The PhD program developed a novel tracking model to assess the stability of diet over time. The ‘Two-stage Estimation Model for Tracking Analysis’ (TEMTA) overcomes difficulties of previous tracking methods and can be widely applied to future epidemiological studies. The model was applied to the MRC National Survey of Health and Development (NSHD) to track fruit, vegetable and fat intake from ages 4-36 years and 36-43 years. Physical activity and under-reporting were also tracked for NSHD participants from ages 36-43 years old. Diet, especially vegetable and fat intakes, were shown not to track from childhood to adulthood. Diet and physical activity were tracking in adulthood but under-reporting was not. In addition, an accumulative life course model was applied to investigate the longitudinal patterns of fruit consumption and obesity. Results suggest the importance of keeping a healthy diet from an early age in life. It is indicated from this study and recent healthy diet can have immediate benefit to health. Discussions and recommendations are made in terms of the advantages and limitations of the TEMTA model, research and public health implications suggested by the tracking analysis in NSHD, and further research opportunities presented by the work.

Three annual cross-sectional surveys between 2001 and 2003 were carried out to determine the extent of all forms of tobacco usage as well as the combined risk of malnutrition (under- and over-nutrition) in adult Bangladeshis. A total of 35,446 individuals, 18 years and above, from both urban and rural areas took part of whom 54.3% were female and 51% were rural dwellers. Data on socio-demographic background and knowledge about the harmful effects of tobacco were collected through a structured questionnaire. In the second and third year, height, weight and waist circumference were measured and body mass index (BMI), waist-to-height ration (WtHR) and conicity index (Cindex) were determined. Nutritional status was defined by the World Health Organization (WHO) BMI cut-offs (< 18.5, 18.5-24.99, 25-29.99 and > 30 for underweight, normal, pre-obese and obesity, respectively) and Asian cut-offs (< 18.5, 18.5-22.99, 23-27.49 and > 27.5 for underweight, low, moderate and high). Cut-offs for central obesity were also used: for waist circumference the international cut-offs for males were > 94cm and > 102cm and females, 80cm and > 88cm (reflecting high and very high levels). There was no increase in tobacco prevalence over the three year study and 37.5% of the sample used at least one form of tobacco. The overall prevalences of smoking, chewing tobacco and gul usage were 20.5%, 20.6% and 1.8% respectively. In addition 8% and 4.5% were past and occasional smoker, respectively. Current smoking and gul usages were significantly higher in males (42.2% and 2.2%, respectively) than females (2.3% and 1.5% respectively) while chewing tobacco was more common in females (21.6%) than males (19.4%). No significant locality difference was observed in smoking rate after adjusting for socio-demographic variables while chewing tobacco was 1.5 times more likely to occur in rural residents and gul usage was 3.6 times more likely to occur in urban residents. Smoking, chewing tobacco and gul usage all decreased with better educational status. The overall mean BMI, WC, WtHR and Cindex were 20.83, 74 cm, 0.48 and 1.20, respectively with higher means in urban residents; no sex difference was found in mean BMI and WC. One-quarter of sample were chronically energy deficient (CED) and all grades of CED were more common in rural residents, manual labourers and farmers and those who had no schooling. Using the WHO cut-offs 59% of the sample were at some risk from either underweight, overweight and smoking which increased to 66% for any tobacco usage. When the Asian cut-offs were used there were substantial increases to 77% for underweight, overweight and smoking and 81% for underweight, overweight and tobacco usage.

