Facial injury is less common in childhood than adulthood. However, it is still a significant cause of morbidity and presentation in hospital emergency departments. The pattern, time trends, and key socio-demographic determinants of facial injuries in Scottish adults admitted to hospital have previously been reported but this is not the case in the paediatric population and the question of whether such injuries are equally distributed across all socio-economic groups has not been answered. In contrast to the epidemiology of facial injuries in the paediatric population, traumatic dental injuries in children and adolescents have become one of the most frequent forms of treatment in dental practice. This suggests that traumatic dental injuries are common in childhood and are a significant cause of morbidity and presentation in hospital emergency departments. The relationship between socio-economic circumstances and the commonest dental disease (dental caries) in Scotland is well established and children resident in the most deprived areas experience more dental disease. However, this relationship with regard to dental injuries in Scotland has not yet been established. In the same way, many answers to questions regarding the sequelae and prognosis of pulpal and endodontic treatment among traumatised teeth remain unclear. Furthermore, numerous studies have been published examining the long term variables that might have an effect on treatment outcomes of avulsed teeth, but none have yet been able to estimate the long term prognosis of avulsed and replanted teeth. Two studies presented in this thesis are largely the result of population based studies investigating the epidemiology of facial and dental injuries in relation to individual demographics. An additional two studies explore the occurrence of paediatric dental trauma and its general treatment outcomes with a specific focus on estimation of the long term prognosis of avulsed and replanted teeth. To establish the epidemiology of paediatric facial injuries in children and adolescents requiring in-patient hospital admission in Scotland and to report the pattern, time trends, and key socio-demographic determinants of paediatric facial injury in Scotland, a descriptive epidemiological population-based study of the incidence of facial injuries in Scotland (2001-2009) was undertaken. Poisson regression models were employed to assess trends. There were 45,388 (4.7 per 1000 population) facial injury cases registered (2001-2009). 60% of injuries were due to non-intentional causes, 15% to motor vehicle incidents, and 9% to assault. 4.5% were alcohol related. The incidence decreased over time from 5.5/1000 in 2001 to 4.0/1000 in 2009. The risk ratio (RR) for males was 1.98 times greater than females (p < 0.001). RR varied significantly between Health Board areas from 0.68 (Dumfries and Galloway) to 1.76 (Grampian) (p < 0.001). There was a significant association between facial injury and deprivation (p<0.001); SIMD 1 (most deprived) had the highest incidence (6.3 per 1000 population; RR =1.89). The findings of this study provide evidence of a continuing increase in the burden of facial injuries in males especially for those who are living in areas of social deprivation, and in certain geographical areas of Scotland. A population-based investigation was undertaken to investigate the pattern and time trends of dental injuries and their socio-demographic determinants among primary one (P1) children in Scotland. Records of Scottish Health Boards' Dental Epidemiological Programme (SHBDEP) and National Dental Inspection Programme (NDIP) for the period 1993 to 2007 were retrieved from the Dental Health Services Research unit - Dundee. Annual incidences of dental injuries were calculated by age, gender, Health Board and DEPCAT (Carstairs deprivation categories). 68,354 P1 children were examined and only 405 (0.6%) had suffered dental injuries (5.9 per 1000 population). There was a significant decrease in incidence over time (1993 figures were three times greater than 2007). Virtually the same incidence rates were recorded for the two genders. However incidence varied significantly between Health Boards (p<0.001); the highest rate being reported in Dumfries (14.2 per 1000 population), which was 11 times greater than Ayrshire (1.3 per 1000 population). There was no significant association between risk of dental injuries and deprivation; in DEPCAT 1 (most affluent) the incidence rate was 6.4 per 1000 population, while in DEPCAT 7 (most deprived) the incidence rate was 5.7 per 1000 population. The findings of this study provide evidence that the incidence of dental injuries had significantly decreased between 1993 and 2007; gender and deprivation level had no effect on the incidence and risk of dental injuries. A retrospective investigation was undertaken to study the sequelae of non-surgical root canal treatment in traumatised anterior permanent teeth of patients referred to a secondary referral centre. The department dental trauma database was used to randomly identify patients who had sustained dental trauma to their permanent anterior teeth between 1994 and 2008 which required pulpal intervention. A data extraction form was designed and completed for each tooth, and then the data was transcribed and processed. The association between treatment outcomes and clinical variables was studied. 