The unique decision making task entrusted to lay juries in adversarial legal systems has attracted the attention of legal psychologists for decades, but more recently technological advances in forensic science have highlighted the importance of understanding how jurors perceive this often ambiguous and complicated type of evidence. This thesis begins by investigating the forensic awareness of lay participants, and the ability of mock jurors to discriminate between varying probative values of forensic evidence. The findings suggest that the perception of weak forensic evidence is affected by contextual information, and there was wide disagreement among participants about the probative value of weak evidence. In an effort to explain the variance in perceived evidence strength, a measure of pre-trial attitudes about forensic science was developed (the Forensic Evidence Evaluation Bias Scale – FEEBS) and administered to 446 participants ranging from students, to jury eligible members of the public, to actual jury venire persons. The results of exploratory and confirmatory factor analyses identified two distinct clusters of attitudes measured by the FEEBS, which correspond conceptually to the hypothesised juror beliefs described in the CSI Effect literature. These attitudes were found to have a significant indirect effect on verdict preference, for trial vignettes describing murder, robbery, and sexual assault scenarios containing weak (or absent) forensic DNA evidence. The implications of these findings for voir dire hearings are discussed, with reference to the cognitive models of juror decision making and the CSI Effect literature.

Background: Smoking cessation interventions are underprovided in primary care. This thesis examines the impact of financial incentives on the provision of smoking cessation interventions, and inequalities in provision, in primary care. Methods: • Systematic review of financial incentives for smoking cessation in healthcare. • Cross sectional study using general practice data from Wandsworth, London, using logistic regression to examine associations between ethnicity and disease group with ascertainment of smoking status and provision of cessation advice following the introduction of the UK’s Quality and Outcomes Framework (QOF). • Before-and-after studies using general practice data from Hammersmith & Fulham, London, looking at the impact of a local financial incentive scheme (QOF+) on smoking outcomes for patients without smoking-related diseases (primary prevention), and antenatal patients, using logistic regression to examine inequalities. Results: Introduction of financial incentives was associated with increased recording of smoking status and advice to smokers, most evident for patients with smoking-related diseases compared with patients without smoking-related diseases, for whom there were much smaller incentives for recording smoking status and none for offering stop smoking advice. However, when specific incentives were provided for primary prevention large improvements in smoking outcomes were seen. The youngest and oldest groups of patients were less likely to be asked about smoking. White British patients were more likely to smoke than other ethnic groups, except Black Caribbean men with depression, 62% of whom smoked. Smoking advice was provided relatively equitably, but Black Caribbean men with depression were less likely to receive advice than White British men with depression (59% vs 81%). Disparities in smoking outcomes with respect to age and ethnicity persisted after the financial incentives were introduced. Conclusions: Introduction of financial incentives was associated with increases in recording smoking status and largely equitable provision of cessation advice, but variations in smoking outcomes between groups persisted. Extending financial incentives to include recording of ethnicity and rewarding quit rates may further improve smoking outcomes in primary care.

Background: Hypertension is the main cardiovascular risk factor. Worldwide, 40% of adults aged 25 years and over have hypertension. Only around 25% of hypertensive patients had controlled hypertension (CH). I assessed the association between individual factors, physician factors and health system factors with having CH. Methods: Population included 17252 hypertensive adults (HA) registered with 28 general practices in Wandsworth, London between 1998 and 2007. Comparison of two blood pressure targets was performed by cross-sectional analysis. Changes in antihypertensive prescribing were evaluated using logistic models. A Bayesian frailty survival model was developed to assess the relationship between potential risk factors and CH with blood pressure <=140/90 mm Hg. Results: In 2007, 26 (93%) practices had more than 70% HA with CH using the QOF target. Using the NICE target of 140/90 mm Hg 4 (14.2%) of practices had more than 70% HAs with CH. In 1998 49.0% HA were not prescribed antihypertensive medication but only 14.8% in 2007. The introduction of the 2006 NICE guidelines was associated with an increase in recommended monotherapy prescribing. 37% of 11373 HA used for survival analysis had CH in 1998 and 62.4% had CH in 2007. Being older, increases in body mass index and higher levels of cholesterol were negatively associated with having CH, hazard ratio (HR) 0.99 (95% credible interval 0.99 – 0.99), HR 0.93 (95% credible interval 0.88 – 099) and HR 0.89 (95% credible interval 0.79 – 0.99) respectively. HA registered after the introduction of Quality and Outcomes Framework (QOF) were more likely to have CH, HR 2.05 (95% credible interval 1.94 – 2.16). Conclusion: In this cohort of HA being treated in primary care, individual factors were mainly associated with not having CH. Blood pressure control rates increased over the period. The use of national hypertension guidelines and the QOF could contribute to this improvement.

