scores at one and two years (area under the receiver operating characteristic curve, 65.6 and 64.1%, respectively; P < 0.001). Patients who reached the postoperative EQ-5D ceiling had significant improvements in OHS, EQ-5D, and EQ-VAS and were more likely to achieve minimal important change and a patient-acceptable symptom state compared with those below the ceiling. CONCLUSIONS:The EQ-5D showed moderate ceiling effects at one and two years following THA, with higher preoperative scores being associated with ceiling attainment. Patients who reached the ceiling experienced greater improvements from baseline and were more likely to achieve clinically meaningful outcomes.

When learning the anterior approach THA (AA-THA), the risk of an early complication doubles. Little is known about the intraoperative processes which may contribute to these problems. The surgical team may also influence surgical error. This study uses Human Reliability Analysis (HRA), a technique used to prevent disaster in aviation and nuclear power, to investigate technical error in AA-THA. Sixty-two primary AA-THA operations were professionally filmed in 61 participants. 6 surgeons (3 experts, 3 learners) performed the operations. An expert derived task sequence analysis identified 10 operative phases. Video footage was analysed, and errors were classified using 10 error modes. Hazard zones, where errors occurred more frequently, were identified. Radiological errors were identified analysing postoperative radiographs. Performance of the wider surgical team was assessed using the NOTECHS II score and correlated with total error. Participants were followed up at 6 weeks to document adverse events. Mean age was 68.5 ± 9.2 years old, mean BMI 27.0 ± 5.6. 44 female, 17 male. 853 errors were recorded across 62 surgeries with a mean 13.8 ± 8.4 errors per case. Most errors happened during acetabular reaming (4.1 ± 4.0 per case), followed by femoral broaching (2.9 ± 3.3) and deep exposure of the hip (2.0 ± 1.6). Learning surgeons made double the total errors compared to experts (25 ± 11.4 vs 12.3 ± 6.9, p = 0.002). Three patients (4.9%) had post-operative femoral fractures. 35.5% of acetabular components were outside the safe zone, 27.4% had a leg length discrepancy > 5mm. As NOTECHs II score increased technical error reduced (Rs = -0.45 (95% CI: -0.64 to -0.21), p <0.001). Strongest subgroup correlations were the scrub nurse (Rs = -0.42 (95% CI: -0.62 to -0.18), p <0.001) and leadership subdomain (Rs = -0.46 (95% CI: -0.64 to -0.23), p <0.001). This study validates the HRA process. Hazard zones were: (1) Acetabular preparation, (2) Femoral broaching and (3) Deep exposure of the hip. Superior teamwork is associated with reduced intraoperative error. This detailed understanding of intraoperative error could guide interventions to reduce patient harm during the learning curve.

Background The minimal clinically important difference, minimal important change, minimal detectable change and patient-acceptable symptom state are poorly defined for the Oxford Shoulder Score following shoulder arthroplasty. The study's aim was to calculate their values. Methods One hundred patients underwent shoulder arthroplasty and completed pre and 1-year postoperative Oxford Shoulder Score. Patient satisfaction was assessed at 1-year using a visual analogue scale from 0 to 100: ‘very satisfied’ (>80), ‘satisfied’ (>60–80), and ‘unsatisfied’ (≤60). The difference between patients recording ‘unsatisfied’ ( n = 11) and ‘satisfied’ ( n = 16) was used to define the minimal clinically important difference. MIC cohort was calculated as the change in Oxford Shoulder Score for those satisfied (>60). Receiver-operating characteristic curve analysis was used to determine the MIC individual and patient-acceptable symptom state. Distribution-based methodology was used for the minimal detectable change. Results The minimal clinically important difference was 6.9 (95% confidence interval 0.7–13.1, p = 0.039). The MIC cohort was 11.6 (95% confidence interval 6.8–16.4) and MIC individual 13. The minimal detectable change was 6.6 and the patient-acceptable symptom state was defined as ≥29. Discussion The minimal clinically important difference and minimal important change can assess whether there is a clinical difference between two groups and whether a cohort/patient has had a meaningful change in their Oxford Shoulder Score, respectively. These were greater than measurement error (minimal detectable change), suggesting a real change. The patient-acceptable symptom state can be used as a marker of achieving satisfaction.

