Using secondary knowledge to support decision tree classification of retrospective clinical data

Retrospective clinical data presents many challenges for data mining and machine learning. The transcription of patient records from paper charts and subsequent manipulation of data often results in high volumes of noise as well as a loss of other important information. In addition, such datasets often fail to represent expert medical knowledge and reasoning in any explicit manner. In this research we describe applying data mining methods to retrospective clinical data to build a prediction model for asthma exacerbation severity for pediatric patients in the emergency department. Difficulties in building such a model forced us to investigate alternative strategies for analyzing and processing retrospective data. This paper describes this process together with an approach to mining retrospective clinical data by incorporating formalized external expert knowledge (secondary knowledge sources) into the classification task. This knowledge is used to partition the data into a number of coherent sets, where each set is explicitly described in terms of the secondary knowledge source. Instances from each set are then classified in a manner appropriate for the characteristics of the particular set. We present our methodology and outline a set of experiential results that demonstrate some advantages and some limitations of our approach. © 2008 Springer-Verlag Berlin Heidelberg.

An investigation into inductive parameter learning in complex hierarchical knowledge structures representing clinical expertise

This dissertation investigates the very important and current problem of modelling human expertise. This is an apparent issue in any computer system emulating human decision making. It is prominent in Clinical Decision Support Systems (CDSS) due to the complexity of the induction process and the vast number of parameters in most cases. Other issues such as human error and missing or incomplete data present further challenges. In this thesis, the Galatean Risk Screening Tool (GRiST) is used as an example of modelling clinical expertise and parameter elicitation. The tool is a mental health clinical record management system with a top layer of decision support capabilities. It is currently being deployed by several NHS mental health trusts across the UK. The aim of the research is to investigate the problem of parameter elicitation by inducing them from real clinical data rather than from the human experts who provided the decision model. The induced parameters provide an insight into both the data relationships and how experts make decisions themselves. The outcomes help further understand human decision making and, in particular, help GRiST provide more accurate emulations of risk judgements. Although the algorithms and methods presented in this dissertation are applied to GRiST, they can be adopted for other human knowledge engineering domains.

Pragmatic development of service based real-time change data capture

This thesis makes a contribution to the Change Data Capture (CDC) field by providing an empirical evaluation on the performance of CDC architectures in the context of realtime data warehousing. CDC is a mechanism for providing data warehouse architectures with fresh data from Online Transaction Processing (OLTP) databases. There are two types of CDC architectures, pull architectures and push architectures. There is exiguous data on the performance of CDC architectures in a real-time environment. Performance data is required to determine the real-time viability of the two architectures. We propose that push CDC architectures are optimal for real-time CDC. However, push CDC architectures are seldom implemented because they are highly intrusive towards existing systems and arduous to maintain. As part of our contribution, we pragmatically develop a service based push CDC solution, which addresses the issues of intrusiveness and maintainability. Our solution uses Data Access Services (DAS) to decouple CDC logic from the applications. A requirement for the DAS is to place minimal overhead on a transaction in an OLTP environment. We synthesize DAS literature and pragmatically develop DAS that eciently execute transactions in an OLTP environment. Essentially we develop effeicient RESTful DAS, which expose Transactions As A Resource (TAAR). We evaluate the TAAR solution and three pull CDC mechanisms in a real-time environment, using the industry recognised TPC-C benchmark. The optimal CDC mechanism in a real-time environment, will capture change data with minimal latency and will have a negligible affect on the database's transactional throughput. Capture latency is the time it takes a CDC mechanism to capture a data change that has been applied to an OLTP database. A standard definition for capture latency and how to measure it does not exist in the field. We create this definition and extend the TPC-C benchmark to make the capture latency measurement. The results from our evaluation show that pull CDC is capable of real-time CDC at low levels of user concurrency. However, as the level of user concurrency scales upwards, pull CDC has a significant impact on the database's transaction rate, which affirms the theory that pull CDC architectures are not viable in a real-time architecture. TAAR CDC on the other hand is capable of real-time CDC, and places a minimal overhead on the transaction rate, although this performance is at the expense of CPU resources.

