Objective: Little is known about health-related behaviours and their co-occurrence among male adolescents in Saudi Arabia. The main purpose of this study was to determine the prevalence of health related behaviours, and to investigate the associations between socio-demographic variables and health related behaviours and the clustering of health risk behaviours. Research Methods: A cross-sectional study using a self-completion anonymous questionnaire was undertaken between February and April, 2008. A stratified random sample of 1501 male adolescents was recruited from one private and public high school in each of the five districts in the city of Riyadh, Saudi Arabia. Main Outcome Measures: Prevalence and associations between health, social and demographic factors and health-related behaviours, including dietary behaviours, oral health, physical activity, smoking, violence, injuries and safety, and mental health factors. Results: The results of this study showed that only 24.2% and 39.9% of the students consumed fruit and vegetables on a daily basis (at least once every day), and only 7% and 13.7% ate fruit and vegetables 3 times or more every day. Also, only 52.4% consumed dairy products at least once every day and only 18.3% of the students consumed dairy products 3 times or more every day. 48.1% reported to not consume any fish products on any day of the week. About 48.7%, 60.2% and 25.2% of the students consumed sweets, soft drinks, and energy drinks at least once every day. The results of this study also showed that only 36.7% of students eat breakfast regularly (≥ 5 days per week). Eating breakfast regularly was positively associated with lower age, liking school, good academic performance, not eating high fat food every day, drinking soft drinks ≤ 1 time/day, drinking milk every day, low BMI, brushing teeth every day, physical activity ≥ 3 days/week, and not engaging in physical fights. Around half (51.3%) participants reported good teeth health status, 22.6% brushed their teeth two times daily, 29.7% brushed their teeth once daily, whereas 47.7% of the subjects do not brush their teeth daily and 54.3% never visited the dentist during the past year. About 29.5% of participants suffered teeth pain sometimes or most of the time and 16.4% missed some school days for this reason. Brushing teeth every day was positively associated with higher standards of parental education, attending private school, living district, good academic performance, liking school, visiting dentist during the last year, good teeth status, and not suffering from teeth pain. Only 18.4% of the students were physically active and only 65.2% participated in physical activity classes in schools. Regular physical activity was positively associated with younger age, liking school, good health status, lower BMI, father’s, mother’s, siblings’ and peers’ physical activity, not smoking, not fighting, not wanting to use drugs or alcohol, and not feeling lonely. Over a third (36.3%) of the participants were overweight or obese. A fifth (20.8%) of the adolescents were current smokers. 20.8% of the students were current smokers. Smoking among students was positively associated with higher age, studying in private school, poor health status, poor school performance, not liking school, father smoking, mother smoking, sibling smoking, peers smoking, low physical activity, wanting to use drugs and alcohol, carrying weapons, fighting, performing car drifting, and being abused by teachers. ii Over half the sample (55.5%) reported an injury, 21.8% had been threatened or injured by weapons. Just under half (49%) of the adolescents reported they were involved in a physical fight. Moreover, fighting among students was positively associated with the interaction of low parental education, not liking school, poor academic performance, skipping breakfast, low physical activity, current smoking, being threatened or injured by weapons, carrying weapons, joining people performing car drifting, bullying others, being abused by teachers. Carrying weapons during the last 30 days was reported by 36.6% of the sample. Carrying weapons was positively associated with higher age, not liking school, poor academic performance, current smoking, fighting, being threatened or injured by weapons, performing car drifting, joining people performing car drifting, taking part in bullying others, and being abused by family. Some (26.1%) of participants reported having been bullied and 24.6% of the students reported bullying others. Many of the adolescents reported being abused by a family member (34.4%) or one of their school teachers (39.5%) during the past 12 months preceding the survey. During this time period, many of the students reported feeling lonely (22.8%), feeling very worried about something that they could not sleep at night sometimes or more (27.0%), and feeling very sad or hopeless