100 permanent anterior teeth (72 patients) were studied. Dental trauma was frequent in the age group 9-11yrs (53.9%). Upper central incisors were the most common teeth involved (43.8%). The male: female ratio was 2:1 with an average age at the time of trauma of 10.31 yrs (SD 2.16 yrs). Home and immediate home environs were the commonest location (18%) while falls (34.8%) and injuries during sport/play (34.8%) were the commonest causes. The commonest injuries in this randomly selected group were enamel-dentine fracture with pulp exposure (34.8%) and avulsion (28%). 66.3% received a first treatment intervention less than 24 hours following the injury. Root canal treatment was the most frequent treatment provided, especially for dental avulsion cases (100%). Treatment outcomes were split into three categories: Success (53.4%); Short-term success but long-term failure (35.6%); and Failure (11%). Significantly fewer failures occurred with: developing roots compared to completed roots (P=0.05); a good quality temporary filling (P<0.003); no mobility (P<0.001); and less than one hour extra alveolar dry time (P =0.02). No significance was reached with regard to: condition of root canal (P=0.095); extra alveolar time (EAT) (P=0.191); and type of storage medium (P=0.43). To assess and identify early clinical variables that are most predictive of treatment outcomes for avulsed and replanted permanent anterior teeth and to develop a model that will allow estimation of treatment outcome based on these variables, a retrospective study was designed and undertaken, where the dental trauma database was used to randomly identify patients who had sustained dental trauma on their permanent teeth leading to avulsion between 1998 and 2007. A data extraction form was designed and completed for each tooth. Demographic, diagnostic and treatment information recorded in the patient’s records, in addition to radiographs, were viewed and then transcribed and processed. The significance for each early clinical variable was assessed using a univariate logistic regression model. Only significant variables (P ≤ 0.05) were considered eligible for the prediction model and a c-index was then constructed for their respective predictive power. 213 patients who had received treatment for avulsed and replanted teeth between 1998 and 2007 were studied and only 105 fulfilled the criteria for evaluation.

Abstract: Ecological influences, observational caries epidemiological trends and associated socioeconomic and geographic dental health inequalities at five years of age in Scotland, 1993/94-2007/08 Introduction: In recent years many national Governments have called for health improvements at the population level and at the same time reductions in health inequalities. To date, dental epidemiology has concentrated mainly on tracking trends in dental health. Methodologies relating to dental health inequalities are, however, not well established. Within Scotland, over the past decade, children’s oral health improvement programmes have been established at national level. Preceding and concurrent with these developments, similar initiatives have been implemented within Greater Glasgow NHS Board. This is Scotland's largest NHS Board with the highest proportion of Scotland’s socio-economic status (SES) deprived population. Recent reports from the National Dental Inspection Programme (NDIP) for five-year-olds show improvements in dental health. The above conditions provide the opportunity to explore dental trends in more detail at geographic level within Scotland and to investigate dental health inequality methodologies within the context of health improvement programmes and overall improvements in dental health. Aim: To examine caries epidemiology data and apply and appraise a range of tests of health inequality to data from Primary 1 (P1) five-year-old children in Scotland during the period 1993/94-2007/08, against a background of health improvement programmes. Furthermore, to apply the selected inequalities tests to the caries data for a) Scotland as a whole and b) the geographic subgroups: 1] Glasgow (GGHB) and 2] the remainder of Scotland, outwith Glasgow (Not-Glasgow). Methods: Secondary analyses were performed on eight successive cross-sectional NDIP five-year-olds' caries datasets, 1993/94 to 2007/08. These permitted both SES and geographic trends in mean d3mft and % dmft=0 to be plotted for the areas: Scotland, GGHB and Not-Glasgow. The metrics selected to model dental health inequalities were: the Significant Caries Index (SIC) and modified SIC10, the Receiver Operator Curve (ROC), the Gini coefficient, the Concentration Curve (CC), Koolman and Doorslaer's transformed Concentration Index (CI), the Slope Index of Inequality (SII), the Relative Index of Inequality (RII) and the Population Attributable Risk (PAR). Odds Ratios and Meta-analyses using Generalised Linear Modelling assessed statistical-inference for dental health and inequality trends. Results: Overall, usable data was retrieved for 