The central theme of this thesis is the longitudinal analysis of health and healthcare data.
Chapter 2 uses the first wave of, and latest longitudinal follow-up to, the Health and Lifestyle Survey (HALS) to investigate the social gradient in cancer, considering both lifetime incidence and duration models of time-to-cancer -- healthy time lived before developing cancer. Contrary to previous claims regarding the relationship between circumstances and the development of cancer, such as Deaton (2002) and Wilkinson (2010), a social gradient in time-to-cancer is observed, with those in the lowest two social classes developing cancer approximately 15% sooner than individuals in the highest social class. This relationship holds after excluding smokers from the sample. No significant gradient is observed when only lifetime incidence of cancer is considered.
Chapter 3 investigates the relationship between smoking and ill-health, with a focus on cancer outcomes. A discrete latent factor model for smoking and health outcomes, allowing for these to be commonly affected by unobserved factors, is jointly estimated, using the British Health and Lifestyle Survey (HALS) dataset. Post-estimation predictions suggest the reduction in time-to-cancer to be 5.7 years for those with a smoking exposure of 30 pack-years, compared to never-smokers. Estimation of posterior probabilities for class membership show that individuals in certain classes exhibit similar observables but highly divergent health outcomes, suggesting that unobserved factors in this model substantially determine these outcomes. The use of a joint model changes the results substantially. The results show that failure to account for unobserved heterogeneity leads to differences in survival times between those in different social classes and with different smoking exposures to be overestimated by more than 50% (males, with 30 pack-years of exposure).
Chapter 4 uses Hospital Episode Statistics, English administrative data from the Department of Health, to further investigate the red herring thesis, as advanced by Zweifel (1999). We use a sample of over 100,000 individuals who used healthcare in the financial year 2005/06 and had died by the end of the financial year 2012/13. We use a panel structure to follow individuals over seven years of this administrative data, containing estimates of inpatient healthcare expenditures (HCE), information regarding individuals' age, time-to-death (TTD), and morbidities at the time of their admission. We find that, while TTD might better explain HCE than does age, TTD itself merely proxies for individuals' morbidities, and no longer explains differences in HCE once we condition on morbidities. Our results point to an important role for including estimates of future changes in morbidity when estimating future HCE.
