Zador Z.,Salford Royal Foundation Trust |
Zador Z.,University of Manchester |
Sperrin M.,Health eResearch Centre | |
King A.T.,Salford Royal Foundation Trust
PLoS ONE | Year: 2016
Background: Traumatic brain injury remains a global health problem. Understanding the relative importance of outcome predictors helps optimize our treatment strategies by informing assessment protocols, clinical decisions and trial designs. In this study we establish importance ranking for outcome predictors based on receiver operating indices to identify key predictors of outcome and create simple predictive models. We then explore the associations between key outcome predictors using Bayesian networks to gain further insight into predictor importance. Methods: We analyzed the corticosteroid randomization after significant head injury (CRASH) trial database of 10008 patients and included patients for whom demographics, injury characteristics, computer tomography (CT) findings and Glasgow Outcome Scale (GCS) were recorded (total of 13 predictors, which would be available to clinicians within a few hours following the injury in 6945 patients). Predictions of clinical outcome (death or severe disability at 6 months) were performed using logistic regression models with 5-fold cross validation. Predictive performance was measured using standardized partial area (pAUC) under the receiver operating curve (ROC) and we used Delong test for comparisons. Variable importance ranking was based on pAUC targeted at specificity (pAUCSP) and sensitivity (pAUCSE) intervals of 90-100%. Probabilistic associations were depicted using Bayesian networks. Results: Complete AUC analysis showed very good predictive power (AUC = 0.8237, 95% CI: 0.8138-0.8336) for the complete model. Specificity focused importance ranking highlighted age, pupillary, motor responses, obliteration of basal cisterns/3rd ventricle and midline shift. Interestingly when targeting model sensitivity, the highest-ranking variables were age, severe extracranial injury, verbal response, hematoma on CT and motor response. Simplified models, which included only these key predictors, had similar performance (pAUCSP = 0.6523, 95% CI: 0.6402-0.6641 and pAUCSE = 0.6332, 95% CI: 0.62-0.6477) compared to the complete models (pAUCSP = 0.6664, 95% CI: 0.6543-0.679, pAUCSE = 0.6436, 95% CI: 0.6289-0.6585, de Long p value 0.1165 and 0.3448 respectively). Bayesian networks showed the predictors that did not feature in the simplified models were associated with those that did. Conclusion: We demonstrate that importance based variable selection allows simplified predictive models to be created while maintaining prediction accuracy. Variable selection targeting specificity confirmed key components of clinical assessment in TBI whereas sensitivity based ranking suggested extracranial injury as one of the important predictors. These results help refine our approach to head injury assessment, decision-making and outcome prediction targeted at model sensitivity and specificity. Bayesian networks proved to be a comprehensive tool for depicting probabilistic associations for key predictors giving insight into why the simplified model has maintained accuracy. © 2016 Zador et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Kovacevic A.,University of Novi Sad |
Dehghan A.,University of Manchester |
Filannino M.,University of Manchester |
Keane J.A.,University of Manchester |
And 2 more authors.
Journal of the American Medical Informatics Association | Year: 2013
Objective: Identification of clinical events (eg, problems, tests, treatments) and associated temporal expressions (eg, dates and times) are key tasks in extracting and managing data from electronic health records. As part of the i2b2 2012 Natural Language Processing for Clinical Data challenge, we developed and evaluated a system to automatically extract temporal expressions and events from clinical narratives. The extracted temporal expressions were additionally normalized by assigning type, value, and modifier. Materials and methods: The system combines rulebased and machine learning approaches that rely on morphological, lexical, syntactic, semantic, and domainspecific features. Rule-based components were designed to handle the recognition and normalization of temporal expressions, while conditional random fields models were trained for event and temporal recognition. Results: The system achieved micro F scores of 90% for the extraction of temporal expressions and 87% for clinical event extraction. The normalization component for temporal expressions achieved accuracies of 84.73% (expression's type), 70.44% (value), and 82.75% (modifier). Discussion: Compared to the initial agreement between human annotators (87-89%), the system provided comparable performance for both event and temporal expression mining. While (lenient) identification of such mentions is achievable, finding the exact boundaries proved challenging. Conclusions: The system provides a state-of-the-art method that can be used to support automated identification of mentions of clinical events and temporal expressions in narratives either to support the manual review process or as a part of a large-scale processing of electronic health databases.
