Modeling Age-Related Changes from Late Middle Age in Men and Women

Arnold Mitnitski, Dalhousie University
Xiaowei Song, Natural Sciences and Engineering Research Council of Canada (NSERC)
Kenneth Rockwood, Dalhousie University

We present a new approach to modeling health transitions in mean and women from late middle age (55+). This approach is based on a parametric representation of health transitions using a modified Poisson distribution for the probabilities of health changes. In the Canadian National Population Health Survey (n=4330, 2548 women, at baseline in 1994), health status was defined by a deficit count, using 33 health-related variables. Changes in deficit count were followed up over 4 consecutive 2-year intervals to 2002. The model stratified by sex predicts health changes (improvement or worsening at any degree, and death) with an exceptionally high accuracy (R2 ~0.90). The model employs only four sex-specific parameters: two represent health transition of survivors, and two represent the probability of death. The parameters reflect the men/women difference in transitions: despite women show higher level of deficit accumulation, they better survive than men at any deficit level.