Objective: The objective of this study was to elucidate the role of Calcium calmodulin-dependent Kinase II (CaMKII) in articular chondrocytes and its involvement in osteoarthritis (OA) pathogenesis. By performing gain and loss of function experiments, the research aimed to determine how CaMKII modulates chondrocyte metabolism, anabolic and catabolic processes, hypertrophic differentiation, and autophagy within the articular cartilage. Design: Articular cartilage was harvested from patients undergoing joint replacement surgery for OA, and adult human articular chondrocytes (AHACs) were isolated and cultured. Recombinant adenoviruses were used to overexpress a constitutively active form of CaMKIIγ (AdCaMKII) or inhibit CaMKII activity (AdAIP). Various assays, including RT-PCR analysis, alcian blue staining of Micromass cultures, immunofluorescence, and Western blotting, were performed to assess the effects of CaMKII modulation on chondrocyte function. Results: Overexpression of activated CaMKIIγ promoted anabolism, evidenced by increased expression of SOX9, COL2A1, and ACAN, and decreased MMP-13 levels. It also enhanced proteoglycan content in AHAC micromass cultures. Furthermore, CaMKII counteracted the catabolic effects of IL-1β and preserved proteoglycan content. We also observed decreased chondrocyte proliferation and increased synthesis of hypertrophic marker Type X Collagen. CaMKII activation was found to induce autophagy, as indicated by increased phosphorylation of Beclin1 and decreased p62 expression. The anabolic effects of CaMKII were dependent on autophagy, as inhibition of autophagy with Bafilomycin prevented the CaMKII-induced increase in glycosaminoglycan content. Conclusions: CaMKII plays a significant role in modulating chondrocyte metabolism and maintaining cartilage homeostasis. It promotes anabolic processes, counteracts catabolic stimuli, and induces autophagy in articular chondrocytes. However, it also promotes hypertrophic differentiation, highlighting the complexity of CaMKII-mediated signalling in cartilage. Understanding these pathways could lead to new therapeutic strategies that leverage CaMKII's anabolic potential while mitigating its pro-degenerative effects. ### Competing Interest Statement The authors have declared no competing interest.

Using data from two ED. departments of 773 patients admitted with SARS-CoV-2, ICD-10 codes derived from the General Practitioner - Summary Care Record (GP-SCR) and Emergency Department (ED.) records were analysed for code discrepancies and whether this related to increased mortality. The average number of ICD-10 codes in both GP-SCR and ED. records was higher for patients who died than patients who survived (all p < .0001). Pre-existing GP digital data provides a better prediction of mortality than data collected manually during admission clerking in the ED. Up to 78.47% of GP-SCR codes were missed in the ED. records and up to 45.49% of the ED. record codes were not in the GP-SCR. A subset of missed ICD-10 codes were identified as being able to predict outcome; a trend towards increasing death rate as the proportion of missed codes increases. Initiatives to make the GP-SCR available to the wider healthcare community should improve patient care and reduce bias during development of machine learning based algorithms.

Background Difficulty kneeling following total knee arthroplasty (TKA) remains highly prevalent, and has cultural, social, and occupational implications. With no clear evidence of superiority, whether or not to resurface the patella remains debatable. This systematic review examined whether resurfacing the patella (PR) or not (NPR) influences kneeling ability following TKA. Methods This systematic review was conducted by following PRISMA guidelines. Three electronic databases were searched utilizing a search strategy developed with the aid of a department librarian. Study quality was assessed using MINROS criteria. Article screening, methodological quality assessment and data extraction were performed by two independent authors, and a third senior author was consulted if consensus was not reached. Results A total of 459 records were identified, with eight studies included in the final analysis, and all deemed to be level III evidence. The average MINORS score was 16.5 for comparative studies and 10.5 for non-comparative studies. The total number of patients was 24,342, with a mean age of 67.6 years. Kneeling ability was predominantly measured as a patient-reported outcome measure (PROM), with two studies also including an objective assessment. Two studies demonstrated a statistically significant link between PR and kneeling, with one demonstrating improved kneeling ability with PR and the other reporting the opposite. Other potential factors associated with kneeling included gender, postoperative flexion, and body mass index (BMI). Re-operation rates were significantly higher in the NPR cohort whereas PR cohorts had higher Feller scores, patient-reported limp and patellar apprehension. Conclusion Despite its importance to patients, kneeling remains not only under-reported but also ill-defined in the literature, with no clear consensus regarding the optimum outcome assessment tool. Conflicting evidence remains as to whether PR influences kneeling ability, and to clarify the situation, large prospective randomized studies are required.