68,398 five-year-old subjects (n=18,174 from GGHB; n=50,224 from Not-Glasgow). In Scotland as a whole, marked SES gradients in caries prevalence and caries burden were related to the DepCat score of children’s home postcode. Between the start and endpoints of the study, the simple absolute SES inequality in mean d3mft between the most affluent and most deprived groups decreased (p<0.02), whilst mean d3mft reduced across the entire SES spectrum. Relative to the baseline year (1993), by 2007, the Odds Ratios for d3mft>0 in Scotland decreased (p<0.0001) to 0.43 (95%CI, 0.40-0.46). Although Scotland's simple absolute SES related dental health inequality (DHI) decreased for mean d3mft (p<0.02), there were no improvements in simple relative SES DHIs over this time period. Simple absolute and simple relative geographic inequalities in weighted %d3mft=0 and mean d3mft were seen when GGHB was compared with Not-Glasgow data. These geographic inequalities metrics tended to increase from 1993/94 until 1999/00. However, by 2007/08 reductions in simple absolute geographic inequality were observed, with marginal improvements in simple relative geographic inequality compared to baseline. Additionally, simple absolute and relative geographic inequality in SIC scores decreased overall against a background of SIC improvements in both GGHB & Not-Glasgow (Meta-analysis, p<0.01, respectively). By 2007/08, relative to 1993/94, Odds Ratios for d3mft>0 in the geographic subgroups GGHB and Not-Glasgow decreased, respectively (p<0.0001), to 0.31 (95%CI, 0.26-0.38) and 0.46 (95%CI, 0.43-0.50). There was evidence of a 'Glasgow (dental health) Effect', whereby GGHB children’s dental health was poorer than in Not-Glasgow during the period 1993 to 1999, after controlling for confounding factors (p<0.01). This ‘Glasgow Effect’ was no longer evident by 2007. Modelling caries data using the complex inequality metrics has given further insights into different dimensions of geographic and SES-related dental health inequalities. For example, in each area from 1993/94-2007/08, the full SIC10 distributions showed respective decreases in complex absolute DHI in affected individuals in population deciles (irrespective of SES). Simultaneously, Scotland's SII indicated that complex absolute SES inequalities decreased (p<0.02). Furthermore, in Glasgow the %PAR decreased by 24 percentage points, itself impacting on Scotland's decreased PAR. However, the RII and transformed CI indicated that complex relative SES DHI increased in each area over the period of study. The ROC, CC & RII plots were comparatively stable over time for Scotland, compared to trends in the GGHB subgroup. There was evidence of some variation in DHI, and the Gini-coefficient (for individual DHI) was counter-intuitive. Discussion: Analysis and interpretation of simple and complex absolute and relative DHI outcomes are not straightforward against a background of population dental health improvements across the SES spectrum. If equivalent absolute dental health improvements are achieved in the best and poorest d3mft groups, as %d3mft>0 and mean d3mft diminish in the denominator group it is increasingly difficult to achieve improvement in relative inequalities. Nonetheless, tests suggest that simple absolute geographic DHI in Scotland's P1's weighted %d3mft=0 and mean d3mft have improved, while simple relative geographic inequality has not deteriorated over the interval 1993-2007. Further insights were obtained from examination of the cross-sectional distributions of SIC10. These showed improvements in complex absolute individual inequality across all population deciles with d3mft>0, over time, at each geographic level. Moreover, comparison of the geographic SIC10 scores for the worst affected deciles demonstrated reductions in simple absolute and relative geographic DHI in five-year-olds' d3mft morbidity for those with the poorest dental health outcomes in 2007 vs. 1993. Furthermore, Scotland's complex absolute SES-related DHI has decreased over time when assessed by SII. Improvements in complex absolute SES-related DHI have occurred more readily than improvements in complex relative SES-related DHI. Conclusions: For the first time, these multiple tests of inequality have been applied to Scotland’s and Glasgow’s child caries datasets. Generally, caries epidemiology trends occurred slowly and smoothly, however, DHI trends from this same data tended to fluctuate (especially in the geographic subgroups). The apparent lack of consonance of the various inequalities metrics demonstrates that measurement, understanding and interpretation of population DHI trends are complicated and require knowledge of the underlying epidemiology trends. Nonetheless, with the exception of the Gini, all results provided useful information which aid understanding of DHI. The complex measures such as the SII and RII had the advantage of using all the available d3mft information within the DepCat domains and weighting results for SES within the denominator populations. Furthermore, in Scotland as a whole, the SIC10 distribution, SII and RII appear to exhibit stable DHI trends, against the background populations' dental health improvements.

