Gradvist ændrende sæsonvariation af kardiovaskulære sygdomme: - Et dansk nationalt kohorte studie
Translated title
Gradually changing seasonal variation of cardiovascular diseases: - A Danish nationwide cohort study
Author
Christensen, Anette Luther
Term
4. term
Education
Publication year
2009
Pages
144
Abstract
This thesis investigates seasonal variation in the incidence of cardiovascular diseases in Denmark from 1980 to 2008 using nationwide hospital registry data. It focuses on first events of acute coronary syndrome, stroke, and venous thromboembolism. Non-Gaussian state space models are applied to separate long-term trends (modeled with cubic splines), seasonal patterns (a single annual cycle in crude analyses and four cycles per year in analyses stratified by sex and age), and a day-of-week effect treated as unstructured seasonality. Model adequacy is assessed through residual analyses. The findings indicate that seasonal patterns in incidence change gradually over the study period. The thesis also includes methodological contributions: implementations of Kalman forecasting and the EM algorithm, and a simulation study comparing geometric models with Poisson regression for modeling seasonality, suggesting Poisson regression performs better with small data sets.
Dette speciale undersøger sæsonvariation i incidensen af kardiovaskulære sygdomme i Danmark fra 1980 til 2008 ved hjælp af landsdækkende hospitalsregisterdata. Fokus er på førstehændelser af akut koronart syndrom, slagtilfælde og venøs tromboemboli. Der anvendes ikke-Gaussiske state space-modeller til at adskille langsigtede trends (modeleret med kubiske splines), sæsonmønstre (én årlig cyklus i overordnede analyser og fire årlige cyklusser i analyser stratificeret efter køn og alder) samt en ugedagseffekt behandlet som ustruktureret sæsonvariation. Modeltilpasningen vurderes via residualanalyser. Resultaterne indikerer, at sæsonmønstre i incidensen ændrer sig gradvist gennem studieperioden. Specialet rummer også metodiske bidrag med implementering af Kalman-prediktion og EM-algoritmen samt et simulationsstudie, der sammenligner geometriske modeller med Poisson-regression til modellering af sæsonvariation, og som tyder på, at Poisson-regression er at foretrække ved små datasæt.
[This apstract has been generated with the help of AI directly from the project full text]