Since 2002, the minimum recommended age worldwide for the introduction of solid foods has been six months, an age when most infants are able to bring food to their mouth and chew it, without assistance. Despite this, the practice of spoon feeding with purées remains prevalent and most research to date has examined the introduction of solid foods from the adult’s perspective rather than the infant’s. As a result, factors that may impact on the food preferences of infants, such as the appearance, smell and haptic qualities of food, have not been investigated, nor has the routine use of puréed foods been challenged. Similarly, while differences have been well documented between the processes of breastfeeding and bottle feeding, the possibility that there may be pertinent differences between spoon feeding and self-feeding has not been explored. Overall, the introduction of solid foods has been researched in nutritional terms, rather than in relation to the infant’s experience and his wider learning and development. This study appears to be the first to explore the introduction of solid food from the infant’s perspective. Ten infants were offered a single food, both as a graspable piece and as a spoon-fed purée. The experience was audio/video-recorded and analysed in depth using a combination of quantitative and qualitative methods. Two interviews were conducted with the mother of each infant, during which they were asked to eat the same food, in the same formats, as their infant, and to comment on the audio/video-recording. The findings indicate that spoon feeding and self-feeding are two contrasting experiences. Self-feeding was seen to be characterised by exploratory behaviour, while spoon feeding showed more evidence of avoidant behaviour by the infant and controlling behaviour by the mother. Possible implications for parental and professional guidance and for future research are discussed.

Particulate air pollution (PM) has been shown by many studies to cause adverse health effects. Traditionally PM exposure was estimated using ambient concentrations. Lately, studies have revealed that this approach poorly reflects differences between individual's exposures and as such results in exposure misclassification. This thesis aims to improve personal exposure predictions by building a model (MEPEX model), which takes into account the temporal and spatial variability of ambient PM, as well as visited microenvironments. For the composition of this model, existing approaches for model components were evaluated, compared and developed. A temporally adjusted land-use regression (LUR-adj) model for predictions of ambient PM2.5 and PM10 was built, validated, and compared to estimates from a dispersion model. Ratios were developed to adjust ambient concentrations for cycling and in-bus transport microenvironments. Additionally, modelling approaches for the home indoor microenvironment were compared, using monitoring data. A secondary aim was to evaluate the performance of different approaches for personal exposure assessment by comparing varying levels of model sophistication. Validation of the LUR-adj model showed good model fit (IA > 0.5) and low error (NMSE < 1) for short-term predictions of PM2.5 and PM10 at locations in London. In comparison to predictions of a dispersion model (ADMS-urban), LUR-adj estimates of PM10 produced better results for model performance parameters at the majority of 26 predicted locations. MEPEX model predictions of monitored daily personal exposure for an individual in London resulted in an R2 of 0.439 for PM2.5 and 0.403 for PM10. Predictions using modelled home outdoor concentrations in comparison were lower with R2 of 0.173 for PM2.5 and 0.086 for PM10. These results provide the first quantifiable evidence that personal exposure models of PM2.5 and PM10 can reduce exposure misclassification compared to estimates based only on ambient PM.

Introduction Breast cancer is the most common cancer in the world and environmental factors such as endocrine disrupting chemicals, as well as reproductive and hormone-related factors play a crucial role in the development of this disease. In order to assess causal pathways between these exposures and disease initiation, biomarkers based on DNA methylation measurements can be used. Methods The potential association between global and locus-specific DNA methylation and breast cancer risk was investigated in two prospective European nested case-control studies. The HM450 array was used to generate epigenomic profiles of archived blood samples, collected from study participants before the onset of disease in 324 matched case-control pairs. The association between endocrine disrupting chemicals measured in blood samples (n=368), reproductive and hormone-related variables assessed by questionnaire (n=324), and hormone levels measured in blood(n=36), and DNA methylation was studied. The meet-in-the-middle approach was applied to identify DNA methylation markers related to both exposures and disease endpoint. Results Global hypomethylation was observed among breast cancer cases compared with controls and locus-specific analyses identified 26 CpGs whose DNA methylation was associated with breast cancer. Cadmium exposure was associated with DNA methylation at 62 CpGs but most associations did not survive adjustment for smoking status. In addition, numerous reproductive and hormone-related variables, as well as the hormones D4 and testosterone were associated with DNA methylation, and three potential meet-in-the-middle candidates were observed. Discussion Despite the relatively low power, results indicated that genome-wide hypomethylation among breast cancer cases may serve as a biomarker for disease risk. More research with bigger sample sizes is needed to disentangle the potential effect of cadmium and smoking on DNA methylation and to further explore possible effects of reproductive and hormone-related factors, as well as hormone levels, on DNA methylation. It is of interest to investigate what the biological consequences of these changes in methylation are.

