No description available.

Long-term outcomes for patients admitted to intensive care units (ICUs) are recognised to be of increasing importance. Published studies indicate that ICU survivors have significant physical impairment, impaired quality of life, and excess mortality during the post-ICU period. The period of excess mortality has been variously estimated as lasting from one to 16 years after ICU discharge. Remarkably little information about long-term mortality and healthcare resource use exists for critical care populations, and outcomes relative to a non-ICU control population are unknown. The aims of the studies presented in the thesis were (i) to describe long-term (five year) mortality and identify factors associated with mortality for patients admitted to ICUs in Scotland (ICU admission cohort) and those surviving to be discharged from hospital alive (ICU survivor cohort); (ii) to compare mortality rates with control populations after adjustment for relevant confounders; (iii) to evaluate the extent of, and factors associated with, long-term (five year) major healthcare resource use of survivors of critical illness (ICU survivor cohort); and (iv)to compare major healthcare resource use with a control hospital inpatient population. I undertook a detailed systematic review of the international literature relating to healthcare resource use in ICU survivors to inform the design of the part of the study relating to resource use. This revealed a paucity of high quality studies but led to recommendations for improving the conduct and reporting of future research in this field. Using both retrospective cohort and matched cohort study designs, I analysed data relating to all patients admitted to Scottish ICUs in 2005 from the Scottish Intensive Care Society Audit Group (SICSAG) database. Two cohorts were defined: an ICU admission cohort, representing all ICU admissions, and a subcohort of those who survived to hospital discharge (ICU survivor cohort). Matched control cohorts of non-ICU hospital inpatients were selected from national datasets. The main outcomes were five-year mortality and major healthcare resource use obtained from linkage to national datasets. Major healthcare use was measured by number of hospital readmissions, number of days spent in hospital and hospital costs during the five years after hospital discharge. Five year mortality was 53% in the ICU cohort compared with 27% for the matched control hospital cohort and 16% for an age/sex-standardised general population. Among hospital survivors, ICU patients had higher five year mortality after adjustment for confounders (HR 1.3, 95%CI 1.2 to 1.4, p<0.001). Age, comorbidity, ICU admission diagnosis and deprivation quintile were independently associated with five-year mortality. The ICU diagnosis with greatest five year mortality (relative to self-poisoning) was variceal bleeding (HR 3.9, 95%CI 2.2 to 6.7, p<0.001). The readmission rate for the 5259 ICU patients surviving to hospital discharge declined from 1.7 readmissions per person in the first year to 0.9 in the fifth year of follow-up. Overall, ICU survivors spent a mean of 29 days in hospital over the five year follow up period, at a cost of £14593 per person. Previous number of admissions was the factor most strongly associated with resource use. ICU patients had a significant increased rate of hospital admission compared with the control cohort throughout the five year follow up period (admission rate ratio 1.21 (95%CI 1.14 to 1.29, p<0.001)). In the programme of work presented in this thesis, I have systematically reviewed evidence for resource use following critical illness, and have demonstrated that ICU patients are more likely to die compared with other hospital inpatients over a five-year horizon, even when only hospital survivor cohorts are considered. Furthermore, I have demonstrated that ICU survivors utilise a significant amount of excess acute hospital resource, which is relevant to health service planning and economic evaluations.

Thesis (M.A.) in Mathematics and Statistics--University of Maine, 2002. / Includes vita. Includes bibliographical references (leaf 45).