Recent advances in ’omics technologies and the development of new computational techniques have greatly contributed to the identification of factors influencing the onset and progression of many common diseases. Yet, despite this great success, it is unlikely that the independent analysis of these data will elucidate the complex web of mechanisms involved in disease development. To enhance our knowledge of disease aetiology, new approaches for linking the large amount of available data need to be developed. As a step towards this goal, the aim of this thesis is to investigate and develop novel statistical methods for the integrative analysis of ’omic data. In particular, in this project we analyse genomic and metabolomic data in relation to the health outcomes in three study populations. To investigate how genetic and metabolic variables act as risk factors in the development of complex disorders, we have developed three novel analytical methodologies, namely ’Differential Network’, ’GEMINi: GEnome Metabolome Integrated Network analysis’ and ’Variance and Covariance regression’ and illustrate their use on real data sets. The results demonstrate the applicability of the new methodologies to identify key molecular changes undetectable with standard approaches. The approaches introduce here have the potential of providing insight into the biological basis of phenotypic variation and aid the generation of new hypotheses about molecular control and regulation in the context of systems biology.

Hypertensive disorders in pregnancy are among the leading causes of death and disabilities in women of child-bearing age in low-income countries. Despite the high burden, the pathophysiology and aetiology of these disorders are poorly understood, making effective prevention difficult. Excessive dietary salt intake is an established risk factor for hypertension, but much less is known about the health effects of water salinity. This epidemiological study investigated the relationship between high salinity in drinking water and the risk of (pre)eclampsia and gestational hypertension among pregnant women in a rural coastal community in Dacope, Bangladesh, where both surface- and groundwater have been affected by severe salinity, exacerbated by the effects of sea-level rise and climate change. The population-based case-control study accrued 202 cases of (pre)eclampsia and gestational hypertension and 1,006 controls. The study showed that women in Dacope were exposed to exceptionally high levels of sodium in their drinking water, and found a significant association with the risk of (pre)eclampsia and gestational hypertension. The risks increased in a dose-response manner for increasing quartiles of sodium concentrations in drinking water (ORs 3.36 [95% CI 2.07 - 5.60], 4.35 [2.61 - 6.94] and 5.40 [3.28 - 8.92]), (p-trend<0.001). The associations persisted after adjusting for multiple potential confounders and in several sensitivity analyses. These findings shed light on a previous observation of high rates of (pre)eclampsia and gestational hypertension in the dry season, when salinity levels in surface and groundwater were higher than in the monsoon season. The study provides an evidence base for creating guidelines on sodium consumption for the prevention of hypertensive disorders in pregnant women and the general population living in coastal areas. Furthermore, the study findings can inform policy makers about the urgent need to find sustainable solutions, for example, by finding alternative sources of drinking water. It is likely that climate change will considerably worsen the current situation, and this adds further impetus to the need for intervention.