The aim of this research was to explore the potential of routinely collected primary care data to support the identification of individuals for cardiovascular risk reduction. The work involved a systematic literature review of reminder interventions operating at the point of care; a randomised controlled trial of a novel software tool to facilitate the targeting of individuals at risk of cardiovascular disease; and an exploration of qualitative issues relevant to the challenge of cardiovascular risk reduction in current practice. The Systematic review resulted in a narrative synthesis and a meta-analysis. It concluded that reminder interventions are generally effective at changing practitioner behaviour, but the effect is inconsistent, probably dependent on organisational context, and difficult to predict. The e-Nudge trial involved 19 practices in Coventry and Warwickshire, who used the e-Nudge software tool for two years. This tool was programmed for the project by the clinical software company EMIS. Whilst the primary outcome (cardiovascular event rate) was not significantly reduced in this timescale, a beneficial effect was demonstrated on the adequacy of data to support risk estimation and on the visibility of the at risk population. A new means of addressing the problem of undiagnosed and late-diagnosed diabetes was also discovered. Qualitative aspects of this area of care are presented through a discussion of ethical issues, a limited series of interviews with members of the public included in the appendix, and extensive field notes taken throughout the research. These provide some context in support of the e-Nudge trial. Routinely collected data of UK general practice provide a potentially rich resource to support primary cardiovascular disease prevention, but practical, ethical and conceptual issues must all be addressed to optimise their impact. This conclusion forms the thesis to be explored and justified through this dissertation.

Postal questionnaires offer one of the least expensive modes of collecting patient based outcomes in health care research. Many methods of increasing response to questionnaires used in educational and market research surveys have been tested. Behavioural theories have also been applied to survey research to understand response decisions. Little attention, however, has focussed specifically on response issues to postal questionnaires used to collect data in clinical trials. This is the subject of this thesis. A systematic review of methods of improving response to postal questionnaire follow-up in health care studies was conducted. A method of improving response was then devised and its effectiveness was tested within an existing clinical trial (the Collaborative Ankle Support Trial - CAST). This method was a 'Trial Calendar'which was a prompting and reminder tool to encourage response. Qualitative data were gathered from clinical trial participants to ascertain factors influencing their response decisions. Finally, the socio-demographic characteristics of CAST participants were examined. The systematic review demonstrated that follow-up reminder systems had the most significant effect on response rates (RR 1.82, Cl 95% 1.11 to 2.99). Incorporating such reminders into a tool such as the 'Trial Calendar', however, had no effect on improving response in CAST. The qualitative study revealed aspects of behavioural theories which could be incorporated into trial information and appeals for response. Analysis of the sociodemographic characteristics of CAST participants revealed that the youngest age group (16-24 years) was less likely to respond at every followup point. It is concluded that rather than anticipating low response rates and striving to devise methods of converting non-responders into responders, efforts should be directed at preventing participants becoming non-responders in the first place. This thesis argues for the area of follow-up to postal questionnaires in clinical trials to become a theoretical research issue in its own right.

This thesis presents designs for a series of clinical trials where instead of designing clinical trials individually, each of the trials is designed as part of a series of trials. The framework of the design is based on a combination of classical frequentist and Bayesian approaches which is sometimes known as the hybrid approach. The unknown parameter of the treatment efficacy is assumed to be random and follows a prior distribution in the design stage but at the end of the trial a frequentist test statistic is used on the observed data to infer the parameter. The design introduced in Chapter 5 aims to determine an optimum sample size for each trial by optimizing the average power of each trial and the overall resources while fixing the conventional type I error. The design has the exibility to either run sequentially or concurrently. The design is then extended to allow interim analyses in each trial (Chapter 6). The focus of the extended design is on a series of Bayesian decision-theoretic phase II trials and one frequentist phase III trial. At each interim stage, a decision is made based on the expected utilities of subsequent actions. There are four possible actions to choose from, namely, to continue the current trial by recruiting more patients, to initiate a new phase II trial, to abandon the development plan or to proceed to a phase III trial with this treatment against a control arm. For the last action, the phase III trial is designed with the hybrid methodology as described above. Finally, the prior distributions for each treatments are assumed to be correlated and as information is gathered from the previous and current trials, the current and following prior distributions are updated (Chapter 7).