Over 8000 total hip arthroplasties (THA) in the UK were revised in 2019, half for aseptic loosening. It is believed that Artificial Intelligence (AI) could identify or predict failing THA and result in early recognition of poorly performing implants and reduce patient suffering. The aim of this study is to investigate whether Artificial Intelligence based machine learning (ML) / Deep Learning (DL) techniques can train an algorithm to identify and/or predict failing uncemented THA. Consent was sought from patients followed up in a single design, uncemented THA implant surveillance study (2010–2021). Oxford hip scores and radiographs were collected at yearly intervals. Radiographs were analysed by 3 observers for presence of markers of implant loosening/failure: periprosthetic lucency, cortical hypertrophy, and pedestal formation. DL using the RGB ResNet 18 model, with images entered chronologically, was trained according to revision status and radiographic features. Data augmentation and cross validation were used to increase the available training data, reduce bias, and improve verification of results. 184 patients consented to inclusion. 6 (3.2%) patients were revised for aseptic loosening. 2097 radiographs were analysed: 21 (11.4%) patients had three radiographic features of failure. 166 patients were used for ML algorithm testing of 3 scenarios to detect those who were revised. 1) The use of revision as an end point was associated with increased variability in accuracy. The area under the curve (AUC) was 23–97%. 2) Using 2/3 radiographic features associated with failure was associated with improved results, AUC: 75–100%. 3) Using 3/3 radiographic features, had less variability, reduced AUC of 73%, but 5/6 patients who had been revised were identified (total 66 identified). The best algorithm identified the greatest number of revised hips (5/6), predicting failure 2–8 years before revision, before all radiographic features were visible and before a significant fall in the Oxford Hip score. True-Positive: 0.77, False Positive: 0.29. ML algorithms can identify failing THA before visible features on radiographs or before PROM scores deteriorate. This is an important finding that could identify failing THA early.

AIM:There are 12 signs of the zodiac, each attributed with its own specific personality traits, desires and attitudes. The aim of the study was to evaluate the effect of zodiac sign on patient-reported outcome measures (PROMS) following primary total knee arthroplasty (TKA). METHOD:Patients undergoing primary TKA during a 2-year period (January 2019 to December 2020) were identified retrospectively. Patient demographics, Oxford Knee Score (OKS), EuroQol 5-dimension (EQ-5D) (baseline, 1 and 2 years) and patient satisfaction scores (1 and 2 years) were collected. Each patient's zodiac sign was assigned from their date of birth. RESULTS:There were 509 patients (228 males [44.8%] and 281 females [55.2%]) with a mean age of 70.9 years and a mean BMI of 30.3. There were no significant differences in gender (p=0.712), age (p=0.088), BMI (p=0.660), or pre-operative OKS (p=0.539). Aries and Gemini (0.366) had the worst and Pisces the best (0.595) pre-operative EQ-5D scores (p=0.038). When adjusting for confounding, Aries (p=0.031) had a greater improvement in EQ-5D at 1 year, although this was not maintained at 2 years. When adjusting for confounding, Pisceans had significantly less of an improvement in OKS at both 1 (p=0.022) and 2 years (p=0.042) and also had a significantly lower risk of satisfaction at 2 years (odds ratio 0.41, p=0.043). CONCLUSION:Zodiac sign was associated with outcome following TKA. Pisceans had the best pre-operative EQ-5D scores, but the least improvement in the post-operative joint specific score (OKS) and were less likely to be satisfied, despite achieving an equal improvement in their health-related quality of life (EQ-5D). Aries started with the lowest pre-operative EQ-5D scores but achieved the best scores at 1 year. Our study shows that an individual's zodiac sign may serve as a useful predictive factor for functional outcomes and satisfaction following TKA. However, our findings are the result of multiple testing in a large dataset following a data trawl, and correlation does not necessarily equal causation even in a real-world registry.

BACKGROUND:Surgical waiting lists have risen dramatically across the UK as a result of the COVID-19 pandemic. The effective use of operating theatres by optimal scheduling could help mitigate this, but this requires accurate case duration predictions. Current standards for predicting the duration of surgery are inaccurate. Artificial intelligence (AI) offers the potential for greater accuracy in predicting surgical case duration. This study aimed to investigate whether there is evidence to support that AI is more accurate than current industry standards at predicting surgical case duration, with a secondary aim of analysing whether the implementation of the models used produced efficiency savings. METHOD:PubMed, Embase, and MEDLINE libraries were searched through to July 2023 to identify appropriate articles. PRISMA extension for scoping reviews and the Arksey and O'Malley framework were followed. Study quality was assessed using a modified version of the reporting guidelines for surgical AI papers by Farrow et al. Algorithm performance was reported using evaluation metrics. RESULTS:The search identified 2593 articles: 14 were suitable for inclusion and 13 reported on the accuracy of AI algorithms against industry standards, with seven demonstrating a statistically significant improvement in prediction accuracy (P < 0.05). The larger studies demonstrated the superiority of neural networks over other machine learning techniques. Efficiency savings were identified in a RCT. Significant methodological limitations were identified across most studies. CONCLUSION:The studies suggest that machine learning and deep learning models are more accurate at predicting the duration of surgery; however, further research is required to determine the best way to implement this technology.

