Sarah R. Kingsbury
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sarah R. Kingsbury has a total of 243 co-authors. Here are the top co-authors by number of shared publications:
- Philip Gerard Conaghan
Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds & NIHR Leeds Biomedical Research Centre, Leeds, UK
30 shared publications - Fiona E. Watt
Department of Immunology and Inflammation, Imperial College London , London
6 shared publications - Andrew J. Grainger
Cambridge University Hospitals
5 shared publications - Tonia L. Vincent
Centre for OA Pathogenesis, Kennedy Institute of Rheumatology, University of Oxford
5 shared publications - David J. Torgerson
York Trials Unit, University of York, York, UK
4 shared publications - Farag Shuweihdi
Leeds Institute of Health Sciences, University of Leeds, Leeds, UK
4 shared publications
Latest Publications
- 28 August 2026
- Published by Oxford University Press (OUP) in Rheumatology
- Vol. 65 (9)
DNA hypo-methylation of the TNFA gene predicts response to MTX in RA but not in PsA
Objectives: The ‘Treat-to-Target’ approach significantly improves treatment outcomes in RA and PsA (RA/PsA). MTX is commonly used as a first-line treatment upon diagnosis. However, 50–70% of treated patients do not achieve clinical remission after 6 months. We previously developed a quantitative methylation-sensitive qPCR (qMSP) assay to quantify DNA-methylation levels of the TNFA gene. We demonstrated its predictive value for RA diagnosis. Here, we assessed the utility of this qMSP assay for predicting MTX treatment outcome and investigated changes in DNA-methylation levels after 6 months of MTX. Methods: Early untreated RA (n = 201) and PsA (n = 77) were included with healthy controls (n = 39). The TNFA-qMSP assay was performed by qPCR on bisulphite-converted DNA extracted from frozen peripheral blood mononuclear cells. The outcome predictive-value of DNA-methylation levels was analysed using binary logistic regression and area under the curve (AUC). Results: Pre-treatment DNA-methylation levels of the TNFA gene were reduced in RA (P = 1.8 × 10−13) and PsA (P = 0.007) compared with controls. Higher DNA-methylation levels were associated with MTX-induced remission in RA (AUC = 0.721) and substantially improved the performance of a predictive model when combined with clinical data (AUC = 0.760) compared with clinical data only (AUC = 0.699). Levels of methylation reduced over time, irrespective of response to MTX in RA. In PsA, DNA-methylation levels were not associated with treatment outcome. Conclusion: The TNF qMSP assay has potential clinical utility as a biomarker for predicting MTX-induced remission specifically in early RA, but not in PsA. However, MTX does not appear to prevent the reduction of DNA-methylation levels at the TNFA loci over time.
- 28 August 2026
- Published by Oxford University Press (OUP) in Rheumatology
- Vol. 65 (9)
DNA hypo-methylation of the TNFA gene predicts response to MTX in RA but not in PsA
Objectives: The ‘Treat-to-Target’ approach significantly improves treatment outcomes in RA and PsA (RA/PsA). MTX is commonly used as a first-line treatment upon diagnosis. However, 50–70% of treated patients do not achieve clinical remission after 6 months. We previously developed a quantitative methylation-sensitive qPCR (qMSP) assay to quantify DNA-methylation levels of the TNFA gene. We demonstrated its predictive value for RA diagnosis. Here, we assessed the utility of this qMSP assay for predicting MTX treatment outcome and investigated changes in DNA-methylation levels after 6 months of MTX. Methods: Early untreated RA (n = 201) and PsA (n = 77) were included with healthy controls (n = 39). The TNFA-qMSP assay was performed by qPCR on bisulphite-converted DNA extracted from frozen peripheral blood mononuclear cells. The outcome predictive-value of DNA-methylation levels was analysed using binary logistic regression and area under the curve (AUC). Results: Pre-treatment DNA-methylation levels of the TNFA gene were reduced in RA (P = 1.8 × 10−13) and PsA (P = 0.007) compared with controls. Higher DNA-methylation levels were associated with MTX-induced remission in RA (AUC = 0.721) and substantially improved the performance of a predictive model when combined with clinical data (AUC = 0.760) compared with clinical data only (AUC = 0.699). Levels of methylation reduced over time, irrespective of response to MTX in RA. In PsA, DNA-methylation levels were not associated with treatment outcome. Conclusion: The TNF qMSP assay has potential clinical utility as a biomarker for predicting MTX-induced remission specifically in early RA, but not in PsA. However, MTX does not appear to prevent the reduction of DNA-methylation levels at the TNFA loci over time.
