ANALYZING MORTALITY IN LONDON - LIFE-TABLES WITH FRAILTY

This paper investigates differentials in mortality between areas and over time by using life-table methods. These methods are applied to ages at death and accompanying individualized covariates ('microlevel' analysis) and to aggregated death-rates. A particular aim is to assess the impact of differences in endowment for longevity ('frailty') on the coefficient expressing the increasing chance of death as age increases (the age gradient), on coefficients expressing the effect of other attributes (e.g. place of birth or social class) on the risk of death, and on life expectancy (essentially the average age at death). The Gompertz model for adult mortality is applied to mortality contrasts between the 32 London boroughs in 1988-90 by using both individual demographic attributes and small area socioeconomic indicators as covariates to predict the risk of death. A parallel analysis is undertaken of trends in mortality, using London level data back to the 1970s. Both parametric and nonparametric methods to represent frailty are used. These have the effect of absolutely enhancing both the age gradient and the effect of certain individual attributes as compared with life-tables which do not allow for differences in frailty. However, in contrast with some other studies, there is no universal tendency for a reduction in expectancy when frailty is included.