AbstractIntroduction30-day emergency readmission is an indicator of treatment related complication once discharged, resulting in readmission. A board-approved quality improvement pathway was introduced to reduce elective re-admissions.MethodThe pathway involved telephone and email contact details provision to patients for any non-life threatening medical assistance, allowing for initial nurse led management of all issues. A new clinic room available 7 days, and same day ultrasound scanning for DVT studies were introduced. A capability, opportunity and behavior model of change was implemented.Readmission rates before and six months after implementation were collected from Model Hospital. A database used to document patient communications was interrogated for patient outcomes.ResultsPrior to implementation, readmission rates following elective primary total knee replacement (TKR) at the 1st business quarter of 2021 (April – June 2021), was 8.7%, (benchmark 3.8%). Following implementation, readmission rates decreased to 4.1% (October – December 2021). 54% of patients making contact were managed with telephone advice. 15% of patients required face-to-face clinic. 32% of those required a same day scan to exclude DVT (1/4).20 out of 684 TKRs performed following protocol introduction were re-admitted within 30 days. Readmissions were 41% surgical, 29% medical. 52% were unaware of the newly implemented protocol. Further improvements have been made to the protocol based on these findings.Implementation of a suitable pathway can significantly reduce re-admission rates in our center and could be used to reduce readmission rates in other national elective treatment centers.

Background: There is a trend towards minimising length of stay (LOS) after total knee arthroplasty (TKA), as greater LOS is associated with poorer outcomes and higher costs. Patient factors known to influence LOS post-TKA include age and ASA grade. Evidence regarding the effect of body mass index (BMI) in particular is conflicting, with some studies finding that increased BMI predicts increased LOS, while others have found no relationship. Few previous studies, which have mostly been conducted outside the UK, have examined the effect of living alone or socioeconomic deprivation, which may be confounders.Methods: We conducted a retrospective cohort study of 1031 consecutive primary TKAs performed between 1 April 2021 and 31 December 2021 in a single high-volume arthro-plasty centre. A multivariable negative binomial regression model was performed for the 860 patients with complete data, using pre-operative (BMI, age, gender, ASA grade, smok-ing, ethnicity, socioeconomic deprivation, living arrangement, EQ5D quality of life score, and indication for surgery) and peri-operative variables (surgeon, surgical approach, tourniquet use, a.m./p.m operation, operation side, duration, and day of the week).Results: Mean LOS was 2.6 days. BMI and socioeconomic deprivation had no effect on LOS (P > 0.05). Increased LOS was associated with living alone, lower EQ5D, age and ASA grade (all P < 0.001), p.m. operation (P < 0.01), female gender and duration of surgery (P < 0.05).Conclusion: BMI and socioeconomic status were not correlated with LOS after TKA. Living alone, which has not been previously reported and lower pre-operative EQ5D status were significant risk factors, which merit consideration in pre-operative planning and counselling.(c) 2023 Elsevier B.V. All rights reserved.

End-stage knee arthropathy is a recognised complication of haemophilia. It is often treated by total knee arthroplasty (TKA), which is more technically challenging in patients with haemophilia (PwH). It remains unclear what factors may predict implant survivorship and deep infection rate. Therefore, we systematically review the evidence regarding TKA survivorship and infection in PwH, compared to the general population, and determine the important factors influencing survivorship, particularly HIV and CD4 + count. A systematic literature review was conducted using MEDLINE, EMBASE, and PubMed for studies reporting Kaplan–Meier survivorship for TKA in PwH (PROSPERO CRD42021284644). Meta-analysis was performed for survivorship, and the results compared to < 55-year-olds from the National Joint Registry (NJR). Meta-regression was performed to determine the impact of relevant variables on 10-year survivorship, with a sub-analysis focusing on HIV. Twenty-one studies were reviewed, totalling 1338 TKAs (average age 39 years). Implant survivorship for PwH at 5, 10, and 15 years was 94