Introduction: Globally, around 2.6 billion people have consumed alcohol in 2017. In the same year, nearly 3 million or 5% of all deaths were attributable to alcohol consumption, the majority of which were non-communicable diseases, such as cancer, digestive and cardiovascular diseases. Chronic heavy alcohol consumption in particular causes harm to the cardiovascular system and is linked to an elevated risk on the occurrence of ischemic heart diseases and cardiomyopathies. The latter constitutes a heterogeneous group of cardiovascular diseases, which can generally be characterized by a weakened heart muscle. The causal link between chronic heavy alcohol consumption and cardiomyopathy has long been recognized, with the Tenth Revision of the International Classification of Diseases (ICD-10) listing alcoholic cardiomyopathy (ACM) as a fully alcohol-attributable diagnosis. For a few, predominately high-income countries, civil registries provide valuable information of ACM mortality. However, for the majority of countries and global population, the cardiomyopathy burden attributable to alcohol consumption needs to estimated. Established methods for estimating alcohol-attributable fractions (AAF), i.e. proportion of an outcome which could be avoided in a scenario of zero alcohol consumption, could not be applied for cardiomyopathy as the link between alcohol consumption levels and risk of cardiomyopathy could not be specified. Accordingly, a global assessment of the contribution of alcohol consumption to the disease burden from cardiomyopathy was lacking. Aims and objectives: First, to develop methods for estimating the contribution of alcohol consumption to cardiomyopathy that can be used globally (study I). Second, to apply the method developed in study I to estimate the global mortality from ACM (study II). Third, to assess differences between this method and an alternative method for estimating the contribution of alcohol consumption to cardiomyopathy proposed during pursuit of these aims (study III). Design: Statistical modelling study with country-level data as unit of analyses. Study I. Based on mortality data from civil registries, the proportion of deaths from ACM among deaths from any cardiomyopathy (=AAF) was used as proxy for the link between alcohol consumption and cardiomyopathy. To generalize this link to countries without available civil registry data, associations of population alcohol exposure and registered AAF were established. Cardiomyopathy deaths that are attributable to alcohol use were quantified in those countries with available registry data. Study II. For countries without available civil registry data, ACM mortality was estimated using population alcohol exposure data based on the methods from study I. As a result, national, regional and global estimates of the mortality attributable to ACM were obtained for the year 2015. Study III. In the alternative method developed by the Global Burden of Disease (GBD) study team, the contribution of alcohol consumption to cardiomyopathy was estimated taking into account that actual ACM deaths may be incorrectly coded as so-called garbage codes (disease codes that do not accurately describe the underlying cause of death). In the alternative method, garbage codes were redistributed to both cardiomyopathy and ACM using statistical procedures. The underlying assumptions for the redistribution of garbage codes were examined by comparing registered and estimated ACM mortality data taking into account the distribution of alcohol exposure. Data sources: Data on population alcohol exposure (alcohol per capita consumption, prevalence of heavy episodic drinking, prevalence of alcohol use disorders) were sourced from publicly available World Health Organization (WHO) data bases. As outcome data, sex-specific mortality counts from different disease groups (ACM, any cardiomyopathy, and selected garbage codes) were obtained at the country level from three different sources: First, WHO mortality data base, which provide civil registry mortality data on nearly half of all member states, coded according to the ICD-10. Second and third, ‘Global Health Estimates’ and ‘GBD Results Tool’ data bases, which provide complete and consistent mortality estimates aggregated into larger disease groups for all WHO member states. Data on covariates were obtained from the United Nations and the World Bank. Statistical analyses: In study I, the dependent variable – AAF for cardiomyopathy – was calculated by dividing deaths from ACM by deaths from any cardiomyopathy, based on civil registry data from N=52 countries. Taking into account country-specific crude mortality rates of ACM, AAF were modeled in two-step sex-specific regression analyses using population alcohol exposure as covariate. AAF were estimated for the same set of N=52 countries, in addition to N=43 countries without civil registry data. Estimated AAF were compared to registered AAF available for N=52 countries. In study II, the global mortality of ACM was estimated by combining civil registry ACM mortality data for N=91 countries and estimated ACM mortality for N=99 countries