- Open Access
- 8 July 2026
- Published by Elsevier in Future Healthcare Journal
- Vol. 13 (3)
Creating a service system architecture for people living with multiple long-term conditions
- 8 July 2026
- Published by Elsevier in Future Healthcare Journal
- Vol. 13 (3)
Open AccessCreating a service system architecture for people living with multiple long-term conditions
- Open Access
- 7 July 2026
- Published by Elsevier in Osteoarthritis and Cartilage
- Vol. 34 (10) , 1539-1549
Systemic Biomarkers of Treatment Response to Methotrexate in People with Painful Knee Osteoarthritis: a Biological Substudy of the PROMOTE Randomised Controlled Clinical Trial
- 7 July 2026
- Published by Elsevier in Osteoarthritis and Cartilage
- Vol. 34 (10) , 1539-1549
Open AccessSystemic Biomarkers of Treatment Response to Methotrexate in People with Painful Knee Osteoarthritis: a Biological Substudy of the PROMOTE Randomised Controlled Clinical Trial
- Open Access
Musculoskeletal Pain and Neuropathic Pain Features Among Individuals with Long COVID and a History of Pain: Characteristics and Impact on Health-Related Quality of Life
Background: Post-COVID-19 syndrome or Long COVID (LC) is a multi-system condition characterised by persistent symptoms beyond three months following SARS-CoV-2 infection. Pain, particularly musculoskeletal and neuropathic pain, is commonly reported and may substantially affect functioning and health-related quality of life (HRQoL). However, there is limited detailed characterisation of pain phenotypes and their differential impact on outcomes in LC. This study aimed to (i) characterise pain associated with LC, (ii) identify factors associated with pain type and severity, and (iii) examine its impact on HRQoL, physical activity, and work. Methods: A cross-sectional survey was conducted among adults with clinician-diagnosed or self-reported LC who reported current or previous joint, muscle, or nerve pain. Participants completed validated patient-reported outcome measures including the C19-YRS, Nordic Musculoskeletal Questionnaire, Pain Detect Questionnaire (PD-Q), Musculoskeletal Health Questionnaire, and EQ-5D-5L. Participants were classified into neuropathic, nociceptive, or ambiguous pain groups based on PD-Q scores. Descriptive analyses, group comparisons, and backward stepwise ordinary least squares regression were used to examine factors associated with overall pain (PD-Q Total). Results: Among 198 participants (73% female; mean age 52 ± 12.7 years; mean LC duration 2.4 years), 96% reported new or worsened pain following COVID-19 infection, with muscle pain (87%), joint pain (84%), and nerve pain (66%) most frequently reported. PD-Q scores suggestive of neuropathic pain were identified in 28% of participants and were associated with significantly greater pain intensity, symptom burden, functional impairment, and lower HRQoL (p < 0.05). These participants reported higher average pain scores across all body regions and greater radiating pain (76%) compared to other groups. Functional disability and anxiety were significant factors associated with pain severity. Conclusions: Pain is highly prevalent in LC, with neuropathic pain representing a distinct and more severe phenotype associated with substantial functional and quality-of-life impacts. This study provides novel comparative evidence on pain phenotypes and their differential burden, highlighting the need for targeted pain assessment and management strategies in LC.