almost every day for two weeks or more (40%). About 14% of the participants in this study reported that they had wanted to use alcohol or drugs. A small but notable proportion (13.9%) of the participants reported that they had thought of attempting suicide and 6.9% had actually attempted suicide. Over a third (36.1%) of adolescents had performed car drifting 12 months preceding the survey. However, car drifting was positively associated with higher age, attending a private school, not liking school, poor academic performance, not brushing teeth every day, current smoking, wanting to use drugs and alcohol, carrying weapons, joining people who performing car drifting, bullying others, and attempting suicide. The majority (78.7%) of participants drove vehicles and 96% and 97.7% reported that they did not use a seat belt when doing so and did not use a seat belt when riding in a car as a passenger, respectively. Only 2.1% and 1.4% of participants wore a helmet when used motorized vehicle or nonmotorized. Conclusions and implications: The results of this study reveal that the adolescents engage in multiple health-risk behaviours, and these risk behaviours are relatively common among adolescents and cluster together. Health related behaviours are associated with several socio-demographic variables (age, father’s and mother’s education, school factors, health status and living districts), although not necessarily in the same order. However, the data emphasized the need for further quantitative and indepth qualitative research throughout Saudi Arabia, including other cities, rural communities, female adolescents, and other Middle Eastern countries. Cross-sectional research to gather evidence on youth health to collect population-based data on a range of health-related behaviours along with physical and social environments amongst school-age students and out-of-school youth facilities are important and highly needed to investigate health-related behaviours and associated risk factors and to measure change over time.

Introduction: Recent research has generated discussion upon the historically noted association between tuberculosis (TB) and diabetes mellitus (DM). However, evidence on the direction of the association, its magnitude, specificity and impact remains sparse. The primary aim of this thesis was to identify whether rates of TB (all sub-types, pulmonary (PTB) and extra pulmonary (EPTB)) are raised amongst those with DM (all sub-types, type 1 (T1DM) and type 2 (T2DM)), or the converse. Further to this, to estimate the magnitude and direction of any such associations. A secondary aim of the thesis was to investigate whether key TB outcomes differ amongst those with co-morbid DM and TB comparative to those with TB disease alone. Methods: The Oxford Record Linkage Study (ORLS) is a database containing records of all hospital admissions and all deaths (regardless of where they occurred), in defined populations within the former Oxford National Health Service Region. ORLS1 covers the years 1963 to 1998 and ORLS2 covers the years 1999 to 2008, the two databases are not linkable. The Health Improvement Network (THIN) is a database containing electronic medical records collected at General Practice clinics throughout the United Kingdom (UK). Retrospective cohort analyses were carried out using data from ORLS1, ORLS2 and THIN. All patients in the datasets were classified as exposed to (having had) or unexposed to (not having had) TB (all sub-types, PTB or EPTB) and exposed or unexposed to DM (all sub-types, T1DM or T2DM). In the ORLS1 and ORLS2 datasets, standardised rate ratios (RR) and corresponding 95% confidence intervals (95% CI) were calculated and compared for DM (all sub- types, T1DM and T2DM) in individuals who have had and have not had TB (all sub- types, PTB or EPTB). Similarly, RR and 95% CI were calculated and compared for TB (all sub-types, PTB or EPTB) in patients with and without DM (all sub-types, T1DM and T2DM). Within THIN datasets, the incidence rate ratio (IRR) and 95% CI of DM (all sub-types, T1DM and T2DM) were calculated using negative binomial regression, with TB (all sub-types, PTB or EPTB) as an explanatory variable. Similarly, the IRR and 95% CI for TB (all sub-types, PTB or EPTB) were calculated using negative binomial regression, with DM (all sub-types, T1DM or T2DM) as an explanatory variable. Systematic searching was performed to identify studies comparing TB outcomes amongst those with and without DM. Data from these studies were utilised to inform meta-analyses that assessed all cause mortality, bacterial clearance rate and TB relapse or recurrence rate amongst individuals with DM and co-morbid TB comparative to those with TB alone. Results Significant increases in TB rates (all sub-types) and pulmonary tuberculosis (PTB) rates were identified amongst individuals with DM comparative to those without DM