Handling varying amounts of missing data when classifying mental-health risk levels

One of the main challenges of classifying clinical data is determining how to handle missing features. Most research favours imputing of missing values or neglecting records that include missing data, both of which can degrade accuracy when missing values exceed a certain level. In this research we propose a methodology to handle data sets with a large percentage of missing values and with high variability in which particular data are missing. Feature selection is effected by picking variables sequentially in order of maximum correlation with the dependent variable and minimum correlation with variables already selected. Classification models are generated individually for each test case based on its particular feature set and the matching data values available in the training population. The method was applied to real patients' anonymous mental-health data where the task was to predict the suicide risk judgement clinicians would give for each patient's data, with eleven possible outcome classes: zero to ten, representing no risk to maximum risk. The results compare favourably with alternative methods and have the advantage of ensuring explanations of risk are based only on the data given, not imputed data. This is important for clinical decision support systems using human expertise for modelling and explaining predictions.

Precompetitive data sharing as a catalyst to address unmet needs in Parkinson's disease

Parkinson's disease is a complex heterogeneous disorder with urgent need for disease-modifying therapies. Progress in successful therapeutic approaches for PD will require an unprecedented level of collaboration. At a workshop hosted by Parkinson's UK and co-organized by Critical Path Institute's (C-Path) Coalition Against Major Diseases (CAMD) Consortiums, investigators from industry, academia, government and regulatory agencies agreed on the need for sharing of data to enable future success. Government agencies included EMA, FDA, NINDS/NIH and IMI (Innovative Medicines Initiative). Emerging discoveries in new biomarkers and genetic endophenotypes are contributing to our understanding of the underlying pathophysiology of PD. In parallel there is growing recognition that early intervention will be key for successful treatments aimed at disease modification. At present, there is a lack of a comprehensive understanding of disease progression and the many factors that contribute to disease progression heterogeneity. Novel therapeutic targets and trial designs that incorporate existing and new biomarkers to evaluate drug effects independently and in combination are required. The integration of robust clinical data sets is viewed as a powerful approach to hasten medical discovery and therapies, as is being realized across diverse disease conditions employing big data analytics for healthcare. The application of lessons learned from parallel efforts is critical to identify barriers and enable a viable path forward. A roadmap is presented for a regulatory, academic, industry and advocacy driven integrated initiative that aims to facilitate and streamline new drug trials and registrations in Parkinson's disease.

Comparison of estimates and calculations of risk of coronary heart disease by doctors and nurses using different calculation tools in general practice: cross sectional study

OBJECTIVE: To assess the effect of using different risk calculation tools on how general practitioners and practice nurses evaluate the risk of coronary heart disease with clinical data routinely available in patients' records. DESIGN: Subjective estimates of the risk of coronary heart disease and results of four different methods of calculation of risk were compared with each other and a reference standard that had been calculated with the Framingham equation; calculations were based on a sample of patients' records, randomly selected from groups at risk of coronary heart disease. SETTING: General practices in central England. PARTICIPANTS: 18 general practitioners and 18 practice nurses. MAIN OUTCOME MEASURES: Agreement of results of risk estimation and risk calculation with reference calculation; agreement of general practitioners with practice nurses; sensitivity and specificity of the different methods of risk calculation to detect patients at high or low risk of coronary heart disease. RESULTS: Only a minority of patients' records contained all of the risk factors required for the formal calculation of the risk of coronary heart disease (concentrations of high density lipoprotein (HDL) cholesterol were present in only 21%). Agreement of risk calculations with the reference standard was moderate (kappa=0.33-0.65 for practice nurses and 0.33 to 0.65 for general practitioners, depending on calculation tool), showing a trend for underestimation of risk. Moderate agreement was seen between the risks calculated by general practitioners and practice nurses for the same patients (kappa=0.47 to 0.58). The British charts gave the most sensitive results for risk of coronary heart disease (practice nurses 79%, general practitioners 80%), and it also gave the most specific results for practice nurses (100%), whereas the Sheffield table was the most specific method for general practitioners (89%). CONCLUSIONS: Routine calculation of the risk of coronary heart disease in primary care is hampered by poor availability of data on risk factors. General practitioners and practice nurses are able to evaluate the risk of coronary heart disease with only moderate accuracy. Data about risk factors need to be collected systematically, to allow the use of the most appropriate calculation tools.