Open AccessMusculoskeletal Pain and Neuropathic Pain Features Among Individuals with Long COVID and a History of Pain: Characteristics and Impact on Health-Related Quality of Life
Background: Post-COVID-19 syndrome or Long COVID (LC) is a multi-system condition characterised by persistent symptoms beyond three months following SARS-CoV-2 infection. Pain, particularly musculoskeletal and neuropathic pain, is commonly reported and may substantially affect functioning and health-related quality of life (HRQoL). However, there is limited detailed characterisation of pain phenotypes and their differential impact on outcomes in LC. This study aimed to (i) characterise pain associated with LC, (ii) identify factors associated with pain type and severity, and (iii) examine its impact on HRQoL, physical activity, and work. Methods: A cross-sectional survey was conducted among adults with clinician-diagnosed or self-reported LC who reported current or previous joint, muscle, or nerve pain. Participants completed validated patient-reported outcome measures including the C19-YRS, Nordic Musculoskeletal Questionnaire, Pain Detect Questionnaire (PD-Q), Musculoskeletal Health Questionnaire, and EQ-5D-5L. Participants were classified into neuropathic, nociceptive, or ambiguous pain groups based on PD-Q scores. Descriptive analyses, group comparisons, and backward stepwise ordinary least squares regression were used to examine factors associated with overall pain (PD-Q Total). Results: Among 198 participants (73% female; mean age 52 ± 12.7 years; mean LC duration 2.4 years), 96% reported new or worsened pain following COVID-19 infection, with muscle pain (87%), joint pain (84%), and nerve pain (66%) most frequently reported. PD-Q scores suggestive of neuropathic pain were identified in 28% of participants and were associated with significantly greater pain intensity, symptom burden, functional impairment, and lower HRQoL (p < 0.05). These participants reported higher average pain scores across all body regions and greater radiating pain (76%) compared to other groups. Functional disability and anxiety were significant factors associated with pain severity. Conclusions: Pain is highly prevalent in LC, with neuropathic pain representing a distinct and more severe phenotype associated with substantial functional and quality-of-life impacts. This study provides novel comparative evidence on pain phenotypes and their differential burden, highlighting the need for targeted pain assessment and management strategies in LC.
- Open Access
- 6 March 2026
- Published by Springer Nature in Clinical Rheumatology
- Vol. 45 (5) , 2837-2844
The relationship between real-time ultrasound-detected extra-articular soft tissue pathologies and knee pain in osteoarthritis: a cross-sectional study
Objective: Extra-articular soft tissue pathologies are relatively understudied in osteoarthritis (OA). We aimed to explore the relationship between structural pathology, sites of knee pain and knee pain severity, as determined using real-time ultrasound (US). Methods: Patients with painful knee OA received an US scan assessing a range of soft tissue structures at four quadrants (anterior, medial, lateral and posterior). US probe site tenderness (sono-palpation) was also assessed at each structure. Participants’ reported outcome measures were recorded, including sites of knee pain using the Knee Pain Map (using the same quadrant locations as structural pathology) and average numeric rating scale (NRS) for pain severity in the past week. Results: Fifty-four knees (26 men, BMI 31.5, mean NRS pain 5.4, 96% Kellgren–Lawrence 2/3) were included. Medial pain was the most frequently reported location. Semimembranosus tendon (SMT) hypoechogenicity, indicating tendinopathy, was the most frequent US finding (53%). However, there was no clear pattern between SMT pathology and knee pain location. Medial collateral ligament (MCL) and lateral collateral ligament (LCL) sono-palpation (reflecting clinical enthesitis) was more frequently observed in medial and lateral knee pain, respectively (43.2% and 52.9%). Higher NRS pain was significantly associated with MCL sono-palpation (1.41 NRS points; CI 0.47–2.36, p = 0.004) and Baker’s cyst sono-palpation (1.78 NRS points; CI 0.35–3.20, p = 0.016). A trend toward greater NRS pain was associated with US-detected patellar tendon pathology (1.1 NRS points; CI − 0.14–2.35, p = 0.08). Conclusion: This exploratory study highlighted the potential importance of extra-articular knee pathologies, particularly the collateral ligaments, in understanding OA knee pain location and severity. Key Points • This study measured the potential role of extra-articular pathology in OA knee pain. • Sono-palpation has been used as a novel method of measuring enthesitis. • Medial and lateral collateral ligament enthesitis may contribute to respective joint pain locations in knee OA. • Enthesitis may be secondary to biomechanical stresses around the knee joint.