within the Oxford Record Linkage Study datasets. The RR of TB (all sub-types) was increased amongst individuals with DM (all sub- types) compared to those without in ORLS1 (RR 1.77 (95% CI 1.45-2.15), P-value <0.001) and ORLS2 (RR 2.56 (95% CI 1.78-3.69), P-value <0.001). The RR of PTB was also increased amongst individuals with DM (all sub-types) compared to those without in ORLS1 (RR 1.72 (95% CI 1.22-2.37), P-value <0.001) and ORLS2 (RR 3.33 (95% CI 1.51-6.62), P-value 0.001). There was a statistically significant elevation of risk for TB amongst those with T2DM compared to those without in ORLS1 (RR 1.58 (95%CI1.15-2.14), P-value 0.003) and ORLS2 (RR 3.33 (95% CI 1.51-6.62), P-value 0.001). There was no significant association between the rates of TB amongst those with T1DM compared to those without in ORLS1. The ORLS 2 dataset was too small to complete this analysis. No significant association was found between the rate of EPTB amongst those with DM comparative to those without in either ORLS1 or ORLS2. There was also no significant association between having had TB (all sub-types, PTB or EPTB) and subsequent risk of DM (all sub-types, T1DM or T2DM) in either ORLS1 or ORLS2. In THIN dataset the risk of TB (all sub-types) was found to be increased amongst individuals with DM (all sub-types), T1DM and T2DM when compared to those without. Thus, the IRR of TB (all sub-types) were significantly increased amongst individuals with DM (all sub-types) (IRR 1.50 (95%CI 1.27-1.76) P-value <0.001), T1DM (IRR 1.46 (95%CI 1.10-1.92) P-value 0.008) and T2DM (IRR 1.54 (95%CI 1.30-1.82) P-value < 0.001) compared to those without DM. The rate of PTB amongst individuals with DM (all sub-types), T1DM, or T2DM compared to those without were not significantly raised within THIN. The rate of EPTB was raised amongst those with T1DM (IRR 2.09 (95%CI 1.19-3.66), P-value 0.010) but was not raised amongst those with DM (all sub-types) (IRR 1.43 (95% CI 0.99-2.07), P-value 0.055) or those with T2DM (IRR 1.39 (95%CI 0.93-2.06), P-value 0.11) when compared to those without DM. In THIN dataset the rates of DM (all sub-types), T1DM and T2DM were found to be raised amongst individuals who have had TB (all sub-types), PTB and EPTB when compared to those who have not. The rate of DM (all sub-types) was increased amongst those who have had TB (all sub-types) (IRR 5.65 (95% CI 5.19-6.16) P-value <0.001), PTB (IRR 5.74 (95% CI 5.08-6.50) P-value <0.001) and EPTB (IRR 4.66 (95% CI 3.94-5.51) P-value <0.001) when compared to those who have not had TB. The rate of T1DM was increased amongst those who have had TB (all sub-types) (IRR 5.49 (95% CI 5.02-6.02) P-value <0.001), EPTB (IRR 0.84 (95% CI 0.35-2.03) P-value <0.001) but not amongst those who have had PTB (IRR 1.09 (95% CI 0.62-1.93) P-value 0.77) when compared to those who have not had TB. The rate of T2DM was increased amongst those who have had TB (all sub-types) (IRR 2.21 (95% CI 1.68-2.91) P-value <0.001), PTB (IRR 5.38 (95% CI 4.73-6.12) P-value <0.001) and EPTB (IRR 4.36 (95% CI 3.65-5.22) P-value <0.001) when compared to those who have not had TB. However, within THIN dataset these estimates of association were being promoted by a significant age by TB interaction effect. Utilising systematic review techniques, twenty five studies were identified which reported upon TB outcomes amongst those with compared to without DM. Meta- analyses showed individuals with co-morbid TB and DM had no significant difference in bacterial clearance rate after 2-3 months of treatment (1,675 participants, 6 trials, Relative Risk (RR) 1.38 (95% CI 0.97-1.97)), no significant difference in risk of TB recurrence & relapse (1,225 participants, 4 trials RR 1.20 (95% CI 0.93-1.54)), but a statistically significant increased risk of all cause mortality (12,128 participants, 18 trials, RR 1.97 (95% CI 1.53-2.55)) comparative to those with TB in isolation. Discussion TB risk is increased in those with compared to those without DM within a UK setting. However, it remains unclear if risk of PTB and EPTB are raised amongst those with DM comparative to those without. It also remains unclear as to whether risk of DM is increased amongst those who have had TB comparative to those who have never had TB. Individuals with co-morbid disease are at a greater risk of mortality during active TB disease than those with TB alone, however risk of TB relapse and recurrence are the same. Consideration of the association between DM and TB may become more important for improving TB control and TB treatment as DM prevalence rises in the UK and globally. In areas where TB is endemic TB screening amongst those with DM and TB prophylaxis may be needed to reduce or stabilise numbers developing active disease. Also, the increasing numbers suffering from co-morbid TB and DM will need heightened clinical attention in order to improve TB mortality outcomes.

