This research has emanated from the geographical concerns raised by organisational change in the British National Health Service (NHS), namely the ongoing debate relating to health and health care inequalities. This thesis develops a flexible, portable and predictive model of health care utilization capable of assisting improved health care planning and analysis. In so doing it contributes to the current resurgence in medical geography. An applied approach to this research is identified which builds upon methods of modelling spatial patterns and processes in geography and the upsurge of interest in Geographical Information Systems (GIS) technology. In these terms, the use of GIS is central to the research; it supports construction and application of the model; facilitates a wide range of analyses; and provides a basis for visualisation and interpretation of model results. The value of modelling in analysing relationships between health inequalities and the location and allocation of health care is identified through a discussion of previous NHS policy initiatives and previous research. From this, a conceptual model of utilization is developed which incorporates components of need, accessibility and provision. A patient survey of asthmatics and diabetics informs the development of the model and validates the choice of indicators used to measure utilization. Indicators of need, accessibility and utilization are thus defined and subsequently measured using a signed chi-square scoring method. The model was developed and tested for primary care General Practitioner services in the Northampton District Health Authority area and outcome measures are proposed and evaluated. Rigorous testing of the model’s sensitivity and robustness is undertaken and potential for its simplification explored. Components are critically evaluated through a comparison with alternative methods of determining spatial inequalities in disadvantage. The potential of the model of utilization for health care planning and analysis is extensively demonstrated through the application of a variety of modelled scenarios. Emergent issues from the research are considered and potential for future geographical research in this area of study, and the impact upon research agendas more generally, is explored

Infant growth is a key indicator of health and a relevant component of paediatric surveillance. Certain growth characteristics are also associated with greater risk for diseases such as obesity and cardiovascular disease. South Asian populations are known to demonstrate poor infant growth and suffer from a high prevalence of non-communicable disease. Relatively little is known about the growth of Pakistani infants, especially following migration. In the United kingdom (UK), infant growth is routinely monitored to detect poor health, and this process produces a repository of largely unutilised data. In 2009, new growth charts, which include a component of the World Health Organisation (WHO) growth standards, were introduced to routine practice. The adoption of prescriptive standards, which are based on breastfed infants living in an unconstrained environment, will have implications for the assessment of growth. To develop and assess the quality of routine growth monitoring data collected in Bradford, UK, so that it can be used to describe the differences in growth between White British and Pakistani infants in the same city. To investigate the factors that influence this growth. To assess the implications of adopting growth standards for practice. The frequency of routine growth monitoring data that are collected at prescribed age periods was assessed. Test-retest growth data were collected from 192 practitioners, and technical error of measurements were calculated. Data on 2464 (boys 51%, White British 45%) infants were submitted to multilevel modelling analysis to produce sex and ethnic specific weight-for-age, abdominal circumference-for-age, head circumference-for-age, and length-for-age growth curves between birth and nine months. Multivariable linear regression models were used to investigate factors that influence size at birth and at nine months. Growth curves were plotted against the WHO standards and the UK 1990 references, Z-scores were calculated, and the relative risks (RR) of underweight, obesity, and poor infant weight gain using the standards compared to the references were assessed. During each prescribed age period for routine growth monitoring generally only 30% to 35% of measurements were recorded. None of the technical error of measurements were excessively large, and coefficients of reliability ranged from 0.96 to 1.00. Multilevel models explained that Pakistani infants were smaller than White British infants, in the first nine months of life, for weight (-210.3g to -321.7g), abdominal circumference (-1.15cm to -0.39cm), head circumference (-0.59cm), and length (-0.32cm). Compared to the WHO standards, infants demonstrated dissimilar weight growth, but similar head circumference and length growth. The common weight growth pattern was slow growth between birth and two months, followed by rapid growth. Using the standards, infants were significantly less likely to be classified as underweight (RR at birth 0.496; 95% Confidence Interval 0.363 to 0.678) and demonstrating poor weight gain from birth to nine months (0.783; 0.644 to 0.952). Growth monitoring data are not collected at prescribed age periods, but following initial training of practitioners are reliable. Integrating research with practice has developed routine data to research calibre and has established protocols to make data more accessible. Pakistani infants were consistently smaller than White British infants, and, despite efforts, the determinants of this phenomenon have not yet been fully elucidated. Growth in weight of infants in Bradford differs significantly from that represented by the WHO standards, and without adequate training of practitioners infant growth may be incorrectly interpreted.