Methodology of photic stimulation revisited:updated European algorithm for visual stimulation in the EEG laboratory

Intermittent photic stimulation (IPS) is a common procedure performed in the electroencephalography (EEG) laboratory in children and adults to detect abnormal epileptogenic sensitivity to flickering light (i.e., photosensitivity). In practice, substantial variability in outcome is anecdotally found due to the many different methods used per laboratory and country. We believe that standardization of procedure, based on scientific and clinical data, should permit reproducible identification and quantification of photosensitivity. We hope that the use of our new algorithm will help in standardizing the IPS procedure, which in turn may more clearly identify and assist monitoring of patients with epilepsy and photosensitivity. Our algorithm goes far beyond that published in 1999 (Epilepsia, 1999a, 40, 75; Neurophysiol Clin, 1999b, 29, 318): it has substantially increased content, detailing technical and logistical aspects of IPS testing and the rationale for many of the steps in the IPS procedure. Furthermore, our latest algorithm incorporates the consensus of repeated scientific meetings of European experts in this field over a period of 6 years with feedback from general neurologists and epileptologists to improve its validity and utility. Accordingly, our European group has provided herein updated algorithms for two different levels of methodology: (1) requirements for defining photosensitivity in patients and in family members of known photosensitive patients and (2) requirements for tailored studies in patients with a clear history of visually induced seizures or complaints, and in those already known to be photosensitive.

Dipeptidyl peptidase IV (DPP IV) and related molecules in type 2 diabetes

Dipeptidyl peptidase IV (DPP IV) is a widely distributed physiological enzyme that can be found solubilized in blood, or membrane-anchored in tissues. DPP IV and related dipeptidase enzymes cleave a wide range of physiological peptides and have been associated with several disease processes including Crohn's disease, chronic liver disease, osteoporosis, multiple sclerosis, eating disorders, rheumatoid arthritis, cancer, and of direct relevance to this review, type 2 diabetes. Here, we place particular emphasis on two peptide substrates of DPP IV with insulin-releasing and antidiabetic actions namely, glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP). The rationale for inhibiting DPP IV activity in type 2 diabetes is that it decreases peptide cleavage and thereby enhances endogenous incretin hormone activity. A multitude of novel DPP IV inhibitor compounds have now been developed and tested. Here we examine the information available on DPP IV and related enzymes, review recent preclinical and clinical data for DPP IV inhibitors, and assess their clinical significance.

Integration issues in the healthcare supply chain

The efficacy, quality, responsiveness, and value of healthcare services provided is increasingly attracting the attention and the questioning of governments, payers, patients, and healthcare providers. Investments on integration technologies and integration of supply chain processes, has been considered as a way towards removing inefficiencies in the sector. This chapter aims to initially provide an in depth analysis of the healthcare supply chain and to present core entities, processes, and flows. Moreover, the chapter explores the concept of integration in the context of the healthcare sector, and indentifies the integration drivers, as well as challenges.

Pharmacogenomic studies of the anticancer and immunosuppressive thiopurines mercaptopurine and azathioprine

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT • 6-Mercaptopurine (6-MP) and azathioprine (AZA) are both inactive prodrugs that require intracellular activation into the active 6-thioguanine nucleotides (6-TGNs). • This metabolic process undergoes three different competitive pathways that are catalysed by three different enzymes; xanthine oxidase (XO), thiopurine methyltransferase (TPMT) and inosine triphosphatase (ITPA), all of which exhibit genetic polymorphisms. • Although the impact of genetic variation in the TPMT gene on treatment outcome and toxicity has been demonstrated, the role of other polymorphisms remains less well known. WHAT THIS STUDY ADDS • New information on the allelic variation of these three enzymes (XO, TPMT and ITPA) and their influence on 6-MP/AZA metabolism and toxicity. • Confirmation of the association of TPMT polymorphism with haematological toxicity. • Identified potential genetic characteristics that may contribute to higher risk of adverse events (such as ITPA IVS2+21A→C mutation). AIMS - To examine the allelic variation of three enzymes involved in 6-mercaptopurine/azathioprine (6-MP/AZA) metabolism and evaluate the influence of these polymorphisms on toxicity, haematological parameters and metabolite levels in patients with acute lymphoblastic leukaemia (ALL) or inflammatory bowel disease (IBD). METHODS - Clinical data and blood samples were collected from 19 ALL paediatric patients and 35 IBD patients who were receiving 6-MP/AZA therapy. All patients were screened for seven genetic polymorphisms in three enzymes involved in mercaptopurine metabolism [xanthine oxidase, inosine triphosphatase (C94→A and IVS2+21A→C) and thiopurine methyltransferase]. Erythrocyte and plasma metabolite concentrations were also determined. The associations between the various genotypes and myelotoxicity, haematological parameters and metabolite concentrations were determined. RESULTS - Thiopurine methyltransferase variant alleles were associated with a preferential metabolism away from 6-methylmercaptopurine nucleotides (P = 0.008 in ALL patients, P = 0.038 in IBD patients) favouring 6-thioguanine nucleotides (6-TGNs) (P = 0.021 in ALL patients). Interestingly, carriers of inosine triphosphatase IVS2+21A→C variants among ALL and IBD patients had significantly higher concentrations of the active cytotoxic metabolites, 6-TGNs (P = 0.008 in ALL patients, P = 0.047 in IBD patients). The study confirmed the association of thiopurine methyltransferase heterozygosity with leucopenia and neutropenia in ALL patients and reported a significant association between inosine triphosphatase IVS2+21A→C variants with thrombocytopenia (P = 0.012). CONCLUSIONS - Pharmacogenetic polymorphisms in the 6-MP pathway may help identify patients at risk for associated toxicities and may serve as a guide for dose individualization.