The integration between health and social care organisations is an acknowledged public policy problem. Issues such as delayed discharge, also known as ‘bed blocking’ have become matters of public debate. Despite government efforts at implementing ‘joined up thinking’, government have found this area frustratingly ‘policy resistant’. Government proposals on fining local authorities over ‘bed blocking’ had to be withdrawn when a systems thinking based report showed their deficiencies. The thesis considers the effectiveness of information systems and the utility of systems thinking approaches in assessing the likelihood of a successful information system led intervention in this area. It relates specifically to social care provision for the elderly, since elderly people are the most significant proportion of patients that experience delayed discharge. A case study approach has been applied looking at the health and social care team responsible only for elderly patients. This concerns the elderly care wards from hospitals at two NHS trusts and a social services department. System dynamics and the use of a rich picture from the Soft Systems Methodology are used. The rich picture has captured the problem situation and informed the building of the system dynamics model, such as of the delays and patient process that exists. The system dynamics model has allowed rigorous testing of the effects of information system policies. These methodologies are used to test the proposition that information systems can achieve a significant improvement in reducing delayed discharges. The focus of this thesis is to look at the effect that integrated health and social care information systems can have on delayed discharging of elderly patients in the UK National Health Service (NHS) and Social Services. This thesis asserts that there has been a need for better information systems not just within separate health and social care industries, but a need for better information systems that span both sectors. Social care and health sectors have both shown a poor history of implementing information systems. There has been a lack of understanding shown of the impacts of information systems using systems thinking approaches, largely by those who use the information systems. This thesis makes the assertion that although information systems will not eradicate all health and social care organisational problems, information systems will make a significant beneficial difference to the way in which organisations operate. Integrated information systems have shown to be useful in informing the capacity changes that are needed throughout the health and social care system. This informs management action to change capacities based on backlogs of patients waiting for services. This led to reduce delayed discharging, as numbers of patients experienced delayed discharges dramatically fall. System dynamics has shown to be useful in allowing the complex behaviour of delayed discharges over time to be understood, by making use of feedback loops and time delays. A rich picture has been useful in capturing stakeholders’ views and the problem situation of delayed discharging. The rich picture has been used as a communication tool to aid stakeholders understanding the problem situation. The conclusions are that these methodologies provide a sound test bed for the proposition that integrated information systems can be useful. However, this is only if they stimulate action across the health and social care sectors, when capacity is limited. Integrated information systems should be used by managers to inform them of the capacity changes that need to be made throughout the patient process, helping to ensure there is a greater response and action to reduce delayed discharges.

