Terence W. O’Neill

Defining Disease Phenotypes in Primary Care Electronic Health Records by a Machine Learning Approach: A Case Study in Identifying Rheumatoid Arthritis.

Objectives 1) To use data-driven method to examine clinical codes (risk factors) of a medical condition in primary care electronic health records (EHRs) that can accurately predict a diagnosis of the condition in secondary care EHRs. 2) To develop and validate a disease phenotyping algorithm for rheumatoid arthritis using primary care EHRs. Methods This study linked routine primary and secondary care EHRs in Wales, UK. A machine learning based scheme was used to identify patients with rheumatoid arthritis from primary care EHRs via the following steps: i) selection of variables by comparing relative frequencies of Read codes in the primary care dataset associated with disease case compared to non-disease control (disease/non-disease based on the secondary care diagnosis); ii) reduction of predictors/associated variables using a Random Forest method, iii) induction of decision rules from decision tree model. The proposed method was then extensively validated on an independent dataset, and compared for performance with two existing deterministic algorithms for RA which had been developed using expert clinical knowledge. Results Primary care EHRs were available for 2,238,360 patients over the age of 16 and of these 20,667 were also linked in the secondary care rheumatology clinical system. In the linked dataset, 900 predictors (out of a total of 43,100 variables) in the primary care record were discovered more frequently in those with versus those without RA. These variables were reduced to 37 groups of related clinical codes, which were used to develop a decision tree model. The final algorithm identified 8 predictors related to diagnostic codes for RA, medication codes, such as those for disease modifying anti-rheumatic drugs, and absence of alternative diagnoses such as psoriatic arthritis. The proposed data-driven method performed as well as the expert clinical knowledge based methods. Conclusion Data-driven scheme, such as ensemble machine learning methods, has the potential of identifying the most informative predictors in a cost-effective and rapid way to accurately and reliably classify rheumatoid arthritis or other complex medical conditions in primary care EHRs.

Clinical assessment of effusion in knee osteoarthritis—A systematic review

Objective The aim of this systematic review was to determine the validity and inter- and intra-observer reliability of the assessment of knee joint effusion in osteoarthritis (OA) of the knee. Methods MEDLINE, Web of Knowledge, CINAHL, EMBASE, and AMED were searched from their inception to February 2015. Articles were included according to a priori defined criteria: samples containing participants with knee OA; prospective evaluation of clinical tests and assessments of knee effusion that included reliability, sensitivity, and specificity of these tests. Results A total of 10 publications were reviewed. Eight of these considered reliability and four on validity of clinical assessments against ultrasound effusion. It was not possible to undertake a meta-analysis of reliability or validity because of differences in study designs and the clinical tests. Intra-observer kappa agreement for visible swelling ranged from 0.37 (suprapatellar) to 1.0 (prepatellar); for bulge sign 0.47 and balloon sign 0.37. Inter-observer kappa agreement for visible swelling ranged from −0.02 (prepatellar) to 0.65 (infrapatellar), the balloon sign −0.11 to 0.82, patellar tap −0.02 to 0.75 and bulge sign kappa −0.04 to 0.14 or reliability coefficient 0.97. Reliability and diagnostic accuracy tended to be better in experienced observers. Very few data looked at performance of individual clinical tests with sensitivity ranging 18.2–85.7% and specificity 35.3–93.3%, both higher with larger effusions. Conclusion The majority of unstandardized clinical tests to assess joint effusion in knee OA had relatively low intra- and inter-observer reliability. There is some evidence experience improved reliability and diagnostic accuracy of tests. Currently there is insufficient evidence to recommend any particular test in clinical practice.