A comparison of the provision of the My Choice Weight Management Programme via general practitioner practices and community pharmacies in the United Kingdom

This study aimed to assess the effectiveness of a novel, community-based weight management programme delivered through general practitioner (GP) practices and community pharmacies in one city in the United Kingdom. This study used a non-randomized, retrospective, observational comparison of clinical data collected by participating GP practices and community pharmacies. Subjects were 451 overweight or obese men and women resident in areas of high socioeconomic deprivation (82% from black and minority ethnic groups, 86% women, mean age: 41.1 years, mean body mass index [BMI]: 34.5 kg m−2). Weight, waist circumference and BMI at baseline, after 12 weeks and after 9 months were measured. Costs of delivery were also analysed. Sixty-four per cent of participants lost weight after the first 12 weeks of the My Choice Weight Management Programme. There was considerable dropout. Mean percentage weight loss (last observation carried forward) was 1.9% at 12 weeks and 1.9% at final follow-up (9 months). There was no significant difference in weight loss between participants attending GP practices and those attending pharmacies at both 12 weeks and at final follow-up. Costs per participant were higher via community pharmacy which was attributable to better attendance at sessions among community pharmacy participants than among GP participants. The My Choice Weight Management Programme produced modest reductions in weight at 12 weeks and 9 months. Such programmes may not be sufficient to tackle the obesity epidemic.

Onions:a global benefit to health

Onion (Allium cepa L.) is botanically included in the Liliaceae and species are found across a wide range of latitudes and altitudes in Europe, Asia, N. America and Africa. World onion production has increased by at least 25% over the past 10 years with current production being around 44 million tonnes making it the second most important horticultural crop after tomatoes. Because of their storage characteristics and durability for shipping, onions have always been traded more widely than most vegetables. Onions are versatile and are often used as an ingredient in many dishes and are accepted by almost all traditions and cultures. Onion consumption is increasing significantly, particularly in the USA and this is partly because of heavy promotion that links flavour and health. Onions are rich in two chemical groups that have perceived benefits to human health. These are the flavonoids and the alk(en)yl cysteine sulphoxides (ACSOs). Two flavonoid subgroups are found in onion, the anthocyanins, which impart a red/purple colour to some varieties and flavanols such as quercetin and its derivatives responsible for the yellow and brown skins of many other varieties. The ACSOs are the flavour precursors, which, when cleaved by the enzyme alliinase, generate the characteristic odour and taste of onion. The downstream products are a complex mixture of compounds which include thiosulphinates, thiosulphonates, mono-, di- and tri-sulphides. Compounds from onion have been reported to have a range of health benefits which include anticarcinogenic properties, antiplatelet activity, antithrombotic activity, antiasthmatic and antibiotic effects. Here we review the agronomy of the onion crop, the biochemistry of the health compounds and report on recent clinical data obtained using extracts from this species. Where appropriate we have compared the data with that obtained from garlic (Allium sativum L.) for which more information is widely available. Copyright © 2002 John Wiley & Sons, Ltd.