- 6 March 2026
- Published by Springer Nature in Clinical Rheumatology
- Vol. 45 (5) , 2837-2844
Open AccessThe relationship between real-time ultrasound-detected extra-articular soft tissue pathologies and knee pain in osteoarthritis: a cross-sectional study
Objective: Extra-articular soft tissue pathologies are relatively understudied in osteoarthritis (OA). We aimed to explore the relationship between structural pathology, sites of knee pain and knee pain severity, as determined using real-time ultrasound (US). Methods: Patients with painful knee OA received an US scan assessing a range of soft tissue structures at four quadrants (anterior, medial, lateral and posterior). US probe site tenderness (sono-palpation) was also assessed at each structure. Participants’ reported outcome measures were recorded, including sites of knee pain using the Knee Pain Map (using the same quadrant locations as structural pathology) and average numeric rating scale (NRS) for pain severity in the past week. Results: Fifty-four knees (26 men, BMI 31.5, mean NRS pain 5.4, 96% Kellgren–Lawrence 2/3) were included. Medial pain was the most frequently reported location. Semimembranosus tendon (SMT) hypoechogenicity, indicating tendinopathy, was the most frequent US finding (53%). However, there was no clear pattern between SMT pathology and knee pain location. Medial collateral ligament (MCL) and lateral collateral ligament (LCL) sono-palpation (reflecting clinical enthesitis) was more frequently observed in medial and lateral knee pain, respectively (43.2% and 52.9%). Higher NRS pain was significantly associated with MCL sono-palpation (1.41 NRS points; CI 0.47–2.36, p = 0.004) and Baker’s cyst sono-palpation (1.78 NRS points; CI 0.35–3.20, p = 0.016). A trend toward greater NRS pain was associated with US-detected patellar tendon pathology (1.1 NRS points; CI − 0.14–2.35, p = 0.08). Conclusion: This exploratory study highlighted the potential importance of extra-articular knee pathologies, particularly the collateral ligaments, in understanding OA knee pain location and severity. Key Points • This study measured the potential role of extra-articular pathology in OA knee pain. • Sono-palpation has been used as a novel method of measuring enthesitis. • Medial and lateral collateral ligament enthesitis may contribute to respective joint pain locations in knee OA. • Enthesitis may be secondary to biomechanical stresses around the knee joint.
- 22 January 2026
- Published by Springer Nature in Drugs & Aging
- Vol. 43 (2) , 123-136
Methotrexate for Osteoarthritis: What Does the Evidence Say?
Osteoarthritis (OA) is a leading cause of disability worldwide, characterised by chronic pain and reduced quality of life. Despite its prevalence, pharmacotherapy options remain limited. Inflammation has emerged as a promising target, with anti-inflammatory agents used in other rheumatological conditions, such as methotrexate (MTX), being explored for OA treatment. MTX is a cornerstone therapy in rheumatoid arthritis (RA), owing to its broad immunomodulatory properties and well-established clinical efficacy. This review summarises evidence from seven randomised controlled trials and two observational studies investigating MTX in knee and hand OA. Studies varied considerably in terms of sample size, study population, MTX dosage and follow-up duration. Overall, study outcomes were conflicting in terms of MTX effect on OA symptoms. However, trials with larger sample sizes and higher MTX doses (> 15 mg/week) consistently reported benefits for pain in knee and hand OA, with a favourable safety profile, supporting MTX as a potential OA treatment. There is still a need for further research to refine dosing strategies, assess longer term use and evaluate cost-effectiveness. Given the complex heterogeneity of OA, stratification by OA phenotype, particularly consideration of local and systemic inflammation, may also be important to underpin selection of a population most likely to respond to MTX treatment. Considerations for the use of MTX in older adults, where comorbidities and polypharmacy may impact use, will also be essential for clinical implementation.