There are advocates of both two-dimensional (2D) and three-dimensional (3D) templating methods for planning total hip replacement. The aim of this study was to compare the accuracy of implant size prediction when using 2D and 3D templating methods for total hip arthroplasty, as well as to compare the inter- and intra-observer reliability in order to determine whether currently available methods are sufficiently reliable and reproducible. Medline, EMBASE and PubMed were searched to identify studies that compared the accuracy of 2D and 3D templating for total hip replacement. Results were screened using the PRISMA flowchart and included studies were assessed for their level of evidence using the Oxford CEBM criteria. Non-randomized trials were critically appraised using the MINORS tool, whilst randomized trials were assessed using the CASP RCT checklist. A series of meta-analyses of the data for accuracy were also conducted. Ten studies reported that 3D templating is an accurate and reliable method of templating for total hip replacement. Six studies compared 3D templating with 2D templating, all of which concluded that 3D templating was more accurate, with three finding a statistically significant difference. The meta-analyses showed that 3D CT templating is the most accurate method. This review supports the hypothesis that 3D templating is an accurate and reliable method of preoperative planning, which is more accurate than 2D templating for predicting implant size. However, further research is needed to ascertain the significance of this improved accuracy and whether it will yield any clinical benefit.

Abstract Introduction Cell-based therapies using lipoaspirate are gaining popularity within surgical fields due to their hypothesised regenerative potential. Several point-of-care lipoaspirate-processing devices have become available to isolate cells for therapeutic use, with published evidence reporting their clinical relevance. However, few studies have analysed the composition of their minimally manipulated cellular products, information that is vital to understand the mechanisms by which these therapies may be efficacious. This review aimed to identify devices using mechanical-only processing of lipoaspirate, their cell yields, viability, phenotype and where available clinical outcomes. Methods MEDLINE, Embase and PubMed databases were systematically searched on 01/09/21 using relevant keywords. PRISMA guidelines were followed (PROSPERO#CRD42021282041), and level of evidence was assessed using the Oxford Centre for Evidence-Based Medicine guidelines. Information was extracted and analysed to summarise the cellular composition derived from these devices and their clinical outcomes. Results 2895 studies were screened and a total of 15 articles (11=Level 5 evidence) fulfilled the inclusion criteria. Overall, 13 devices were identified. All the studies reported cell yield for their devices (range 0.005–21×106). 10 reported viability (range 60–98%), 11 performed immuno-phenotypic analysis of the cell-subtypes and 4 investigated clinical outcomes of their cellular products. Only 2 studies reported all four parameters. Conclusion Although many devices are available to mechanically process lipoaspirate, few have published peer-reviewed literature of their products’ composition. Significant heterogeneity in the reporting of these studies makes it difficult to summarise their clinical potential. This review is unable to make any recommendations on the clinical use of these devices. Take-home message Many point-of-care devices have become available to mechanically process lipoaspirate to form a cellular product for therapeutic use in surgery. However, few have publications on their products’ composition which is information required to understand the mechanisms by which these therapies may be efficacious.

Purpose Cell-based therapies using lipoaspirate are gaining popularity in orthopaedics due to their hypothesised regenerative potential. Several ‘point-of-care’ lipoaspirate-processing devices/systems have become available to isolate cells for therapeutic use, with published evidence reporting their clinical relevance. However, few studies have analysed the composition of their ‘minimally-manipulated’ cellular products in parallel, information that is vital to understand the mechanisms by which these therapies may be efficacious. This scoping review aimed to identify devices/systems using mechanical-only processing of lipoaspirate, the constituents of their cell-based therapies and where available, clinical outcomes. Methods PRISMA extension for scoping reviews guidelines were followed. MEDLINE, Embase and PubMed databases were systematically searched to identify relevant articles until 21 st April 2022. Information relating to cellular composition and clinical outcomes for devices/systems was extracted. Further information was also obtained by individually searching the devices/systems in the PubMed database, Google search engine and contacting manufacturers. Results 2895 studies were screened and a total of 15 articles (11 = Level 5 evidence) fulfilled the inclusion criteria. 13 unique devices/systems were identified from included studies. All the studies reported cell concentration (cell number regardless of phenotype per millilitre of lipoaspirate) for their devices/systems (range 0.005–21 × 10 6 ). Ten reported cell viability (the measure of live cells- range 60–98%), 11 performed immuno-phenotypic analysis of the cell-subtypes and four investigated clinical outcomes of their cellular products. Only two studies reported all four of these parameters. Conclusion When focussing on cell concentration, cell viability and MSC immuno-phenotypic analysis alone, the most effective manual devices/systems were ones using filtration and cutting/mincing. However, it was unclear whether high performance in these categories would translate to improved clinical outcomes. Due to the lack of standardisation and heterogeneity of the data, it was also not possible to draw any reliable conclusions and determine the role of these devices/systems in clinical practice at present. Level of Evidence Level V Therapeutic.