During the transition from childhood to adulthood young people are vulnerable to relatively unique health problems and risks, whilst also developing personal autonomy, and learning when and how to access health services. This thesis describes three studies, the overall aim of which was to gain a greater understanding of how teenagers use general practice, and to identify those factors that influence this behaviour and ultimate health outcomes. The aim of the first two studies was to identify demographic, health, and attitudinal factors associated with both overall general practice consultation rates and consultation for specific types of condition. The first study involved identifying annual consultation rates and reasons for consultation from the medical records of 836 (94.4%) of 886 teenagers aged 13 to 15 registered with five general practices across the East Midlands, covering the 12 month period prior to the second study. In the second study, 713 (80.5%) of the 886 teenagers from the first study responded to a postal questionnaire survey, carried out between May and August 1997, which explored health concerns, help-seeking behaviour, health related behaviour, use of health services, and attitudes to general practice. The results from each of the first two studies were linked, in order to identify associations between self-reported health status, attitudes and behaviour and recorded consultation behaviour, based on the 678 teenagers for whom complete data sets were available. The median annual consultation rate was two, with 76.1% of teenagers consulting at least once and 23.8% consulting on four or more occasions. Consultation rates increased with age amongst girls, who had significantly higher rates than boys by age 15. Most consultations were for respiratory and skin problems, with consultations for psychological problems being least frequent. Teenagers reported a wide range of health concerns and, whilst general practitioners were identified as the most frequent source of health advice from formal health services, friends and family were cited far more frequently. Although 91.8% of survey respondents rated confidentiality as important, there was no association between attitudes towards confidentiality and actual consultation behaviour. In contrast, teenage girls who expressed concerns about embarrassment were less likely to consult about gynaecological problems and contraception. The third study was a case control study in which the general practice consultation patterns of 240 young women who subsequently became pregnant (having a recorded termination, delivery or miscarriage between January 1995 and January 1998) were compared with those of 719 age-matched controls without a history of pregnancy. Cases were significantly more likely than controls to have consulted in the year prior to conception with 93% of cases consulting at least once and 71% having discussed contraception at some time. Teenagers whose pregnancy ended in a termination were significantly more likely than controls to have received emergency contraception. In conclusion, whilst teenagers have been shown to use general practice for a range of health problems, the results from these studies suggest that there is a need to facilitate access for teenagers with more sensitive problems, and to improve identification and follow-up of those at greatest risk of adverse outcomes.

This thesis investigates the representation of a stochastic epidemic process as a directed random graph; we use this representation to impute the missing information in final size data to make Bayesian statistical inference about the model parameters using MCMC techniques. The directed random graph representation is analysed, in particular its behaviour under the condition that the epidemic has a given final size. This is used to construct efficient updates for MCMC algorithms. The MCMC method is extended to include two-level mixing models and two-type models, with a general framework given for an arbitrary number of levels and types. Partially observed epidemics, that is, where the number of susceptibles is unknown or where only a subset of the population is observed, are analysed. The method is applied to several well known data sets and comparisons are made with previous results. Finally, the method is applied to data of an outbreak of Equine Influenza (H3N8) at Newmarket in 2003, with a comparison to another analysis of the same data. Practical issues of implementing the method are discussed and are overcome using parallel computing (GNU OpenMP) and arbitrary precision arithmetic (GNU MPFR).

Breast cancer, which is the most common cancer in women, is a complex disease characterised by multiple molecular alterations. Current routine clinical management relies on availability of robust clinical and pathologic prognostic and predictive factors, like the Nottingham Prognostic Index, to support decision making. Recent advances in highthroughput molecular technologies supported the evidence of a biologic heterogeneity of breast cancer. This thesis is a multi-disciplinary work involving both computer scientists and molecular pathologists. It focuses on the development of advanced computational models for the classification of breast cancer into sub-types of the disease based on protein expression levels of selected markers. In a previous study conducted at the University of Nottingham, it has been suggested that immunohistochemical analysis may be used to identify distinct biological classes of breast cancer. The objectives of this work were related both to the clinical and technical aspects. From a clinical point of view, the aim was to encourage a multiple techniques approach when dealing with classification and clustering. From a technical point of view, one of the goals was to verify the stability of groups obtained from different unsupervised clustering algorithms, applied to the same data, and to compare and combine the different solutions with the ones available from the previous study. These aims and objectives were considered in the attempt to fill a number of gaps in the body of knowledge. Several research questions were raised, including how to combine the results obtained by a multi-techniques approach for clustering and whether the medical decision making process could be moved in the direction of personalised healthcare. An original framework to identify core representative classes in a dataset was developed and is described in this thesis. Using different clustering algorithms and several validity indices to explore the best number of groups to split the data, a set of classes may be defined by considering those points that remain stable across different clustering techniques. This set of representative classes may be then characterised resorting to usual statistical techniques and validated using supervised learning. Each step of this framework has been studied separately, resulting in different chapters of this thesis. The whole approach has been successfully applied to a novel set of histone markers for breast cancer provided by the School of Pharmacy at the University of Nottingham. Although further tests are needed to validate and improve the proposed framework, these results make it a good candidate for being transferred to the real world of medical decision making. Other contributions to knowledge may be extracted from this work. Firstly, six breast cancer subtypes have been identified, using consensus clustering, and characterised in terms of clinical outcome. Two of these classes were new in the literature. The second contribution is related to supervised learning. A novel method, based on the naive Bayes classifier, was developed to cope with the non-normality of covariates in many real world problems. This algorithm was validated over known data sets and compared with traditional approaches, obtaining better results in two examples. All these contributions, and especially the novel framework may also have a clinical impact, as the overall medical care is gradually moving in the direction of a personalised one. By training a small number of doctors it may be possible for them to use the framework directly and find different sub-types of the disease they are investigating.