Ainsworth J.,Health eResearch Centre | |
Ainsworth J.,University of Manchester |
Buchan I.,Health eResearch Centre | |
Buchan I.,University of Manchester
Methods of Information in Medicine | Year: 2015
Objectives: In this paper we aim to characterise the critical mass of linked data, methods and expertise required for health systems to adapt to the needs of the populations they serve–more recently known as learning health systems. The objectives are to: 1) identify opportunities to combine separate uses of common data sources in order to reduce duplication of data processing and improve information quality; 2) identify challenges in scaling-up the reuse of health data sufficiently to support health system learning. Methods: The challenges and opportunities were identified through a series of e-health stakeholder consultations and workshops in Northern England from 2011 to 2014. From 2013 the concepts presented here have been refined through feedback to collaborators, including patient/citizen representatives, in a regional health informatics research network (www.herc.ac.uk). Results: Health systems typically have sepa rate information pipelines for: 1) commissioning services; 2) auditing service performance; 3) managing finances; 4) monitoring public health; and 5) research. These pipelines share common data sources but usually duplicate data extraction, aggregation, cleaning/preparation and analytics. Suboptimal analyses may be performed due to a lack of expertise, which may exist elsewhere in the health system but is fully committed to a different pipeline. Contextual knowledge that is essential for proper data analysis and interpretation may be needed in one pipeline but accessible only in another. The lack of capable health and care intelligence systems for populations can be attributed to a legacy of three flawed assumptions: 1) universality: the generalizability of evidence across populations; 2) time-invariance: the stability of evidence over time; and 3) reducibility: the reduction of evidence into specialised subsystems that may be recombined. Conclusions: We conceptualize a population health and care intelligence system capable of supporting health system learning and we put forward a set of maturity tests of progress toward such a system. A factor common to each test is data-action latency; a mature system spawns timely actions proportionate to the information that can be derived from the data, and in doing so creates meaningful measurement about system learning. We illustrate, using future scenarios, some major opportunities to improve health systems by exchanging conventional intelligence pipelines for networked critical masses of data, methods and expertise that minimise dataaction latency and ignite system-learning. © Schattauer 2015.
Verburg I.W.M.,University of Amsterdam |
De Keizer N.F.,University of Amsterdam |
Holman R.,University of Amsterdam |
Dongelmans D.,University of Amsterdam |
And 2 more authors.
Critical Care Medicine | Year: 2016
Objectives: The performance of ICUs can be compared by ranking them into a league table according to their risk-adjusted mortality rate. The statistical quality of a league table can be expressed as its rankability, the percentage of variation between ICUs attributable to unexplained differences. We examine whether we can improve the rankability of our league table by using data from a longer period or by grouping ICUs with similar performance constructing a league table of clusters rather than individual ICUs. Design: We developed a league table for risk-adjusted mortality rate with its rankability. The effect of assessment period was determined using a resampling procedure. Hierarchical clustering was used to obtain clusters of similar ICUs. Patients: We used data from ICUs participating in the Dutch National Intensive Care Evaluation registry between 2011 and 2013. Measurements and Main Results: We constructed league tables using 157,394 admissions from 78 ICUs with risk-adjusted mortality rate between 5.9% and 13.9% per ICU over the inclusion period. The rankability was 73% for 2013 and 89% for the whole period 2011-2013. Rankability over the year 2013 increased till 98% when clustering ICUs, reaching an optimum at a league table of seven clusters. Conclusions: We conclude that, when using data from a single year, the rankability of a league table of Dutch ICUs based on risk-adjusted mortality rate was unacceptably low. We could improve the rankability of this league table by increasing the period of data collection or by grouping similar ICUs into clusters and constructing a league table of clusters of ICUs rather than individual ICUs. Ranking clusters of ICUs could be useful for identifying possible differences in performance between clusters of ICUs.