Diabetes has become one of the most severe deceases due to an increasing number of diabetes patients globally. A large amount of digital data on diabetes has been collected through various channels. How to utilize these data sets to help doctors to make a decision on diagnosis, treatment and prediction of diabetic patients poses many challenges to the research community. The thesis investigates mathematical models with a focus on neural networks for large scale diabetes data modelling and analysis by utilizing modern computing technologies such as grid computing and cloud computing. These computing technologies provide users with an inexpensive way to have access to extensive computing resources over the Internet for solving data and computationally intensive problems. This thesis evaluates the performance of seven representative machine learning techniques in classification of diabetes data and the results show that neural network produces the best accuracy in classification but incurs high overhead in data training. As a result, the thesis develops MRNN, a parallel neural network model based on the MapReduce programming model which has become an enabling technology in support of data intensive applications in the clouds. By partitioning the diabetic data set into a number of equally sized data blocks, the workload in training is distributed among a number of computing nodes for speedup in data training. MRNN is first evaluated in small scale experimental environments using 12 mappers and subsequently is evaluated in large scale simulated environments using up to 1000 mappers. Both the experimental and simulations results have shown the effectiveness of MRNN in classification, and its high scalability in data training. MapReduce does not have a sophisticated job scheduling scheme for heterogonous computing environments in which the computing nodes may have varied computing capabilities. For this purpose, this thesis develops a load balancing scheme based on genetic algorithms with an aim to balance the training workload among heterogeneous computing nodes. The nodes with more computing capacities will receive more MapReduce jobs for execution. Divisible load theory is employed to guide the evolutionary process of the genetic algorithm with an aim to achieve fast convergence. The proposed load balancing scheme is evaluated in large scale simulated MapReduce environments with varied levels of heterogeneity using different sizes of data sets. All the results show that the genetic algorithm based load balancing scheme significantly reduce the makespan in job execution in comparison with the time consumed without load balancing.

The concentration on healthcare information technology has never been determined than it is today. This awareness arises from the efforts to accomplish the extreme utilization of Electronic Health Record (EHR). Due to the greater mobility of the population, EHR will be constructed and continuously updated from the contribution of one or many EPRs that are created and stored at different healthcare locations such as acute Hospitals, community services, Mental Health and Social Services. The challenge is to provide healthcare professionals, remotely among heterogeneous interoperable systems, with a complete view of the selective relevant and vital EPRs fragments of each patient during their care. Obtaining extensive EPRs at the point of delivery, together with ability to search for and view vital, valuable, accurate and relevant EPRs fragments can be still challenging. It is needed to reduce redundancy, enhance the quality of medical decision making, decrease the time needed to navigate through very high number of EPRs, which consequently promote the workflow and ease the extra work needed by clinicians. These demands was evaluated through introducing a system model named SVSEPRS (Searching and Viewing Segments of Electronic Patient Records Service) to enable healthcare providers supply high quality and more efficient services, redundant clinical diagnostic tests. Also inappropriate medical decision making process should be avoided via allowing all patients‟ previous clinical tests and healthcare information to be shared between various healthcare organizations. Multidimensional data model, which lie at the core of On-Line Analytical Processing (OLAP) systems can handle the duplication of healthcare services. This is done by allowing quick search and access to vital and relevant fragments from scattered EPRs to view more comprehensive picture and promote advances in the diagnosis and treatment of illnesses. SVSEPRS is a web based system model that helps participant to search for and view virtual EPR segments, using an endowed and well structured Centralised Multidimensional Search Mapping (CMDSM). This defines different quantitative values (measures), and descriptive categories (dimensions) allows clinicians to slice and dice or drill down to more detailed levels or roll up to higher levels to meet clinicians required fragment.