Introduction. This is an international study of the epidemiology of family medicine in three practice populations from the Netherlands, Malta and Serbia. Distributions of incidence and prevalence rates and diagnostic associations between common reasons for encounter and episodes titles are compared, and similarities and differences are described and analysed. The research project also involved two systematic reviews of the literature, and research into the duration of episodes of care. Methodology. Participating family doctors (FDs) recorded details of all their patient contacts in an episode of care(EoC) structure using the International Classification of Primary Care (ICPC). Reasons for encounter (RfEs) presented by the patient and episode titles (diagnostic labels of EoCs) were classified with ICPC. The relationships between RfEs and episode titles were studied with Bayesian methods. Results. Distributions of incidence and prevalence rates and of diagnostic odds ratios from the three population databases are presented and compared. Conclusions. ICPC, the RfE and the EoC data model are appropriate tools to study the process of diagnosis in Family Medicine (FM) . Distributions of diagnostic associations between RfEs and episode titles in the Transition Project international populations show remarkable similarities and congruencies in the process of diagnosis from both the RfE and episode title perspectives. The congruence of diagnostic associations between populations supports the use of such data from one population to inform diagnostic decisions in another. Differences in the magnitude of such diagnostic associations are significant, and population-specific data are therefore desirable. It is proposed that both an international (common) and a local (health care system specific) content of FM exist, and that the empirical distributions of diagnostic associations presented in this paper are a reflection of both these effects. A new indicator of the chronicity of a health problem is proposed.

'Well-being' is a word that has appeared in policy documentation and academic papers with increasing frequency during the last few decades. However, it is far from clear as to what the word means or to what it refers, and it is the existence of that ambiguity that constitutes the rationale for this study. My general strategy for dealing with the observed obscurity was to investigate the available academic and policy literatures, and explore how those involved in policy formation and development configured and deployed the word well-being in their written and spoken discourse. To that end, I collected multiple sources of qualitative and quantitative data. My primary data involved the collection of 17 semi-structured interviews with academics and policy makers engaged with the study of well-being. My secondary data were derived from a study of 591 randomly selected academic papers drawn from six separate fields of inquiry. I analysed my data using various quantitative and qualitative techniques, including modified forms of content analysis and thematic analysis. Three key discoveries emerged from the research. First, the word well-being, which appears with increasing frequency across academic and policy discourse, has become increasingly 'psychologicalised'. Contemporary explanations perceive well-being as an epiphenomenon, which arises from the dialectical relationship between the availability of resources and a person's ability to use these capitals for personal betterment over the life course. Second, the word appears to function as a useful political, boundary object. In this respect, it is able to conscript others - individuals, departments, agencies, and organisations - into taking responsibility for well -being. Third, multiple interpretations of well-being abound in academic and policy discourse, and while we have yet to reach consensus on a definition of well -being, there is agreement that it is a phenomenon, which is capable of measurement and quantification.

Introduction: In many low income countries the delivery of quality health services is hampered by health system-wide barriers which are often interlinked, however empirical evidence on how to assess the level and scope of these barriers is scarce. It has been recognised that taking a more comprehensive approach to assessing these barriers is more likely to provide lessons on what works and why. WHO has been advocating the use of systems wide approaches such as systems thinking to guide intervention design and evaluation. This thesis reports system-wide assessment of a complex health system intervention in Zambia known as Better Health Outcome through Mentorship and Assessment (BHOMA) that aimed to improve service quality at the health facility and influence service demand from the community. Methodology: This study is nested within a cluster randomised trial of the BHOMA intervention that aims to strengthen the health system in three rural districts covering 42 health facilities in Zambia. The main trial has a stepped wedge design where the intervention is being rolled-out to all the 42 health facilities over a period of 4 years. A baseline health facility survey was done in 2011. This was followed by a 12 months post-intervention