- 22 January 2026
- Published by Springer Nature in Drugs & Aging
- Vol. 43 (2) , 123-136
Methotrexate for Osteoarthritis: What Does the Evidence Say?
Osteoarthritis (OA) is a leading cause of disability worldwide, characterised by chronic pain and reduced quality of life. Despite its prevalence, pharmacotherapy options remain limited. Inflammation has emerged as a promising target, with anti-inflammatory agents used in other rheumatological conditions, such as methotrexate (MTX), being explored for OA treatment. MTX is a cornerstone therapy in rheumatoid arthritis (RA), owing to its broad immunomodulatory properties and well-established clinical efficacy. This review summarises evidence from seven randomised controlled trials and two observational studies investigating MTX in knee and hand OA. Studies varied considerably in terms of sample size, study population, MTX dosage and follow-up duration. Overall, study outcomes were conflicting in terms of MTX effect on OA symptoms. However, trials with larger sample sizes and higher MTX doses (> 15 mg/week) consistently reported benefits for pain in knee and hand OA, with a favourable safety profile, supporting MTX as a potential OA treatment. There is still a need for further research to refine dosing strategies, assess longer term use and evaluate cost-effectiveness. Given the complex heterogeneity of OA, stratification by OA phenotype, particularly consideration of local and systemic inflammation, may also be important to underpin selection of a population most likely to respond to MTX treatment. Considerations for the use of MTX in older adults, where comorbidities and polypharmacy may impact use, will also be essential for clinical implementation.
- Open Access
- 3 April 2025
- Published by Oxford University Press (OUP) in Rheumatology
- Vol. 64 (8) , 4589-4598
Primary care prediction of hip and knee replacement 1–5 years in advance using Temporal Graph-based Convolutional Neural Networks (TG-CNNs)
Objective: This study aimed to predict the risk of requiring a primary hip or knee replacement 1 and 5 years in advance, using clinical codes. Methods: Primary care clinical codes, sourced from ResearchOne Electronic Health Records between 1999 and 2014, were used to represent patient pathways prior to hip or knee replacement. Patient records were used to train and test models for hip or knee replacement, 1 and 5 years in advance. Temporal graphs were constructed from clinical codes to predict hip and knee replacement risk, where nodes are clinical codes, and edges are the time between primary care visits. Hip and knee replacement cases were matched to controls by age, sex and Index of Multiple Deprivation. The model was validated on unseen data, with performance measured using area under the receiver operator curve (AUROC), calibration and area under the precision recall curve (AUPRC), recalibrating for class imbalance. Results: For knee replacement prediction, AUROC was 0.915 (95% CI 0.914, 0.916) (1 year) and 0.955 (95% CI 0.954, 0.956) (5 years) with AUPRCs of 0.353 (95% CI 0.302, 0.403) and 0.442 (95% CI 0.382, 0.503), respectively. For hip replacement prediction, AUROC was 0.919 (95% CI 0.918, 0.920) (1 year) and 0.967 (95% CI 0.966, 0.968) (5 years), with AUPRCs of 0.409 (95% CI 0.366, 0.451) and 0.879 (95% CI 0.833, 0.924), respectively. Conclusion: Hip and knee replacement risk can be predicted up to 5 years in advance, with a temporal-graph based artificial intelligence model achieving the best performance. This may be used for planning preventative treatment or triaging patients.