Coronary heart disease rates show substantial geographic variation in England, which could be due to environmental variables (e.g. climate, air quality) or behavioural risk factors for coronary heart disease within populations. Previous work investigating this geographic variation has either used ecological analysis (i.e. areas as units of observation) or individual-level analysis. Ecological studies have been unable to account adequately for differences in behavioural risk factors within populations; individual-level studies have been under-powered at the area-level to include all potentially explanatory environmental variables. This thesis reports on ecological multi-level and spatial error regression analyses of coronary heart disease mortality and hospitalisation rates for all wards in England using environmental variables and synthetic estimates of the prevalence of behavioural risk factors as explanatory variables. Existing sets of synthetic estimates were subjected to studies of their validity. Validated synthetic estimates of the prevalence of smoking, low fruit and vegetable consumption, raised blood pressure, obesity and raised cholesterol were combined into a single index of unhealthy lifestyle to take account of collinearity between them. Final models successfully explained around 80% of large scale geographic variation (i.e. variation between wards in different areas of the country) in mortality rates for coronary heart disease and 60% in hospitalisation rates, and around 20% of the small scale geographic variation (i.e. variation between wards in close proximity) in mortality rates, and 30% in hospitalisation rates. The climate explained around 15% of large scale geographic variation in coronary heart disease rates after adjustment for the index of unhealthy lifestyle and socioeconomic deprivation. Urbanicity and air pollution explained a small amount of small scale geographic variation in coronary heart disease rates. The majority of explained geographic variation was due to the index of unhealthy lifestyle and deprivation. The results of this thesis confirm and extend findings from the British Regional Heart Study, report on the validity of synthetic estimates currently used to guide healthcare resource allocation, and introduce an index of unhealthy lifestyle that could be used in future ecological studies of chronic disease.

Prostate cancer has emerged as the most frequently diagnosed cancer, except for non-melanoma skin cancer, among men in many Western countries in the last decade. In the United Kingdom (UK), prostate cancer accounts for nearly a quarter of all new male cancer diagnoses. Increasing age and some genetic and ethnic risk factors have been identified but few modifiable risk factors are known. The introduction of Prostate Specific Antigen (PSA) testing has increased the detection of previously undiagnosed disease but its contribution to the observed increases in prostate cancer incidence is not clear. Considerable variations in the incidence of prostate cancer have been observed in different geographic regions and socio-economic groups across the UK but it is not known whether, or to what extent, these may be attributed to differential uptakes of PSA testing. Prostate cancer is the third most common cause of cancer death in men but many cases do not progress. There is therefore an important clinical need for better prognostic markers so that the increasing numbers of men with prostate cancer can be appropriately managed. This thesis begins with a descriptive epidemiological study using cancer registry incidence data from the West of Scotland from 1991 to 2007. The aim was to determine whether the incidence of prostate cancer was continuing to rise and to describe any demographic or socio-economic patterns that might suggest particular at-risk groups. To understand whether any socio-economic differentials in incidence might be due to PSA testing, I examined Gleason grade-specific prostate cancer incidence by socio-economic groups over time. Socio-economic circumstances were measured using census-derived Carstairs scores. Overall (age adjusted) prostate cancer incidence increased by 70% from 44 per 100,000 in 1991 to 75 per 100,000 in 2007, an average annual growth of 3.59%. This pattern was driven by significant increases in both low and high grade cancers with no convincing change in their proportions over time. Incidence was inversely associated with deprivation with the highest rates among the most affluent groups. To explore the role of potentially modifiable risk factors on prostate cancer incidence, the Midspan and Collaborative prospective cohort studies were analysed. An analysis of the relationship between cholesterol and prostate cancer incidence was conducted on the Midspan cohort, which comprises 12,926 men who were enrolled between 1970 and 1976 and followed up to 31st December 2007. Cox Proportional Hazards Models were used to evaluate the association between baseline plasma cholesterol and Gleason grade-specific prostate cancer