Biomarkers can provide a near error-free assessment of the intake of certain nutrients: for example, protein intake can be inferred from urinary Nitrogen levels and energy intake from Doubly Labelled Water method measurement of energy expenditure. However, historical cohort studies have often not included such calibration methods, and biomarker derived measurements of nutrient intake which are of calibration standard are not available for most nutrients. We develop a Bayesian method for correcting the value of the regression coefficient which measures the strength of the relationship between a disease related response variable and a dietary covariate. Calibration is implemented using informative prior distributions on the value of certain error-related model parameters; these informative priors are inferred from other studies which have included nutrient intake measurements of calibration quality. The model is successfully applied to cross-sectional and longitudinal epidemiological datasets. A limitation of these models is that where measurement error has caused a statistically significant relationship between the response variable and a dietary covariate to appear insignificant, the model is not able to recover the true significance of the relationship, and performs in this respect no better than an ordinary linear regression of the response variable on the uncorrected nutrient intake covariates. We conjecture that this a general limitation of measurement error correcting methods without internal calibration data.

No description available.

The methods developed in the thesis are largely motivated by one particular application. The function of transplanted lungs often deteriorates through the years after the transplant, usually attributed to a condition known as bronchiolitis obliterans syndrome. The presence of the underlying disease can only be strictly confirmed by an invasive biopsy procedure. Therefore, in practice, lung function is determined by repeated measurements of forced expiratory volume in one second. A set of data from Papworth Hospital's transplant programme is used to illustrate many of the models. Staged progression can be described using multi-state models on Markov processes. These are initially described in contexts where each patient's true status is known at a series of observation times. Extensions to these models are presented for situations where the underlying disease stage is not observed. These hidden Markov models can either be based on repeated observations of a disease stage with error, or repeated observations of the continuous biomarker. Methods illustrated for fitting these models to data include maximum likelihood and Bayesian estimation. The main contrasting approach directly models the evolution of the continuous process through time using longitudinal markers. A general model is presented, which can include random effects growth curves, continuous stochastic evolution, and semi-parametric estimation of marker trajectories. Various special cases of this model are illustrated on the lung function marker data. Finally, a hierarchical Bayesian model is developed which can encompass different underlying patterns of disease evolution among a set of patients. It is based on finite population mixtures. In the illustrative application, disease onset can either occur suddenly, inducing change-point in the biomarker trajectory, or smoothly, described by a linear model for the marker. Markov chain Monte Carlow techniques allow the simultaneous Bayesian estimation of all unknown quantities in the hierarchy.

Statistical process control methods were originally implemented in the industrial context. With increasing interest in the measurement and comparison of health outcomes, quality control tools are now being applied to medical data. However, outcomes measured on patients may have greatly differing associated risks, making standard quality control tools often inappropriate. Nevertheless, if patient risk an be adequately explained by a set of measurable patient covariates, specially developed statistical monitoring tools can be employed that take the risk into account. A comprehensive discussion of risk-adjusted quality control charts and methods is given, the theoretical form of existing and developed methods being described, as well as issues concerning considerations of design and enhancements to the methods. With a focus on discrete data types and particular case-mix structures, the charts are compared under various optimality criteria and applied to some example datasets. Multivariate risk-adjusted charts are also discussed in depth and the particular problem of parallel variables addressed via an example. Estimation of the level of a process throughout monitoring, and, most importantly, following signal of a chart, is of especial interest here. The exponentially weighted moving average (EWMA) chart is the chart seeming to be most suited to the estimation of level, but use of the EWMA as a monitoring tool is thought to be more approachable from a Bayesian standpoint. The Bayesian origin of the EWMA as an estimator, or smoother, of process level is described in detail. Similar Bayesian models are also described and related to the EWMA. Based upon the discussed models, a possible Bayesian monitoring scheme that produces an and estimate of process level as a by-product is developed and a demonstration of its application given.

No description available.