Accuracy of injection and short-term pain relief following intra-articular corticosteroid injection in knee osteoarthritis – an observational study

BackgroundIntra-articular corticosteroid injections (IACI) are effective treatments for pain in knee osteoarthritis (KOA) but treatment response varies. There is uncertainty as to whether structural factors such as accurate placement of IACI affect outcome. We examined this question in a pragmatic observational study, using ultrasound (US) to verify accuracy of IACI.Methods105 subjects with KOA (mean age 63.1xa0years, 59% female) routinely referred for IACI underwent assessment of demographic factors, x-ray and US of the knee before aspiration and IACI (based on clinical landmarks) with 40xa0mg triamcinolone acetonide with lignocaine plus a small amount of atmospheric air by an independent physician. US demonstration of intra-articular mobile air, i.e. a positive air arthrosonogram, was used to determine accurate placement of injection. Both patients and injecting physicians were blind to the US findings. Pain at baseline, three and nine weeks post injection was assessed using the 500xa0mm WOMAC pain subscale and response defined asu2009≥u200940% reduction in pain from baseline. Inter-observer reliability of air-arthrosonogram assessment was good: κ 0.79 (three raters).ResultsSixty-three subjects (60.6%) were responders at three weeks and 43 (45.7%) at nine weeks. Seventy-four subjects (70.5%) had a positive arthrosonogram. A positive air arthrosonogram did not associate with a higher rate of response to treatment (p 0.389 at three weeks, p 0.365 at nine weeks). There was no difference in US effusion depth, power Doppler signal or radiographic grade between responders and non-responders to the injection, but female gender associated with response at 3xa0weeks and previous injection with non-response at 9xa0weeks.ConclusionsAccurate intra-articular injection of corticosteroid results did not result in superior outcome in terms of pain compared to inaccurate injection in symptomatic knee OA.

Structural predictors of response to intra-articular steroid injection in symptomatic knee osteoarthritis

BackgroundThe aim was to examine if structural factors could affect response to intra-articular steroid injections (IASI) in knee osteoarthritis (OA).MethodPersons with painful knee OA participated in an open-label trial of IASI where radiographic joint space narrowing (JSN) and Kellgren-Lawrence (KL) grade, whole-organ magnetic resonance imaging (MRI) scores (WORMS) and quantitative assessment of synovial tissue volume (STV) were assessed on baseline images. Participants completed the Knee Injury and Osteoarthritis Outcome Score (KOOS) and a question about knee pain with a visual analogue scale for pain during nominated activity (VASNA), and Outcome Measures in Rheumatology (OMERACT)-Osteoarthritis Research Society International (OARSI) criteria were used to assess responder status within 2xa0weeks (short term) and 6xa0months (longer term). Regression models were used to examine predictors of short and longer term response to IASI.ResultsSubjects (nu2009=u2009207) attended and had IASI. Information on responder status was available on 199 participants. Of these, 188 subjects, mean age 63.2xa0years (standard deviation (SD) 10.3), 97 (51.6%) female, had x-rays and 120 had MRI scans available. Based on the OMERACT-OARSI criteria, 146 (73.4%) participants responded to therapy and 40 (20.1%) were longer term responders. A few factors were associated with a reduced KOOS-pain and VASNA response though none were associated with OMERACT-OARSI responder status in the short term. Higher MRI meniscal damage (odds ratio (OR) = 0.74; 95% CI 0.55 to 0.98), increasing KL maximal grade (ORu2009=u20090.43; 95% CI 0.23 to 0.82) and joint space narrowing (JSN) maximal score (ORu2009=u20090.60; 95% CI 0.36 to 0.99) were each associated with a lower odds of longer term responder status. Baseline synovitis was not associated with treatment response. The predicted probability of longer term response decreased from 38% to 12% as baseline maximal JSN increased from grade 0 to 3.ConclusionCompared with those who have mild structural damage, persons with more severe knee damage on either MRI or x-ray are less likely to respond to knee IASI.Trial registrationISRCTN.com, ISRCTN07329370. Registered 21 May 2010. Retrospectively registered

With a biomechanical treatment in knee osteoarthritis, less knee pain did not correlate with synovitis reduction