evaluation survey. At the time of the follow up survey 24 health facilities had received the intervention while 18 had not. Data collection used both quantitative and qualitative methods. The study was guided by a systems thinking theoretical framework which was inspired by the WHO building blocks for health system strengthening. Results: The baseline survey validated tools and indicators for assessing health system building blocks. Research paper 2 applied an innovative measure of health worker motivation which was initially applied in Kenya. The results showed that this simple tool was reliable with cronbach’s alpha of 0.73 for the 21 item measures of health workers’ motivation. Baseline assessment of health worker motivation showed variation in motivation score based on gender and access to training. Research paper 3 tested and applied a new tool for measuring health systems governance at health facility level. The new tool for measuring governance was reliable with the 16 item one side cronbach’s alpha ranging between 0.69-0.74.The tool was simple to use and found to be applicable in the Zambian health care setting. A balanced scorecard approach was applied to measure the baseline health system characteristics for the target districts. Differences in performance were noted by district and residence in most domains with finance and service delivery domains performing poorly in all study districts. Regression modelling showed that children’s clinical observation scores were negatively correlated with drug availability (coeff 20.40, p = 0.02) while Adult clinical observation scores were positively association with adult service satisfaction score (coeff 0.82, p = 0.04) Baseline qualitative results are presented in paper 5. The results showed close linkages between health system building blocks. Challenges noted in service delivery were linked to human resources, medical supplies, information flow, governance and finance building blocks either directly or indirectly. The 12 months post intervention survey applied both quantitative and qualitative methods. Research paper 6 presents 12 months post intervention quantitative results applying the balanced scorecard approach as at baseline. Comparison was made between the control and intervention health facilities. The results showed significant mean differences between intervention (I) and control (C) sites in the following domains: Training domain (Mean I:C;87.5.vs 61.1, mean difference 23.3,p=0.031),adult clinical observation domain (mean I:C;73.3 vs.58.0, mean difference 10.9,p=0.02 ). The 12 months post intervention qualitative evaluation applied systems thinking approach and the conceptual framework developed before the intervention. The findings are presented in research paper 7. The overall results showed that the community had accepted the intervention with increasing demand for services reported in all sites where the BHOMA intervention was implemented. The indications were that in the short term there was increased demand for services but the health workers’ capacity was not severely affected. However, from a systems thinking perspective, it was clear that several unintended consequences also occurred during the implementation of the BHOMA. Conclusion: In evaluation of complex interventions such as the BHOMA attention should be paid to context. Using system wide approaches and triangulating data collection methods seems to be important to successful evaluation of such complex intervention.

In this thesis I take up the topic of our understanding of questions in a detailed case study of non-utility measures of health-related quality of life. I argue that efforts to standardize these measures lead to limitations in our ability to understand and measure quality of life. In the first half of this thesis I describe two types of bias that affect quality of life measures despite efforts to validate them. On the one hand, quality of life measures can perpetuate ethnocentric understandings of quality of life. On the other hand, respondents often understand the questions in these measures very differently than researchers imagined. I argue that the residual bias found in quality of life measures is the result of two assumptions built into the use of construct validity: 1) when a measure's outcomes confirm our hypotheses, we are warranted in having greater confidence in the accuracy of our theory 2) respondents understand the questions and answers in our measures in the same way as researchers imagined they would. In the second half of this thesis I argue that the limitations of construct validity stem from the logic of asking questions, a logic which precludes standardization. I propose that quality of life measures ought to be understood differently-they are not independent instruments capable of unambiguous claims, but rather one element in a dialogic framework whose questions and outcomes serve as the starting point for further inquiry. Finally, I examine what might have motivated the misguided use of construct validity. I suggest that the motivation lies in an erroneous picture of the human subject. I argue for an alternative picture that allows me to introduce an ethical dimension to our questions about quality of life.