A new algorithm for the relationship between vision and ametropia

Refraction simulators used for undergraduate training at Aston University did not realistically reflect variations in the relationship between vision and ametropia. This was because they used an algorithm, taken from the research literature, that strictly only applied to myopes or older hyperopes and did not factor in age and pupil diameter. The aim of this study was to generate new algorithms that overcame these limitations. Clinical data were collected from the healthy right eyes of 873 white subjects aged between 20 and 70 years. Vision and refractive error were recorded along with age and pupil diameter. Re-examination of 34 subjects enabled the calculation of coefficients of repeatability. The study population was slightly biased towards females and included many contact lens wearers. Sex and contact lens wear were, therefore, recorded in order to determine whether these might influence the findings. In addition, iris colour and cylinder axis orientation were recorded as these might also be influential. A novel Blur Sensitivity Ratio (BSR) was derived by dividing vision (expressed as minimum angle of resolution) by refractive error (expressed as a scalar vector, U). Alteration of the scalar vector, to account for additional vision reduction due to oblique cylinder axes, was not found to be useful. Decision tree analysis showed that sex, contact lens wear, iris colour and cylinder axis orientation did not influence the BSR. The following algorithms arose from two stepwise multiple linear regressions: BSR (myopes) = 1.13 + (0.24 x pupil diameter) + (0.14 x U) BSR (hyperopes) = (0.11 x pupil diameter) + (0.03 x age) - 0.22 These algorithms together accounted for 84% of the observed variance. They showed that pupil diameter influenced vision in both forms of ametropia. They also showed the age-related decline in the ability to accommodate in order to overcome reduced vision in hyperopia.

Eye alignment:a novel approach

This thesis describes the design and development of an eye alignment/tracking system which allows self alignment of the eye’s optical axis with a measurement axis. Eye alignment is an area of research largely over-looked, yet it is a fundamental requirement in the acquisition of clinical data from the eye. New trends in the ophthalmic market, desiring portable hand-held apparatus, and the application of ophthalmic measurements in areas other than vision care have brought eye alignment under new scrutiny. Ophthalmic measurements taken in hand-held devices with out an clinician present requires alignment in an entirely new set of circumstances, requiring a novel solution. In order to solve this problem, the research has drawn upon eye tracking technology to monitor the eye, and a principle of self alignment to perform alignment correction. A handheld device naturally lends itself to the patient performing alignment, thus a technique has been designed to communicate raw eye tracking data to the user in a manner which allows the user to make the necessary corrections. The proposed technique is a novel methodology in which misalignment to the eye’s optical axis can be quantified, corrected and evaluated. The technique uses Purkinje Image tracking to monitor the eye’s movement as well as the orientation of the optical axis. The use of two sets of Purkinje Images allows quantification of the eye’s physical parameters needed for accurate Purkinje Image tracking, negating the need for prior anatomical data. An instrument employing the methodology was subsequently prototyped and validated, allowing a sample group to achieve self alignment of their optical axis with an imaging axis within 16.5-40.8 s, and with a rotational precision of 0.03-0.043°(95% confidence intervals). By encompassing all these factors the technique facilitates self alignment from an unaligned position on the visual axis to an aligned position on the optical axis. The consequence of this is that ophthalmic measurements, specifically pachymetric measurements, can be made in the absence of an optician, allowing the use of ophthalmic instrumentation and measurements in health professions other than vision care.

Health related quality of life and perception of stigmatisation in adolescents living with sickle cell disease in Nigeria:a cross sectional study

Background: Sickle cell disease impacts the physical, emotional and psychological aspects of life of the affected persons, often times exposing them to disease associated stigma from the society and alters the health related quality of life (HRQoL). This study compared the HRQoL of adolescents with sickle cell disease with their healthy peers, identified socio-demographic and clinical factors impacting HRQoL, and determined the extent and effects of SCD related stigma on quality of life. Procedure: We conducted a cross-sectional survey among 160 adolescents, 80 with SCD and 80 adolescents without SCD. Socio-demographic and clinical data were collected using a pre-tested questionnaire. HRQoL was investigated using the Short Form (SF-36v2) Health Survey. SCD perceived stigma was measured using an adaptation of a perceived stigma questionnaire. Results: Adolescents with SCD have significantly worse HRQoL than their peers in all of the most important dimensions of HRQoL (physical functioning, physical roles limitation, emotional roles limitation, social functioning, bodily pain, vitality and general health perception) except mental health. Recent hospital admission and SCD related complication further lowered HRQoL scores. Over seventy percent of adolescents with SCD have moderate to high level of perception of stigmatisation. Hospitalisation, SCD complication, SCD stigma were inversely, and significantly associated with HRQoL. Conclusions: Adolescents living with SCD in Nigeria have lower health related quality of life compared to their healthy peers. They also experience stigma that impacts their HRQoL. Complications of SCD and hospital admissions contribute significantly to this impairment. Pediatr Blood Cancer 2015;62:1245-1251.