A Parametric Approach to Modeling Health Transitions

Arnold Mitnitski, Dalhousie University
Nader Fallah, Dalhousie University
Kenneth Rockwood, Dalhousie University

We present a novel approach to summarizing health transitions as a stochastic process of changes in health states. Instead of analyzing health transitions separately in each direction (improvement, worsening and death) we suggest a general parametric approach which allows estimating the probabilities of changes as a function of the current state. From the four population specific parameters, two represent health transition of survivors, and two represent the probability of death. The approach is illustrated in the representative Canadian population: (i) transitions in general health status (deficits accumulation), (ii) transitions in cognitive states represented by the errors in the Modified Mini Mental Scale Examination Score, and (iii) transitions between cognitive impairments of different degrees. The performances of the models are very high, R2 ranging from 0.89 to 0.99, and robust across settings. This approach has a potential for the wide range of applications in analyzing age-related health transitions.