This thesis examines the mechanisms of policy change in the hospital sector in three countries (England, Sweden and Japan), and argues that pressure on central government to respond to public concerns can significantly alter conventional institutional arrangements. By analysing four types of pressure (two mainly political, i.e. local campaigns against hospital closure and corporatisation of public hospitals; two mainly technical, i.e. quality assurance system-building, and malpractice incidents), the thesis sheds light on the fact that, when institutional vulnerabilities are exposed to public criticism, central governments exhibit their capacity to reform the hospital sector irrespective of institutional constraints. Under these circumstances, the varieties of the institutions in the three countries do not matter, as the observed responses were similar. In order to compare and contrast the 'responsiveness' of central government within the different 'logics' of the respective health care systems, this thesis investigates selected parliamentary and unitary states with universal health coverage, each however with different degrees of state involvement in the hospital sector: England (nationally-run) as part of the United Kingdom, Sweden (locally-run) and Japan (predominantly privately-run). By differentiating the types of pressure and examining the saliency of each issue in the printed media, it is demonstrated that the responsiveness of government to pressure is largely affected by the institutional arrangements in which they operate. However, when the saliency of non-redistributive technical issues is high, institutional constraints are overcome and institutional choices by government are reversed under heightened pressure. The analysis of dynamic policy change questions the constraining nature of political institutions on health reforms, and explains how policy convergence comes about to an extent that goes beyond path dependency in this predominantly profession-driven policy sector.

This thesis examines technology diffusion within the UK NHS. Motivated by increasing health expenditure over the last years, it is important to understand the diffusion process of medical technology in order to determine the factors that enhance or delay the incorporation of technologies into common practice. Given the uncertainty inherent in new technology and its presupposed competitive advantage, the diffusion process is approached through the informational sources available to agents as a mechanism to overcome uncertainty. Information increases physicians' knowledge on product quality and consequently influences technology choice. The set of regulatory and financial incentives provided by the health care system are also considered. Throughout the thesis dynamic panel data methods are used to estimate technology demand equations. The first case study looks at diffusion within the primary care sector of three drug groups at the therapeutical class level using prescription data from IMS Health. The second empirical case explores within-group therapeutical diffusion with emphasis on competition amongst branded products. The question addressed relates to the informational and product characteristics that consolidate different prescription trends and product uptake. Results suggest that prescription experience is the most important source of information; however, physicians access additional informative channels when the technology is a breakthrough innovation. Additionally, drug diffusion is unaffected by the health system organisation. The final empirical work addresses diffusion of two surgical procedures in the secondary care sector using HES data. Specifically, it considers the impact of competition introduced by the NHS reforms initiated in the 90s. Patient follow-up also allows exploration of the impact that surgical innovation has on patients' health outcomes using a competing risk model. Findings suggest higher diffusion in less concentrated markets, with specialised and university providers having faster uptake. Moreover, diffusion presents long-term effects on improved quality of care.

Longitudinal electronic health records are a valuable resource for research because they contain information on many patients over long follow-up periods. Missing data commonly occur in these data because it was collected for clinical and not research purposes. Analysing data with missing values can potentially bias estimates and standard errors resulting in invalid inferences. Multiple imputation, commonly used in research to impute missing values, is increasingly regarded as the standard method for handling missing data in medical research because of its practicality and flexibility under the assumption the data is missing at random (MAR). Until now, few imputation approaches are sufficiently flexible to account for the longitudinal and dynamic structure of electronic health records. However, the two-fold fully conditional specification (FCS) algorithm was proposed to impute missing values in longitudinal data, but this methods was not currently validated in the complex setting of longitudinal electronic health records. I propose to adapt, evaluate and implement the two-fold FCS algorithm to impute missing data from large primary care database. To achieve this, first I investigate the extent and patterns of missing data in a longitudinal clinical database for health indicators associated with cardiovascular disease risk to determine if the MAR assumption is plausible. Additionally, I develop methods to identify and remove outliers, which can potentially bias imputations, from data with repeated measurements before imputation. Next, I adapt and develop the two-fold FCS multiple imputation algorithm to impute missing values in longitudinal clinical data for health indicators associated with cardiovascular disease risk and I validate the two-fold FCS algorithm to assess bias and precision through challenging simulation studies. I develop a new software programme which implements this adapted version of the two-fold FCS algorithm to impute missing values in longitudinal data. Finally, I apply the two-fold FCS algorithm in THIN to (i) model cardiovascular disease risk and (ii) understand factors associated with greater total cholesterol reduction in patients with type II diabetes.