incidence. Following up to 37 years’ follow-up, 650 men developed prostate cancer. Their baseline plasma cholesterol level was positively associated with hazard of high grade (Gleason score ≥8) prostate cancer incidence (n=119). The association was greatest among men in the 4th highest quintile for cholesterol, 6.1 to <6.69 mmol/l, Hazard Ratio 2.28, 95% CI 1.27 to 4.10, compared with the baseline of <5.05 mmol/l. This association remained significant after adjustment for age, body mass index, smoking and socio-economic status. Evidence on the possible role of tea and coffee consumption in the development of prostate cancer remains limited to a small number of studies with short follow-up and small numbers of cases. Therefore to understand the relationship of tea and coffee consumption with overall as well as grade-specific prostate cancer, a prospective cohort study of 6016 men was carried out, who were enrolled in the Collaborative cohort study between 1970 and 1973 and followed up to 31st December 2007. Three hundred and eighteen men developed prostate cancer in up to 37 years’ follow-up. I found a positive association between consumption of tea and overall risk of prostate cancer incidence (p=0.02). The association was greatest among men who drank ≥7 cups of tea per day (HR 1.50, 95% CI 1.06 to 2.12) compared with the baseline of 0-3 cups per day. However, I did not find any significant association between tea intake and low (Gleason < 7) or high grade (Gleason 8-10) prostate cancer incidence. Higher coffee consumption was inversely associated with risk of high grade disease (HR 0.46, 95% CI 0.21-0.99) but not with overall risk of prostate cancer. These associations remained significant after adjustment for age, Body Mass Index, smoking, social class, cholesterol level, systolic blood pressure and alcohol consumption. Although survival of prostate cancer patients has improved over time, little is known about the major prognostic factors. To understand the socio-economic differences and major determinants of survival, an investigation was carried out using cancer registry incidence data from the West of Scotland from 1991 to 2007, linked with General Registrar Office (Scotland) death records up to 31st December 2008. Socio-economic circumstances were measured using the Scottish Index for Multiple Deprivation (SIMD). Age, sex and deprivation specific mortality rates were obtained from General Registrar Office for Scotland (GRO(S)). One, three and five year relative survival was estimated using the complete approach. Survival gradients across deprivation quintiles were estimated using linear regression, weighted by the variance of the relative survival estimate, using STATA software (StataCorp, version 11). Five year relative survival increased from 58.2% to 78.6% in men over the same period (an average deprivation adjusted increase of 10.2% between six years periods). Despite substantial improvements in survival of prostate cancer patients, there was a deprivation gap (that is, better survival for the least deprived compared with the most deprived) between the three time periods. The deprivation gap in five year relative survival widened from -4.76 in 1991-1996 to -10.08 in 2003-2007. Age, Gleason grade and socio-economic status appeared as significant determinants of survival. There is some evidence that systemic inflammation may be associated with survival in patients with prostate cancer although its relationship to tumour grade and socio-economic circumstances has not been previously studied. I therefore investigated the association between inflammation-based prognostic scores and survival, using the modified Glasgow Prognostic Score (mGPS) and Neutrophil Lymphocyte Ratio (NLR) as well as Gleason grade. The patient cohort within the Glasgow Inflammation Outcome Study who had a diagnosis of prostate cancer was included in this study. The mGPS is a categorical score constructed by combining serum C-reactive protein and albumin levels, while the NLR is obtained by calculating the ratio of neutrophils to lymphocytes. The relationship between mGPS and NLR and five-year relative survival was explored after adjusting for age, socio-economic circumstances and Gleason grade. Of the 897 prostate cancer patients in the Glasgow Inflammation Outcome Study, 422 (47%) died during a maximum follow-up of 6.2 years. Systemic inflammation had a significant prognostic value. The mGPS predicted poorer 5-year overall and relative survival independent of age, socio-economic circumstances, disease grade and NLR. Raised mGPS also had a significant association with excess risk of death (mGPS 2: Relative Excess Risk = 2.08, 95% CI 1.13-3.81) among aggressive, clinically significant prostate cancer (Gleason score 8-10). Prostate cancer patients with a raised mGPS had significantly higher risks of death overall as well as for high grade disease. Inflammation-based prognostic scores can potentially predict patient outcome and a further prospective study is warranted to assess their clinical value.