Despite extensive epidemiological research and plausible biological mechanisms being elucidated, it is unclear whether vitamin D reduces risks of cancer incidence and mortality. Only for colorectal cancer does the observational evidence seem persuasive, whereas for other cancer types an anti-carcinogenic role has not been established convincingly, with rarer cancers seldom investigated. Similarly, whether vitamin D has a beneficial role on other chronic disease end-points and all-cause mortality remains uncertain, despite extensive research. Prospective studies which directly measure actual circulating 25-hydroxyvitamin D (25(OH)D) are viewed as the 'gold standard' approach to assess vitamin D-disease associations. However, these studies are expensive to carry out (as circulating 25(OH)D usually has to be measured in all participants) and a single measurement of circulating 25(OH)D may not reflect long-term exposures (due to within-person variability). An alternative approach, not yet used in European populations, is to create predictor scores of circulating 25(OH)D levels. This cost effective approach provides the opportunity to examine associations between predicted 25(OH)D and multiple outcomes (including less common diseases). Sex-specific predictor scores were derived in 4,089 participants from the European Investigation into Cancer and Nutrition (EPIC) study by quantifying the relationships between correlates/determinants of circulating 25(OH)D levels (using multivariable linear regression models). The predictor scores were validated in 2,029 participants with measured circulating 25(OH)D levels. In summary, the predictor scores provided poor estimates of absolute circulating 25(OH)D levels but were more successful at ranking individuals similarly by their actual and predicted levels. The predictor scores were also able to replicate results from previous EPIC colorectal cancer incidence and prostate cancer incidence nested case-control studies which used actual circulating 25(OH)D measurements. Overall, this evidence suggests that the predictor scores may have utility for epidemiological research but not in a clinical setting. The predictor scores were then applied to the full EPIC cohort to assess the associations between predicted 25(OH)D levels with risks of cancer incidence and mortality, all-cause mortality, and cause-specific mortality. In summary, significant inverse predicted 25(OH)D score associations were observed for: overall cancer incidence and mortality; colorectal cancer incidence; lung cancer incidence and mortality; kidney cancer incidence; stomach and oesophageal cancer incidence; pancreatic cancer incidence and mortality; thyroid cancer incidence; prostate cancer mortality; all-cause mortality; circulatory disease mortality; respiratory disease mortality; and digestive disease mortality. However, due to the methodological limitations specific to 25(OH)D predictor scores - such as providing poor estimates of absolute levels - and observational epidemiology in general, it is important to acknowledge that alternative explanations may explain some or all of these observed relationships.

Chronic kidney disease has become a worldwide public health problem as it is linked with an increase in the incidence and prevalence rate of patients with renal failure who require renal replacement therapy with high healthcare costs and poor outcomes. Moreover, chronic kidney disease (CKD) in its earlier stages is associated with a higher prevalence of many adverse outcomes such renal failure, cardiovascular disease, and premature death. In Saudi Arabia, the total number of patients who are receiving renal replacement therapy has been increasing rapidly in recent decades, as the Saudi Centre for Organ Transplantation data has demonstrated, while the estimated annual cost for treating end-stage renal disease uses around 3.8% of the total Ministry of Health's budget in Saudi Arabia. This means that more attention is required for the prevention and management of CKD and its risk factors in Saudi Arabia. Therefore, this thesis was formulated to review the quality of healthcare management for patients with chronic kidney disease in Saudi Arabia. This review was achieved based on: (I) assessing and describing the epidemiology of end-stage renal disease, (II) assessing pre end-stage renal disease care practice, (III) examining the indications for hospital admission among haemodialysis patients. The findings have revealed that the burden of end-stage renal disease has increased substantially in Saudi Arabia particularly among middle-aged diabetic and hypertensive patients, while the most prevalent co-morbid conditions were hypertension and Hepatitis C virus infection. Detecting patients with chronic kidney disease in the earlier stages, and pre end-stage renal disease care practices and its outcomes should be improved in Saudi Arabia. Furthermore, cardiovascular diseases play a major role in increasing the rate of hospitalization and mortality among patients with end-stage renal disease. In conclusion, this review stressed the needs and the importance of multidisciplinary preventive strategies for leading causes of end stage renal disease (ESRD) and improving pre ESRD nephrology care practices and its outcomes in Saudi Arabia. This will not be possible to achieve without collaboration between healthcare providers and a comprehensive national health information system to allow consistent assessment of the quality of healthcare management and provide an updated evidence-based health policy that will help healthcare providers to choose the best intervention that will help to improve healthcare practice outcomes and allocate healthcare resources.

