Frailty Modeling of Canadian and Swedish Mortality at Adult and Advanced Ages

Kirill F. Andreev, United Nations
Robert R. Bourbeau, Université de Montréal

A newly established Canadian database provides an opportunity to study survival in Canada at adult and advanced ages on a cohort basis. Age-specific mortality patterns at ages 65 and over have been analyzed with a four parameter frailty model incorporating Gompertz baseline mortality and Aalen-Hougaard family of frailty distributions. The results suggest the model fits data better than Gompertz due to capturing deceleration in the rate of mortality increase. However, no particular form of this model including traditional gamma-Gompertz model has emerged as the best fitting model in all cases. A particular type of the best fitting model depends highly on population, starting age, and type of life table, period or cohort. Moreover, a thorough analysis of female period data suggests that even this model is not able to capture age pattern of mortality properly because of acceleration in mortality increase in the age range 65-80.