Author Description

Stroke and Coronary Heart Disease (CHD) result in reduced physical functioning and health status and associated burden to the individual and society. The main objectives of this thesis are to examine the frequency of transitions between health states in older adults, with and without stroke and CHD, predictors of these transitions and how they can be used as input to a calculation of disability free life expectancy.Data from the Established Populations for Epidemiologic Studies of the Elderly (EPESE) which consists of three large epidemiological cohort studies of samples of persons over 64 years of age living in East Boston, Iowa and New Haven were used for this study.Markov process models were used to generate health expectancy estimates. Using a longitudinal method, models accounting for time, found that women and men at age 74 years with stroke will spend 1.0 years and 1.35 years disability-free respectively while women and men with CHD will spend 1.19 years and 1.26 years respectively disability-free over the next five years. Women with and without disease experience more disability compared to men in these groups.Variables including demographic risk factors, modifiable health behaviours, symptomatology for disease, selected health characteristics and comorbidity were assessed as predictors of the probability of transitions. Final models across all five transitions considered reduced models with six important variables, age, sex, cognition, depression, BMI, smoking, alcohol consumption. Advanced age was associated with poorer functioning. Women with stroke who are eighty years of age or older, are depressed with impaired cognition, together with a BMI ≥30 kg/m2 were at most risk for moving to a worsened physical state and less likely than persons without these risk factors to move to an improved state.The cross sectional analysis, models without time, produce life expectancy estimates that are very close to estimates from the longitudinal data. If the gold standard of these methods is the Markov chain method including time, then the cross sectional estimates can be used as a close approximation.