Growing evidence suggests an association between attention-deficit/hyperactivity disorder (ADHD) and obesity, although very little is understood about the nature of this link. The aims of this thesis were to examine the following aspects of the ADHD-obesity association: (1) the directionality of the link from childhood to adolescence, (2) behavioural mediators during childhood and adolescence, and (3) prenatal risk factors common for both disorders. Participants were from the Northern Finland Birth Cohort (NFBC) 1986 (N=9479). Data were obtained on pregnancy and birth factors, and child/adolescent mental health, obesity, and lifestyle factors. Regression analyses showed that ADHD symptoms significantly predicted obesity, rather than in the opposite direction, from childhood to adolescence. Mediation analyses examined potential underlying behavioural factors - physical activity and binge-eating, and showed that physical inactivity mediated the longitudinal ADHD symptom-obesity association. Further, there was a bidirectional, longitudinal association between physical inactivity and ADHD symptoms. ADHD and obesity may share common prenatal risk factors, including prenatal exposure to cortisol. This was studied using a quasi-experimental approach by examining the impact of prenatal exposure to synthetic glucocorticoids (sGC). Results from propensity-score and mixed-effects methods showed that prenatal sGC increased the risk for general psychiatric disturbance and inattention symptoms, but not obesity, in childhood. Placental size may represent another common prenatal contributing factor; placental size was positively associated with behaviour problems, including ADHD symptoms, in child and adolescent boys, but was not associated with obesity. This thesis addresses important unexplored aspects of the association between ADHD and obesity, and provides insight into risk factors for both disorders. The direction of the association was driven from ADHD symptoms to obesity, and physical inactivity was a behavioural mediator underlying the link. Although there was no evidence that both disorders share common prenatal risk, prenatal sGC and placental size were positively associated with ADHD symptoms.

Logistic regression models are often proposed to describe dose-response relationships in dose-escalation clinical trials to determine the maximum tolerated dose. In a Bayesian setting, the 1-parameter continual reassessment method and the 2-parameter escalation with overdose control designs have been implemented assuming acceptable tolerability thresholds of between 20% and 35%. The literature is sparse on the operating characteristics of 2-parameter Bayesian logistic regression models (BLRM) when sample sizes are small (i.e. <50); response rates <20% or >35% are of interest; and expert beliefs are available for incorporating into prior distributions for model parameters. Motivated by a case study of a new infertility treatment, this thesis describes the operating characteristics of the 2-parameter BLRM in a dose-escalation setting, with small sample sizes, and applied to response rates consistent with both safety and efficacy endpoints i.e. 10% to 90%. When information external to the trial is available from expert beliefs, ways in which those beliefs may be elicited in a structured manner are evaluated. Simulation is used to assess the impact of these prior distributions on trial conclusions. I have demonstrated that elicitation can be performed in a structured manner in both academic and industry settings and I provide specific recommendations for the structured planning and execution of elicitation sessions. Simulations show that there is no single set of priors that always produce unbiased estimates with minimum variance across a range of target response rates, so simulations specific to the planned trial must be conducted. Furthermore, when only discrete doses are available, simulations show that choosing the available dose closest to that recommended by the model is more likely to lead to an unbiased estimate of the dose that attains a pre-specified response rate. Recommendations are provided for how to improve the study design and analysis for the motivating case study.

Background Sickle cell disorders (SCD) are the most common inherited blood disorders in England. Without prompt diagnosis and proper treatment, they can be a serious source of morbidity and mortality. Sickle cell diseases affect mainly black minority and ethnic populations, and have so far received relatively low priority from a health policy perspective. Antenatal and newborn screening, the development of minimum standards, antibiotic prophylaxis, comprehensive immunisations, and preventive diagnostic tests have positively influenced SCD management. There remains an unclear picture as to the trends and health care utilization of patients with SCD in England. Aims This study looks at the burden of disease in England by assessing hospital admissions, readmissions and related costs. It also aims to identify gaps in care and prevention which may identify possible contributors to avoidable admissions. Findings Using Hospital Episode Statistics (HES) data, trends for SCD hospital admissions in England showed a rise in 50% of hospital admissions over a 10 year period. The most deprived areas had a higher rate of readmission and in-patient mortality among those with SCD. Adolescents had a higher rate of readmission possibly identifying a gap in health care access. Local findings in a high prevalence area showed that the majority of admissions were for a short length of stay and 74% of patients accounted for multiple admissions. A patient focus group and questionnaire both identified potential gaps in care and prevention. Conclusion Through the use of 6 studies which showed the SCD admission rates in England, the readmission rates, local admissions, costs associated with admission and patient perspectives in both care and prevention, there is a clearer picture as to the trends and health care utilization of patients with SCD in England. The studies suggest that ascertaining the prevalence of at-risk groups in England as well as addressing inequalities in health care access among minority groups and areas of high disease prevalence can further aid in disease management. Shifting diagnostic and follow-up care from acute care facilities to primary care facilities and promoting preventive care measures and adherence to standards and guidelines may possibly decrease the cost burden, reduce avoidable hospital admissions and increase the timeliness and effectiveness of disease management. Investing in training and education of primary care physicians for sickle cell diseases may also improve quality of care.