Cerebrovascular disease, has an enormous, and increasing, impact on global health. As well as causing clinical stroke, cerebrovascular disease is thought to be a major contributor to cognitive decline and dementia. Socioeconomic status (SES) is associated with risk of stroke. Those in the lowest SES group are estimated to be at twice the risk of stroke compared to those in the highest SES group. Those with low SES may also have a more severe stroke and a poorer outcome. It is imperative that the extent and mechanism of this association is clarified. This thesis aims to determine if the association between SES and stroke is explained by a greater prevalence of traditional vascular risk factors amongst those of low SES. It also explains the link with a novel risk factor, poor oral health. Lastly it addresses the long-term cognitive outcome in older people at risk of vascular disease. A systematic review and meta-analysis was undertaken to establish if vascular risk factors explain the association between SES and stroke incidence / post-stroke mortality. This demonstrated that lower SES was associated with an increased risk of stroke and that a greater burden of vascular risk factors in those with low SES explained about 50% of the additional risk of stroke. However this meta-analysis could not clarify what vascular risk factors are most critical. Low SES was also associated with increased mortality risk in those who have a stroke although study results were heterogeneous and this link was not readily explained by known vascular risk factors. A prospective study of 467 consecutive stroke and transient ischameic attack (TIA) patients from three Scottish hospitals was undertaken with the aim of establishing whether those with low SES carry higher levels of vascular risk factors, have a more severe stroke and have equal access to stroke care services and investigations. Stroke / TIA patients with low SES were younger and more likely to be current smokers but there was no association with other vascular risk factors /co-morbidity. Those who had lower SES had a more severe stroke. The lowest SES group were less likely to have neuroimaging or an electrocardiogram although differences were not significant on multivariate analysis. There was however equal access to stroke unit care. A secondary analysis of a prospective cohort study of 412 stroke patients was conducted. The aim was to explore oral health after acute stroke and assess if poor oral health explains the association between SES and stroke. Dry mouth amongst acute stroke patients was very common, however there was no association between oral health and low SES. There was an association of dry mouth with pre-stroke disability and Urinary Tract Infection. There was also a link with oral Candida glabrata colonisation, although the clinical relevance of this is uncertain. In the acute phase after stroke there was no convincing association of dry mouth with dysphagia or pneumonia. Therefore there was no association between SES and poor oral health as measured in this study but oral health may still be part of the explanation of the association between SES and acute stroke and this needs further investigation. Vascular disease is an important contributor to cognitive decline and dementia. Low SES may be associated with an increased risk of cognitive decline in later life and vascular disease may be a mediating factor. More effective prevention of vascular disease may slow cognitive decline and prevent dementia in later life, particularly in low SES groups. Lipid lowering with statins might be effective in preventing dementia but so far evidence from randomised control trials does not show benefit from statins in preventing cognitive decline and dementia. However the duration of follow-up in these trials was short and there may be benefit in the long-term. My aim was therefore to establish if long-term follow-up of the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER) study was feasible. I found that it was feasible to follow-up 300 elderly survivors from the Scottish arm of the PROSPER study and the methods could be extended to the whole group. As expected nearly half of the PROSPER participants were dead. Additionally a large proportion of traceable participants had significant cognitive impairment. Smoking cessation, control of blood pressure and management of other vascular risk factors should be made a priority in areas of low SES. Additionally further research is needed to fully clarify the association between SES and stroke incidence. Avenues for exploration might include the possibilities of poorer access to effective stroke care, reduced uptake of care and poorer oral health in lower SES groups. In addition public health campaigns regarding smoking cessation should be directed at lower SES groups. I have shown that a large scale follow-up of the PROSPER participants is feasible and may determine new and novel risk factors for dementia and assess the long-term effect of a period of treatment with pravastatin.