- 3 April 2025
- Published by Oxford University Press (OUP) in Rheumatology
- Vol. 64 (8) , 4589-4598
Open AccessPrimary care prediction of hip and knee replacement 1–5 years in advance using Temporal Graph-based Convolutional Neural Networks (TG-CNNs)
Objective: This study aimed to predict the risk of requiring a primary hip or knee replacement 1 and 5 years in advance, using clinical codes. Methods: Primary care clinical codes, sourced from ResearchOne Electronic Health Records between 1999 and 2014, were used to represent patient pathways prior to hip or knee replacement. Patient records were used to train and test models for hip or knee replacement, 1 and 5 years in advance. Temporal graphs were constructed from clinical codes to predict hip and knee replacement risk, where nodes are clinical codes, and edges are the time between primary care visits. Hip and knee replacement cases were matched to controls by age, sex and Index of Multiple Deprivation. The model was validated on unseen data, with performance measured using area under the receiver operator curve (AUROC), calibration and area under the precision recall curve (AUPRC), recalibrating for class imbalance. Results: For knee replacement prediction, AUROC was 0.915 (95% CI 0.914, 0.916) (1 year) and 0.955 (95% CI 0.954, 0.956) (5 years) with AUPRCs of 0.353 (95% CI 0.302, 0.403) and 0.442 (95% CI 0.382, 0.503), respectively. For hip replacement prediction, AUROC was 0.919 (95% CI 0.918, 0.920) (1 year) and 0.967 (95% CI 0.966, 0.968) (5 years), with AUPRCs of 0.409 (95% CI 0.366, 0.451) and 0.879 (95% CI 0.833, 0.924), respectively. Conclusion: Hip and knee replacement risk can be predicted up to 5 years in advance, with a temporal-graph based artificial intelligence model achieving the best performance. This may be used for planning preventative treatment or triaging patients.
- Open Access
- 2 November 2024
- Published by Wiley in Musculoskeletal Care
- Vol. 22 (4) , e70003
Arthroplasty Follow‐Up Pathways: A Qualitative Study of Current Practice and Healthcare Professionals' Attitudes and Motivations for Change
Background: There is considerable diversity in arthroplasty follow‐up pathways. This qualitative study aimed to understand healthcare professionals' practice and attitude to follow‐up, their motivation for change and what evidence they considered before implementing new pathways.Methods: The main UKSAFE study enroled 38 centres providing revision procedures across the United Kingdom. A purposive sample of professional leads and service managers was identified from site contacts. Individual interviews were conducted by telephone, transcribed verbatim and analysed using a theoretical framework derived from current evidence and the data collected in our previous studies.Results: We found that there had been a shift away from bringing patients back into the clinic for repeat follow‐up assessments, but that this was not universal, and some centres had long‐established care pathways that involved long‐term follow‐up. The way in which those services were provided might be different or have common features, but centres were likely to face common problems including large patient numbers and funding restraints.Conclusion: The reliability of newer prosthetics and surgical skill has influenced some changes by increasing confidence in a pathway which does not routinely provide long‐term follow‐up. Service commissioners also have a role to play in how follow‐up care pathways are configured, but scrutiny of the ratio of new to follow‐up appointments can put pressure on clinical staff to follow‐up only patients with identified clinical need. Virtual clinics can provide a service to patients and use scarce resources efficiently, but NHS IT systems that would be needed to support more remote working, for example, telemedicine and plans to collect PROMS data online from patients to assist with monitoring were not advanced.
- 2 November 2024
- Published by Wiley in Musculoskeletal Care
- Vol. 22 (4) , e70003
Open AccessArthroplasty Follow‐Up Pathways: A Qualitative Study of Current Practice and Healthcare Professionals' Attitudes and Motivations for Change
Background: There is considerable diversity in arthroplasty follow‐up pathways. This qualitative study aimed to understand healthcare professionals' practice and attitude to follow‐up, their motivation for change and what evidence they considered before implementing new pathways.Methods: The main UKSAFE study enroled 38 centres providing revision procedures across the United Kingdom. A purposive sample of professional leads and service managers was identified from site contacts. Individual interviews were conducted by telephone, transcribed verbatim and analysed using a theoretical framework derived from current evidence and the data collected in our previous studies.Results: We found that there had been a shift away from bringing patients back into the clinic for repeat follow‐up assessments, but that this was not universal, and some centres had long‐established care pathways that involved long‐term follow‐up. The way in which those services were provided might be different or have common features, but centres were likely to face common problems including large patient numbers and funding restraints.Conclusion: The reliability of newer prosthetics and surgical skill has influenced some changes by increasing confidence in a pathway which does not routinely provide long‐term follow‐up. Service commissioners also have a role to play in how follow‐up care pathways are configured, but scrutiny of the ratio of new to follow‐up appointments can put pressure on clinical staff to follow‐up only patients with identified clinical need. Virtual clinics can provide a service to patients and use scarce resources efficiently, but NHS IT systems that would be needed to support more remote working, for example, telemedicine and plans to collect PROMS data online from patients to assist with monitoring were not advanced.