Sperrin M.,Health eResearch Centre | |
Marshall A.D.,University of Manchester |
Higgins V.,University of Manchester |
Buchan I.E.,Health eResearch Centre | |
Renehan A.G.,Health eResearch Centre |
International Journal of Obesity | Year: 2014
Background: The prevalence of excess body weight, commonly measured as body mass index (BMI)≥25 kg m-2, has increased substantially in many populations worldwide over the past three decades, but the rate of increase has slowed down in some western populations. Objective: We address the hypothesis that the slowing down of BMI trend increases in England reflects a majority sub-population resistant to further BMI elevation. Design: Pseudo-panel data derived from annual cross-sectional surveys, the Health Surveys for England (1992-2010). Trends in median BMI values were explored using regression models with splines, and gender-specific mixture model (latent class analysis) were fit to take an account of increasing BMI distribution variance with time and identify hidden subgroups within the population. Subjects: BMI was available for 164 155 adults (men: 76 382; women: 87 773). Results: Until 2001, the age-adjusted yearly increases in median BMI were 0.140 and 0.139 kg m -2 for men and women, respectively, decreasing thereafter to 0.073 and 0.055 kg m-2 (differences between time periods, both P-values<0.0001). The mixture model identified two components - a normal BMI and a high BMI sub-population - the proportions for the latter were 23.5% in men and 33.7% in women. The remaining normal BMI populations were 'resistant' with minimal increases in mean BMI values over time. By age, mean BMI values in the normal BMI sub-population increased greatest between 20 and 34 years for men; for women, the increases were similar throughout age groups (slope differences, P<0.0001). Conclusion: In England, recent slowing down of adult BMI trend increases can be explained by two sub-populations - a high BMI sub-population getting 'fatter' and a majority 'resistant' normal BMI sub-population. These findings support a targeted, rather than a population-wide, policy to tackle the determinants of obesity. © 2014 Macmillan Publishers Limited All rights reserved.
Herrett E.,London School of Hygiene and Tropical Medicine |
Gallagher A.M.,Datalink |
Gallagher A.M.,University Utrecht |
Bhaskaran K.,London School of Hygiene and Tropical Medicine |
And 6 more authors.
International Journal of Epidemiology | Year: 2015
The Clinical Practice Research Datalink (CPRD) is an ongoing primary care database of anonymised medical records from general practitioners, with coverage of over 11.3 million patients from 674 practices in the UK. With 4.4 million active (alive, currently registered) patients meeting quality criteria, approximately 6.9% of the UK population are included and patients are broadly representative of the UK general population in terms of age, sex and ethnicity. General practitioners are the gatekeepers of primary care and specialist referrals in the UK. The CPRD primary care database is therefore a rich source of health data for research, including data on demographics, symptoms, tests, diagnoses, therapies, health-related behaviours and referrals to secondary care. For over half of patients, linkage with datasets from secondary care, disease-specific cohorts and mortality records enhance the range of data available for research. The CPRD is very widely used internationally for epidemiological research and has been used to produce over 1000 research studies, published in peer-reviewed journals across a broad range of health outcomes. However, researchers must be aware of the complexity of routinely collected electronic health records, including ways to manage variable completeness, misclassification and development of disease definitions for research. © The Author 2015. Published by Oxford University Press on behalf of the International Epidemiological Association.