Technologies such as geographical information systems (GIS) have emerged during the past two decades as part of the Information Revolution and include functions such as data storage, management, integration, analysis and presentation. GIS have wide and diverse applications in disciplines such as engineering, business/marketing, urban planning and environmental management but remain underused in public health. The thesis reports the findings of a mixed methods study examining the views and perceptions of public health practitioners on the value of GIS mapping in decision-making. A case study design was chosen; the case issue (childhood lead [Pb] exposure) represents an example of the "case" which is defined as 'decision- making in public health'. The exploratory phase of the study combined heterogeneous data to produce a visualisation of lead contamination in Newcastle. The value of GIS in public health was explored in an interview phase. Twenty-two semi-structured interviews were conducted with decision-makers involved either directly or indirectly in public health practice. Interview recordings were transcribed and coded thematically for analysis. Decision-makers tended to be positive about the use of GIS in public health and many volunteered potential opportunities to apply GIS mapping techniques further. Four discourses were highlighted through analyses, namely: data origins (Ontological Discourse), status (Power Discourse), application (Functionality Discourse) and reciprocity (Collaboration Discourse). The power of maps to integrate multiple, disparate datasets was found to be important and respondents felt, overall, that GIS mapping was a democratic means of communication. Complexity frameworks are drawn upon to make sense of the research findings and to illuminate the need for non-reductionist models of decision-making in the public health context. The lessons learnt through this study can be translated to other fields, thereby sharing skills, knowledge and experience to promote collaboration and integrated thinking across the public health landscape.

This thesis describes research into the facilitation of mediated communication of health updates and support needs across the social network, on behalf of individuals experiencing acute or chronic health problems. This led to the user-centred design, development and evaluation of a prototype software tool. Investigatory applied research was conducted with the parents of sick newborn infants who were (or had previously been) cared for in a Neonatal Unit, and their social networks of family, friends, colleagues and neighbours. The thesis makes contributions to knowledge within Social Networks, Health Informatics, Adaptive Systems and User Modelling. The user-centred research was conducted using a Grounded Theory approach, progressively focussing on developing themes. An iterative approach was taken to evaluation of the resulting theory. In the Social Networks domain, a novel, intuitive mechanism for capturing the membership and structure of an individual’s personal social network has been defined and developed, grounded in the work of evolutionary anthropologist Robin Dunbar. Use of the highly visual mechanism requires low levels of literacy and computer skills. It is cross-culturally applicable, and makes no prior assumptions about an individual’s relationships. In the domains of Health Informatics, Adaptive Systems and User Modelling, a model has been defined for adaptive information sharing across the personal social network. This model provides a number of new insights about information sharing choices made by an individual experiencing a health crisis (the ego) and their supporters (alters).

Background: Excess adiposity, diabetes mellitus, and ischaemic heart disease are common important causes of morbidity and premature mortality in postmenopausal women in the UK. A large amount of data exists on known risk factors for these conditions, and for risk factors men and women share there is little evidence to suggest sex-based differences. It has been suggested that factors unique to women (such as parity and breastfeeding) may also influence risk. The nature of the relationship between childbearing and these conditions remains to be clarified. In this thesis I explore the association between women’s childbearing histories and their adiposity, and risk of diabetes or ischaemic heart disease after the menopause, to provide evidence on the character, repeatability and public health relevance of the associations. Aim: To explore the hypothesis that childbearing (specifically parity and breastfeeding) is associated with women’s body weight and risk of excess adiposity, and also with women’s risk of diabetes mellitus, and ischaemic heart disease after the menopause. Methods: Data are analysed from a large population-based cohort of middle-aged UK women recruited in 1996 to 2001 (the Million Women Study) with complete childbearing information, and who had baseline anthropometry, and were followed for incident diabetes or ischaemic heart disease through repeat survey questionnaires, hospital admission records, and central registry databases. Results: In a large ethnically homogeneous population of postmenopausal UK women increasing parity was associated with an increase in BMI, however this increase was offset in women who breastfed. The associations between parity, breastfeeding and BMI were of a similar order of magnitude to established risk factors known to be associated with BMI, for example smoking, and physical activity. The associations between childbearing and women’s risk of diabetes mellitus after the menopause appear to be largely due to the effects of childbearing on maternal BMI. There is only limited evidence to suggest a direct effect of childbearing on women’s risk of diabetes after the menopause. There is statistically significant evidence of an association between childbearing and women’s risk of ischaemic heart disease after the menopause. Parity was associated with a modest increase in risk whereas breastfeeding was associated with a small decrease in risk, however the effects were small in comparison to known important risk factors. Conclusions In a large population of UK women childbearing was found to have a persistent influence on women’s mean BMI after the menopause, and through this postmenopausal risk of diabetes mellitus. Childbearing was also found to be mod-estly associated with women’s risk of ischaemic heart disease after the menopause.