BackgroundBraces are used to treat pain in patellofemoral joint osteoarthritis (PFJOA). In a trial, we previously reported pain improvement after 6-weeks brace use. The pain reduction did not correlate with changes in Magnetic Resonance Imaging (MRI) assessed Bone Marrow Lesion volume or static synovial volume. Studies show that changes in the synovium on dynamic contrast enhanced (DCE) MRI are more closely associated with symptom change than static synovial volume changes. We hypothesised change in synovitis assessed using dynamic imaging could explain the reduction in pain.MethodOne hundred twenty-six men and women aged 40–70xa0years with painful radiographically confirmed PFJOA were randomised to either brace wearing or no brace for 6-weeks. Pain assessment and DCE-MRI were performed at baseline and 6xa0weeks. DCE data was analysed using Tofts’s equation. Pain measures included a VAS of pain on nominated aggravating activity (VASNA), and the KOOS pain subscale. Paired t-tests were used to determine within person change in outcome measures and Spearman’s correlation coefficients were used to determine the correlation between change in pain and change in the DCE parameters.ResultsMean age of subjects was 55.5xa0years (SDxa0=xa07.5) and 57% were female. There was clear pain improvement in the brace users compared to controls (VASNAxa0−xa016.87xa0mm, pxa0=xa0<0.001). There was no significant change to the dynamic synovitis parameters among brace users nor was pain change correlated with change in dynamic synovitis parameters.ConclusionThe reduction in knee pain following brace wearing in patients with PFJOA is not explained by changes in synovitis.Trial registrationTrial registration number UK. ISRCTN50380458/Registered 21.5.2010.

Mechanisms of Osteoarthritis (OA) Pain

Purpose of ReviewOsteoarthritis (OA) is a major cause of pain and disability worldwide. There is, however, a relatively poor correlation between the severity of OA based on plain radiograph changes and symptoms. In this review, we consider the mechanisms of pain in OA.Recent FindingsIt is now widely recognised that OA is a disease of the whole joint. Data from large observational studies which have used magnetic resonance imaging (MRI) suggest that pain in OA is associated with a number of structural factors including the presence of bone marrow lesions (BMLs) and also synovitis. There is evidence also of alterations in nerve processing and that both peripheral and central nerve sensitisation may contribute to pain in OA.SummaryIdentification of the causes of pain in an individual patient may be of benefit in helping to better target with appropriate therapy to help reduce their symptoms and improve function.

A prospective cohort study measuring cost-benefit analysis of the Otago Exercise Programme in community dwelling adults with rheumatoid arthritis.

BackgroundFalls are one of the major health problems in adults with Rheumatoid Arthritis (RA). Interventions, such as the Otago Exercise Programme (OEP), can reduce falls in community dwelling adults by up to 35%. The cost-benefits of such a programme in adults with RA have not been studied.The aims of this study were to determine the healthcare cost of falls in adults with RA, and estimate whether it may be cost efficient to roll out the OEP to improve function and prevent falls in adults living with RA.MethodsPatients with Rheumatoid Arthritis aged ≥18xa0years were recruited from four rheumatology clinics across the Northwest of England. Participants were followed up for 1xa0year with monthly fall calendars, telephone calls and self-report questionnaires. Estimated medical cost of a fall-related injury incurred per-person were calculated and compared with OEP implementation costs to establish potential economic benefits.ResultsFive hundred thirty-five patients were recruited and 598 falls were reported by 195 patients. Cumulative medical costs resulting from all injury leading to hospital services is £374,354 (US

Erratum to “Where and how to inject the knee—A systematic review” [Seminars in Arthritis and Rheumatism 2013;43:195-203]

540,485). Average estimated cost per fall is £1120 (US

A prospective study of fall risk factors in adults with rheumatoid arthritis.

1617). Estimated cost of implementing the OEP for 535 people is £116,479 (US

Influence of Inflammatory Polyarthritis on Quantitative Heel Ultrasound Measurements

168,504) or £217.72 (US