Greve S.V.,Health eResearch Centre |
Journal of Hypertension | Year: 2016
BACKGROUND:: Carotid–femoral pulse wave velocity (cfPWV) adds significantly to traditional cardiovascular risk prediction, but is not widely available. Therefore, it would be helpful if cfPWV could be replaced by an estimated carotid–femoral pulse wave velocity (ePWV) using age and mean blood pressure, and previously published equations. The aim of this study was to investigate whether ePWV could predict cardiovascular events independently of traditional cardiovascular risk factors and/or cfPWV. METHOD:: cfPWV was measured and ePWV was calculated in 2366 patients from four age groups of the Danish MONICA10 cohort. Additionally, the patients were divided into four cardiovascular risk groups based on Systematic COronary Risk Evaluation (SCORE) or Framingham risk score (FRS). In 2006, the combined cardiovascular endpoint of cardiovascular death, nonfatal myocardial infarction, stroke and hospitalization for ischemic heart disease was registered. RESULTS:: Most results were retested in 1045 hypertensive patients from a Paris cohort. Bland–Altman plot demonstrated a relative difference of −0.3% [95% confidence interval (CI) −15 to 17%] between ePWV and cfPWV. In Cox regression models in apparently healthy patients, ePWV and cfPWV (per SD) added independently to SCORE in prediction of combined endpoint [hazard ratio (95%CI)?=?1.38(1.09–1.76) and hazard ratio (95%CI)?=?1.18(1.01–1.38)] and to FRS [hazard ratio (95%CI)?=?1.33(1.06–1.66) and hazard ratio (95%CI)?=?1.16(0.99–1.37)]. If healthy patients with ePWV and/or cfPWV at least 10?m/s were reclassified to a higher SCORE risk category, net reclassification index was 10.8%, P less than 0.01. These results were reproduced in the Paris cohort. CONCLUSION:: ePWV predicted major cardiovascular events independently of SCORE, FRS and cfPWV indicating that these traditional risk scores have underestimated the complicated impact of age and blood pressure on arterial stiffness and cardiovascular risk. Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.
Kraal J.J.,University of Amsterdam |
Peek N.,Health eResearch Centre | |
Van den Akker-Van Marle M.E.,Leiden University |
Kemps H.M.,University of Amsterdam
European journal of preventive cardiology | Year: 2014
BACKGROUND: Home-based exercise training in cardiac rehabilitation (CR) has the potential to improve CR uptake, decrease costs and increase self-management skills. The FIT@Home study evaluates home-based CR with telemonitoring guidance using coaching interventions including strategies for behavioural changes with the aim to maintain adherence to a healthy lifestyle and to improve long-term effects. In this interim analysis we provide short-term results on exercise capacity, quality of life and training adherence of the first 50 patients included in the FIT@Home study.DESIGN: The study design was a randomised controlled trial.METHODS: Low to moderate risk CR patients were randomised to a 12-week home-based training (HT) programme or a 12-week centre-based training (CT) programme. In both groups, training was performed at 70-85% of maximal heart rate (HRmax) for 45-60 min, 2-3 times per week. The HT group received three supervised training sessions, before commencing training with a heart rate monitor in their home environment. These patients received individual coaching by telephone weekly, based on training data uploaded on the Internet. The CT programme was performed under the direct supervision of a physical therapist. Exercise capacity and health-related quality of life were assessed at baseline and at 12 weeks.RESULTS: CT (n = 25) and HT (n = 25) both showed a significant improvement in peak oxygen uptake (peak VO2) (10% and 14% respectively) and quality of life after 12 weeks of training, without significant between-group differences. The average training intensity of the HT group was 73.3 ± 3.5% of HRmax. Training adherence was similar between groups.CONCLUSION: This analysis shows that HT with telemonitoring guidance has similar short-term effects on exercise capacity and quality of life as CT in CR patients. © Authors 2014 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav.
Dehghan A.,University of Manchester |
Keane J.A.,University of Manchester |
Nenadic G.,University of Manchester |
Nenadic G.,Health eResearch Centre |
Proceedings - 2013 IEEE International Conference on Systems, Man, and Cybernetics, SMC 2013 | Year: 2013
In addition to structured data, electronic health records contain unstructured clinical notes and narratives. The identification and classification of mentions of relevant clinical concepts is a crucial preprocessing step in designing and developing clinical decision support systems. While this task has gained significant attention in recent years, there are still a number of issues that need further investigation. This paper explores a variety of common challenges faced by clinical named entity recognition and classification methods as well as current approaches to handling them. © 2013 IEEE.