In this thesis, three stochastic epidemic models for intervention for emerging diseases are considered. The models are variants of real-time, responsive intervention, based upon observing diagnosed cases and targeting intervention towards individuals they have infected or are likely to have infected, be they housemates or named contacts. These models are: (i) a local tracing model for a disease spreading amongst a community of households, wherein intervention (vaccination and/or isolation) is directed towards housemates of diagnosed individuals, (ii) a contact tracing model for a disease spreading amongst a homogeneously-mixing population, with isolation of traced contacts of a diagnosed individual, (iii) a local tracing and contact tracing model for a disease spreading amongst a community of households, with intervention directed towards housemates of both diagnosed and traced individuals. These are quantified by deriving threshold parameters that determine whether the disease will infect a few individuals or a sizeable proportion of the population, as well as probabilities for such events occurring.

Pneumonia is a common diagnosis in general practice and is associated with significant morbidity and mortality. Current estimates of pneumonia incidence in the UK are based on studies more than a decade ago and little is known about longer term outcomes in pneumonia patients. Though much is known about the aetiology of pneumonia and predictors of mortality, an emerging area for research is the relationship between commonly prescribed drugs in general practice and pneumonia. The aims of this thesis were first, to determine overall incidence and mortality for pneumonia and how these vary by socio-demographic characteristics like age, sex, deprivation; and second, to investigate whether statins, angiotensin converting enzyme inhibitors (ACEIs) and gastric acid suppressants like proton pump inhibitors (PPIs) and histamine 2 receptor antagonists (H2RAs) modify the risk of acquiring pneumonia and its prognosis. This study used data from The Health Improvement Network (THIN) database, a longitudinal database of anonymised computerised medical patient records from 330 United Kingdom (UK) general practices at the time of data extraction in 2006. A cohort design was used to determine pneumonia incidence and mortality in the UK. Case-control, case-series and cohort study designs were used to investigate associations between the various drug exposures and pneumonia. The overall incidence of pneumonia was 237 per 100,000 person-years (95 % confidence interval (CI): 235 to 239) and this rate was stable between 1991 and 2003. Pneumonia was more common in men and in children under the age of four years and adults over the age of 65 years. There was an increased incidence of pneumonia with higher levels of socioeconomic disadvantage. Pneumonia cases showed much higher all-cause mortality as compared to the general population, both in the short and long-term and this increase was independent of underlying comorbidity. After adjusting for potential confounders, current prescriptions for statins and ACE inhibitors were associated with a significant reduction in the risk of acquiring pneumonia. Current prescriptions for PPIs were associated with an increased risk of pneumonia. With regards the impact on mortality: the use of statins was associated with a lower risk of short and long-term mortality following pneumonia whereas the use of ACEIs was associated with a decreased mortality risk only in the short-term. No relationship was observed between prescriptions for PPIs, H2RAs and pneumonia mortality. This study shows that caution must be exercised while prescribing proton pump inhibitors especially in patients known to be at high risk of pneumonia. There is also a potential role for statins in preventing pneumonia in at-risk patients and improving pneumonia outcomes but this will necessitate clinical trials to determine adequate dose, duration and safety profiles before any prescribing policy recommendations are made.