without available civil registry data. For the latter set of countries, ACM mortality data were calculated by estimating AAF based on the methodology outlined in the first study and subsequently applied to all cardiomyopathy deaths. As a proxy for under-reporting of ACM in civil registries, estimated ACM deaths were compared to registered ACM deaths for N=91 countries. In study III, ACM mortality estimates from the GBD study were compared against registered ACM mortality data for N=77 countries, aiming to test underlying assumptions for redistribution of garbage-coded deaths in the alternative method. For this purpose, descriptive statistics and Pearson correlations were used to assess the association of estimated and registered deaths and to examine consistency of estimates with population alcohol exposure. Results: In study I, population alcohol exposure and ACM mortality were closely linked (spearman correlation=0.7), supporting the proposed modelling strategy. For N=95 countries, the AAF for cardiomyopathy was estimated at 6.9% (95% confidence interval (CI): 5.4-8.4%), indicating that one in 14 of all cardiomyopathy deaths were attributable to alcohol in the year 2013 or the last available year. The findings were robust, with 78% of all estimated AAF deviating less than 5% from registered AAF. In study II, it was estimated that 25,997 (95% CI: 17,385-49,096) persons died from ACM in 2015 globally, with 76.0% of ACM deaths being located in Russia. Globally, 6.3% (95% CI: 4.2-11.9%) of all deaths from cardiomyopathy were estimated to be caused by alcohol. Furthermore, indications of underreporting in civil registration mortality data were found, with two out of three global ACM deaths being possibly misclassified. In study III, findings suggested that only one in six ACM deaths were correctly coded in civil registries of N=77 countries. However, the algorithm accounting for misclassifications in the GBD study was not aligned with population alcohol exposure, which has led to implausibly high ACM mortality estimates for people aged 65 years or older. Specifically, registered and estimated ACM mortality rates diverged in the elderly, which was corroborated with decreasing correlations in these age groups. Conclusions: For countries without civil registry data, the contribution of alcohol consumption to mortality from cardiomyopathy could be quantified using population alcohol exposure and estimated mortality data for any cardiomyopathy. The proposed method was adapted by the WHO in 2018, allowing for a more complete picture of the alcohol-attributable global disease burden for nearly 200 countries. Notably, ACM mortality was hardly present in countries with low to moderate alcohol consumption levels, corroborating that ACM is the result of sustained and very high alcohol consumption levels. In civil registries, at least two out of three ACM deaths are misclassified, thus, presented mortality figures are likely underestimated. As with other alcohol-attributable diseases, misclassification of ACM mortality is a systematic phenomenon, which may be caused by low resources, lacking standards and severe stigma associated with alcohol use disorders. With transition from ICD-10 to ICD-11, new methods will be required as ACM will not remain a unique diagnosis in the new classificatory system. Future methods should account for mortality misclassifications by redistributing garbage codes while taking into consideration the distribution of alcohol exposure. Further, measures to reduce stigma may improve diagnostic accuracy for ACM and other alcohol-attributable diseases. This will not only improve public health statistics but also – and more importantly – improve health prospects of persons with heavy alcohol consumption.:Statement for a publication-based dissertation I Contents II List of tables IV List of figures V Abbreviations VI Abstract VII 1 Introduction 10 1.1 Global extent of alcohol use 10 1.2 Alcohol-attributable disease burden 11 1.3 Estimating the alcohol-attributable burden 12 1.4 Cardiomyopathy 18 1.5 Alcohol and cardiomyopathy 19 2 Aims and objectives 21 3 Study design and methodology 21 3.1 Study design 21 3.2 Data sources 22 4 Study I - Quantifying the global contribution of alcohol consumption to cardiomyopathy 25 4.1 Background 26 4.2 Methods 27 4.3 Results 32 4.4 Discussion 38 4.5 Conclusion 41 5 Study II - National, regional and global mortality due to alcoholic cardiomyopathy in 2015 42 5.1 Introduction 43 5.2 Methods 44 5.3 Results 45 5.4 Discussion 51 6 Study III - Mortality from alcoholic cardiomyopathy: Exploring the gap between estimated and civil registry data 57 6.1 Introduction 58 6.2 Experimental section 59 6.3 Results 62 6.4 Discussion 67 7 General discussion 72 7.1 Summary of the findings 72 7.2 Strengths and limitations 72 7.3 Implications for future research 75 7.4 Implications for alcohol policy 79 7.5 Outlook 80 7.6 Conclusion 81 8 References 83 9 Appendix A (study I) 97 10 Appendix B (study II) 99 10.1 Methods 99 10.2 Results 103 11 Appendix C (study III) 119 11.1 Methods 119 11.2 Results 124 12 Erklärung gemäß § 5 der Promotionsordnung 128