Background. Health information systems are inadequate in many countries. For childhood diarrhoea and pneumonia specifically, the leading infectious causes of child mortality worldwide, current data collection methods are not providing sufficient information for surveillance. The collection of health data could be greatly assisted with the use of mobile devices (mHealth). Mobile phone text messaging is widely used, but its potential for health data collection has not yet been realised. Aim. To explore the application of mHealth-based collection of information relevant to childhood diarrhoea and pneumonia in rural China. Methods. A mixed methods approach was used: (i) a survey and semi-structured interviews to assess the usage of mobile phones by caregivers of young children; (ii) cognitive interviews, usability testing and a cluster randomised cross-over study to determine the validity of a text messaging survey on care-seeking for childhood diarrhoea and pneumonia; and (iii) researchers' observations and structured interviews with participants of the cross-over study to evaluate factors influencing participation in mHealth-based studies. Results. Many of the 1854 survey participants (1620; 87.4%) used mobile phones. Of 1014 participants in the cross-over study, 662 (65.3%) responded to the first text message. Of 651 participants willing to participate, 356 (54.7%) completed the text messaging survey. Overall, text message data were moderately to substantially equivalent to face-to-face data. The text messaging survey was acceptable to parents, but grandparents were often unable to use text messages. Among many factors influencing participation were trust, perceived usefulness and ease of use. Conclusions. Text messaging can be applied to collect data on care-seeking for childhood diarrhoea and pneumonia in rural China, but several questions remain, including how to improve accuracy and response rates. Further work needs to advance innovative mHealth-based data collection methods that can improve health surveillance, enhance implementation of appropriate interventions and ultimately save children's lives.

Latin America is a region of diversity, inequality, poverty and an outstanding capacity to remain stable despite these challenges. The HIV epidemic in the region resembles these same characteristics, with a wide range of risk behaviours, a disproportionate burden in vulnerable groups and yet perhaps a highly effective response. Brazil and Colombia have extensively deployed prevention strategies and delivered antiretroviral treatment and there are still further expansions in sight. However, the likely impact of these HIV programmes on the epidemic has never been evaluated. This thesis addresses these gaps by retrospectively evaluating the impact of antiretroviral treatment and prevention campaigns on new HIV infections. This is done by means of mathematical models that represent HIV transmission in these settings and which creates a counterfactual projection for the trajectory the epidemic might otherwise have taken. These estimations are interpreted in the context of ambitious plans to scale treatment further, along with a growing realisation of the long-term costs that these programmes imply. Tracking the epidemic is essential for the evaluation of programmes in the next phase of the response. To support this, a new method for incidence estimation is proposed. This method relies exclusively on case-report data, which is robust in these settings, and a flexible model specification that should be suitable for a wide range of epidemic scenarios. The parameters for the model are estimated in Bayesian framework and applied to the case study of Colombia. These resulting estimates of the historic course of the epidemic in Colombia are strikingly different to that which has previously been estimated and casts new light on the nature of epidemics in this region and the response to it that is now required. Overall, these results stand as the first analysis of this kind in the region and present useful results and methods that should support a continued effective response to HIV epidemic in this region.

Using administrative data to measure the quality and safety of hospital care offers many opportunities. However, progress has been limited to few countries and predominantly to a small subset of broad measures, such as Hospital Standardised Mortality Rates. In this thesis, I investigate the potential advantages and feasibility - in terms of validity and applicability - of specialty-specific indicators. In the first part of my PhD work, I examine the case for specialty-specific indicators. I also present potential applications which overcome some of the existing shortcomings of previous uses of indicators based on administrative data. In the next stage of the project I focus on assessing feasibility by focusing on two specialties - stroke and obstetric care - conducting systematic reviews and consulting with experts to develop two indicator sets. As part of this, I identified the shortcomings in current use of indicators in these specialties. To investigate the limitations of these indicators, I applied the indicator definitions to English hospital administrative data (Hospital Episode Statistics, HES) and evaluated whether they can be used to discriminate between hospitals based on their performance and, importantly, to understand the effect of differences in coding practice. The final aspect of the research was to investigate alternative applications for the indicators which can overcome some of the shortcomings highlighted in both the prior analyses and existing literature. In doing so, I raise serious, robust shortcomings on the quality and safety of weekend care.

Since 2000, we have been trying to characterize and classify HIV epidemics to guide the strategic design of HIV prevention policies and focus HIV programmes and resource allocation by a regions' epidemic type. We have used arbitrary thresholds of HIV prevalence across different risk-groups in a given population, 'static' mathematical models and classical epidemiological measures of the population attributable fraction that do not account for chains of transmission. As a result, these traditional approaches could be missing the underlying transmission dynamics and the role of key populations - such as female sex workers and their clients - on HIV spread. In this thesis, I build on a growing paradigm shift on how we should re-classify HIV epidemics based on the epidemiological features that lead to HIV emergence and persistence (i.e. the 'epidemic drivers' that influence the basic reproductive ratio, R0). I examine the extent to which our traditional approaches have been underestimating the contribution of sex work to HIV spread and likely misclassifying epidemic type by developing dynamic mathematical models of HIV transmission and simulating a large number of plausible 'synthetic' HIV epidemics. I then develop - as proof-of-concept - a novel algorithm to diagnose epidemic type using these synthetic epidemics and glean the key epidemiological data that would be most useful to help distinguish between 'epidemic drivers', and therefore would be most useful to collect as part of HIV surveillance and future empirical research.