The aim was to identify independent preoperative factors associated with changes in health-related quality of life (HRQoL) following total knee arthroplasty (TKA), and whether these could be used as indicators for surgery. A retrospective study of 3127 TKA patients was undertaken that included 1194 (38.2 ASA grade 3, grade II obesity, a better preoperative EQ-5D or OKS were independently associated with a lesser improvement in HRQoL. The thresholds identified in the EQ-5D or OKS for a clinically significant improvement in HRQoL may be used as potential indicators for referral for TKA. Patella resurfacing was not independently associated with a clinically important improvement in HRQoL. Retrospective diagnostic study, Level III.

Abstract Introduction Total hip and knee arthroplasty are common orthopaedic procedures that require post-operative radiographs to confirm implant positioning and identify complications. Artificial intelligence (AI) technology has the potential to automate image analysis. This systematic review reports on how AI-based technologies are currently being used and their accuracy in image analysis following THA and TKA. Methods EMBASE, Medline and PubMed libraries were systematically searched for articles published until 15/09/2021 using terms related to “x-ray analysis”, “total hip/knee arthroplasty”, and “AI”. The review was performed according to the PRISMA guidelines (PROSPERO#CRD42021276876). Study quality was assessed using a modified MINORS tool. AI performance was reported using the area under the curve (AUC) and accuracy. Results Of the 455 studies identified, 12 were included: nine reported implant identification, three described prediction of implant failure and three compared AI performance with orthopaedic surgeons. AI-based implant identification was precise (AUC 0.992–1) and most algorithms reported accuracy >90%. Two of these studies reported AI performance to be similar or superior to human experts. AI prediction of dislocation risk following THA was acceptable (AUC 76.67), diagnosis of hip implant loosening was good (accuracy 88.3%) and measurement of acetabular angles on post-operative x-rays was comparable to humans (Cohen's kappa 0.76–1.00). Conclusion AI technology can be trained to identify implant models on post-operative x-rays with a performance that is comparable to that of human experts. However, the technology requires further development to enable analysis of other post-operative radiographic features following arthroplasty surgery that could improve patient care. Take-home message Artificial intelligence image analysis through deep learning can classify hip and knee implants, and measure malposition and detect features of loosening of hip implants.

AIMS:Total hip arthroplasty (THA) and total knee arthroplasty (TKA) are common orthopaedic procedures requiring postoperative radiographs to confirm implant positioning and identify complications. Artificial intelligence (AI)-based image analysis has the potential to automate this postoperative surveillance. The aim of this study was to prepare a scoping review to investigate how AI is being used in the analysis of radiographs following THA and TKA, and how accurate these tools are. METHODS:The Embase, MEDLINE, and PubMed libraries were systematically searched to identify relevant articles. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for scoping reviews and Arksey and O'Malley framework were followed. Study quality was assessed using a modified Methodological Index for Non-Randomized Studies tool. AI performance was reported using either the area under the curve (AUC) or accuracy. RESULTS:Of the 455 studies identified, only 12 were suitable for inclusion. Nine reported implant identification and three described predicting risk of implant failure. Of the 12, three studies compared AI performance with orthopaedic surgeons. AI-based implant identification achieved AUC 0.992 to 1, and most algorithms reported an accuracy > 90%, using 550 to 320,000 training radiographs. AI prediction of dislocation risk post-THA, determined after five-year follow-up, was satisfactory (AUC 76.67; 8,500 training radiographs). Diagnosis of hip implant loosening was good (accuracy 88.3%; 420 training radiographs) and measurement of postoperative acetabular angles was comparable to humans (mean absolute difference 1.35° to 1.39°). However, 11 of the 12 studies had several methodological limitations introducing a high risk of bias. None of the studies were externally validated. CONCLUSION:These studies show that AI is promising. While it already has the ability to analyze images with significant precision, there is currently insufficient high-level evidence to support its widespread clinical use. Further research to design robust studies that follow standard reporting guidelines should be encouraged to develop AI models that could be easily translated into real-world conditions. Cite this article: Bone Joint J 2022;104-B(8):929-937.