Introduction Patients are describing their healthcare experiences online using rating websites. There has been substantial professional opposition to this, but the government in England has promoted the idea as a mechanism to improve healthcare quality. Little is known about the content and effect of healthcare rating and review sites. This thesis aims to look at comments left online and assess whether they might be a useful measure of healthcare quality. Method I used a variety of different approaches to examine patients' comments and ratings about care online. I performed an examination of the comments left on the NHS Choices website, and analysed whether there was a relationship between the comments and traditional patient surveys or other measures of clinical quality. I used discrete choice experiments to look at the value patients place on online care reviews when making decisions about which hospital to go to. I used natural language processing techniques to explore the comments left in free text reviews. I analysed the tweets sent to NHS hospitals in England over a year to see if they contained useful information for understanding care quality. Results The analysis of ratings on NHS Choices demonstrates that reviews left online are largely positive. There are associations between online ratings and both traditional survey methods of patient experience and outcome measures. There is evidence of a selection bias in those who both read and contribute ratings online - with younger age groups and those with higher educational attainment more likely to use them. Discrete choice experiments suggest that people will use online ratings in their decisions about where to seek care, and the effect is similar to that of a recommendation by friends and family. I found that sentiment analysis techniques can be used classify free text comments left online into meaningful information that relates to data in the national patient surveys. However, the analysis of comments on Twitter found that only 11% of tweets were related to care quality. Conclusions Patients rating their care online may have a useful role as a measure of care quality. It has some drawbacks, not least the non-random group of people who leave their comments. However, it provides information that is complementary to current approaches to measuring quality and patient experiences, may be used by patients in their decision-making, and provides timely information for quality improvement. I hypothesise that it is possible to measure a 'cloud of patient experience' from all of the sources where patients describe their care online, including social media, and use this to make inferences about care quality. I find this idea has potential, but there are many technical and practical limitations that need to be overcome before it is useful.

River Blindness (onchocerciasis) has been identified by the World Health Organization as potentially eliminable. Until recently, the African Programme for Onchocerciasis Control focused on annual distribution of ivermectin to reduce morbidity, but encouraged by success in some foci, it has embarked on eliminating the infection from the continent. To this end, increasing the treatment frequency to twice yearly (biannual) has been suggested. However, this may not be cost-effective everywhere in Africa, so it is necessary to assess under which epidemiological scenarios it would be advisable. The central aim of this thesis is to develop further an onchocerciasis transmission model (EpiOncho) to evaluate the impact and cost of biannual vs. annual ivermectin treatment in a range of scenarios typical of savannah onchocerciasis foci in Africa. The analyses and methods are divided into three main components. First, a mathematical model of the dynamics of onchocercal disease was developed and linked to infection output from EpiOncho. Results indicate that although long-term annual ivermectin treatment reduces dramatically onchocerciasis related disease burden, its overall impact on infection depends strongly on baseline levels of endemicity. Second, a study was conducted in Ghana to assess the economic cost of biannual relative to annual ivermectin distribution. Results indicate that the (per year) cost of biannual ivermectin treatment is approximately 60% higher than that of annual treatment (and not simply double, as assumed by others). Third, the health impact, programmatic cost, and projected duration of biannual vs. annual ivermectin treatment were evaluated. Findings indicate that although biannual treatment yields only small additional health benefits over those of annual treatment, its benefit is pronounced in the context of elimination goals, shortening timeframes to reach proposed operational thresholds for stopping treatment and potentially generating programmatic cost savings. Notwithstanding these conclusions, the feasibility of increasing from one to two treatments yearly will vary with the specific programmatic circumstances.

Currently, regulators make marketing authorisation decisions based on the benefit-risk balance of a treatment with little input from patients and the public. The aims of this thesis are to apply systematic methods of benefit-risk assessment proposed for use in regulatory decision-making, and examine how they can incorporate patient and public involvement. These aims were met through the investigation of three case studies: efalizumab, rimonabant, and natalizumab. Efalizumab (indication: plaque psoriasis) was withdrawn from the market due to the occurrence of progressive multifocal leukoencephalopathy. I examined the feasibility of applying descriptive and quantitative frameworks to determine the benefit-risk balance of efalizumab, and evaluated where patient and public involvement can be applied in the decision-making process. Rimonabant (indication: overweight and obesity) was withdrawn from the market due to the occurrence of psychiatric disorders. I conducted a discrete choice experiment to elicit patient and public preferences regarding the benefits and risks of treatment, and developed a probabilistic analysis method to analyse the preference data and determine the benefit-risk balance of rimonabant. Natalizumab (indication: relapsing remitting multiple sclerosis) received authorisation, but the benefit-risk balance was reassessed due to the occurrence of progressive multifocal leukoencephalopathy. Authorisation was maintained with risk minimisation measures. I developed a framework to guide the application, reporting, and evaluation of patient and public involvement in benefit-risk assessment. I then tested the feasibility of the framework with protocols designed to elicit patient and public preferences on the benefits and risks of natalizumab. Based on the findings of these case studies, preferences can be elicited from patients and the public and used to determine the benefit-risk balance of a medicine. This thesis provides a methodological foundation to guide the application, reporting and evaluation of PPI in the benefit-risk assessment of medicines and improve the legitimacy, transparency, and quality of regulatory decision-making.