The emergence of distributed applications operating on large-scale, heterogeneous and decentralised networks poses new and challenging problems of concern to society as a whole, in particular for data security, privacy and confidentiality.  Trust management and authorisation policy language have been proposed to address access control and authorisation in this context. Still, many key problems have remained unsolved. Existing systems are often not expressive enough, or are so expressive that access control becomes undecidable; their semantics is not formally specified; and they have not been shown to meet the requirements set by actual real-world applications. This dissertation addresses these problems. We present CASSANDRA, a role-based language and system for expressing authorisation policy, and the results of a substantial case study, a policy for a national electronic health record (HER) system, based on the requirements of the UK National Health Service’s National Programme for Information Technology (NPfIT). CASSANDRA policies are expressed in a language derived from Datalog with constraints. CASSANDRA supports credential-based authorisation (e.g. between administrative domains), and rules can refer to remote policies (for credential retrieval and trust negotiation). The expressiveness of the language (and its computational complexity) can be tuned by choosing an appropriate constraint domain. The language is small and has a formal semantics for both query evaluation and the access control engine. There has been a lack of real-world examples of complex security policies: our NPfIT case study fills this gap. The resulting CASSANDRA policy (with 375 rules) demonstrates that the policy language is expressive enough for a real-world application. We thus demonstrate that a general-purpose trust management system can be designed to be highly flexible, expressive, formally founded and meet the complex requirements of real-world applications.

Childhood diseases remain an important public health issue, particularly in developing countries where they are still a significant cause of mortality. It is only through understanding the underlying mechanisms of such diseases that we can hope to control, and ultimately eradicate them. Here I attempt to address the epidemiological effects of spatial scale, from within-city through to pan-European dynamics. I also consider which factors are the major causes of the variabilities observed in the dynamics of childhood diseases, focusing on demography in particular. Questions of space (be they spatial scale or spatial heterogeneities) are of increasing interest in ecology, from individual interactions through to the population level and beyond to metapopulation studies. Space may also be of considerable importance in epidemiological studies, since disease prevention and control are highly dependent upon population size and the proximity to other susceptible individuals. This thesis presents a review of research in these areas (chapter one), analysis of within-city infection dynamics (in particular measles notifications in London boroughs, chapter two), and comparisons of these data with equivalent data from cities in England and Wales. These results are then compared with findings from both stochastic and deterministic models (chapter three). Chapter four introduces novel Scottish city data and employs a time-series SIR (susceptible, infectious, recovered) model to compare the various datasets. The study area is then widened still further and pan-European measles and whooping cough data are considered, with particular attention paid to the effects of variations in the underlying demographic parameters, alongside the potential for population dynamic interference between the two diseases (chapter five). These phenomena are then assessed with stochastic and deterministic models (chapter six). Chapter seven then unifies the data and the models, and suggests possible extensions of the model. Finally, chapter eight reviews the thesis and proposes possible avenues for future research.

The specific aim of this thesis was to combine data on mutation prevalence and risk from both high risk families and population based series, in order to develop a model for familial breast and ovarian cancer which incorporates both the effects of BRCA1, BRCA2 and other genes. The principal methodology used was segregation analysis and the genetic models were constructed using the computer program MENDEL. The first dataset consisted of 112 families containing two or more relatives with epithelial ovarian cancer. BRCA1 and BRCA2 germline mutations were detected in 50% of these families. When the effects of BRCA1, BRCA2 and a third ovarian cancer susceptibility gene were modelled simultaneously none of the models fitted gave significant evidence for a third gene. BRCA1 and BRCA2 were estimated to account for at least 38 % of the excess familial risk of ovarian cancer. Using data on the families of twelve BRCA1 mutation carriers in a study of 374 ovarian cancer cases unselected for family history, the estimated ovarian cancer risk by age 70 was 66% and the corresponding breast cancer risk was 45%. The breast cancer dataset consisted of 1484 women diagnosed with breast cancer under the age 55 from whom blood samples were analysed for mutations in BRCA1 and BRCA2. Using information on breast and ovarian cancer history in first degree relatives, and on the mutation status of the index cases we estimated the effects of BRCA1, BRCA2, a third gene BRCA3 and a polygenic effect. For this purpose the Hypergeometric Polygenic model was implemented in MENDEL.