- Open Access
- 23 October 2024
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 158, 102999
A systematic review of networks for prognostic prediction of health outcomes and diagnostic prediction of health conditions within Electronic Health Records
- 23 October 2024
- Published by Elsevier in Artificial Intelligence in Medicine
- Vol. 158, 102999
Open AccessA systematic review of networks for prognostic prediction of health outcomes and diagnostic prediction of health conditions within Electronic Health Records
- Open Access
Developing the Temporal Graph Convolutional Neural Network Model to Predict Hip Replacement using Electronic Health Records
Background: Hip replacement procedures improve patient lives by relieving pain and restoring mobility. Predicting hip replacement in advance could reduce pain by enabling timely interventions, prioritising individuals for surgery or rehabilitation, and utilising physiotherapy to potentially delay the need for joint replacement. This study predicts hip replacement a year in advance to enhance quality of life and health service efficiency. Methods: Adapting previous work using Temporal Graph Convolutional Neural Network (TG-CNN) models, we construct temporal graphs from primary care medical event codes, sourced from ResearchOne EHRs of 40-75-year-old patients, to predict hip replacement risk. We match hip replacement cases to controls by age, sex, and Index of Multiple Deprivation. The model, trained on 9,187 cases and 9,187 controls, predicts hip replacement one year in advance. We validate the model on two unseen datasets, recalibrating for class imbalance. Additionally, we conduct an ablation study and compare against four baseline models. Results: Our best model predicts hip replacement risk one year in advance with an AUROC of 0.724 (95% CI: 0.715-0.733) and an AUPRC of 0.185 (95% CI: 0.160-0.209), achieving a calibration slope of 1.107 (95% CI: 1.074-1.139) after recalibration. Conclusions: The TG-CNN model effectively predicts hip replacement risk by identifying patterns in patient trajectories, potentially improving understanding and management of hip-related conditions.
Open AccessDeveloping the Temporal Graph Convolutional Neural Network Model to Predict Hip Replacement using Electronic Health Records
Background: Hip replacement procedures improve patient lives by relieving pain and restoring mobility. Predicting hip replacement in advance could reduce pain by enabling timely interventions, prioritising individuals for surgery or rehabilitation, and utilising physiotherapy to potentially delay the need for joint replacement. This study predicts hip replacement a year in advance to enhance quality of life and health service efficiency. Methods: Adapting previous work using Temporal Graph Convolutional Neural Network (TG-CNN) models, we construct temporal graphs from primary care medical event codes, sourced from ResearchOne EHRs of 40-75-year-old patients, to predict hip replacement risk. We match hip replacement cases to controls by age, sex, and Index of Multiple Deprivation. The model, trained on 9,187 cases and 9,187 controls, predicts hip replacement one year in advance. We validate the model on two unseen datasets, recalibrating for class imbalance. Additionally, we conduct an ablation study and compare against four baseline models. Results: Our best model predicts hip replacement risk one year in advance with an AUROC of 0.724 (95% CI: 0.715-0.733) and an AUPRC of 0.185 (95% CI: 0.160-0.209), achieving a calibration slope of 1.107 (95% CI: 1.074-1.139) after recalibration. Conclusions: The TG-CNN model effectively predicts hip replacement risk by identifying patterns in patient trajectories, potentially improving understanding and management of hip-related conditions.