TY  - JOUR
A1  - Caswell, Hal
A1  - van Daalen, Silke
T1  - Healthy longevity from incidence-based models: More kinds of health than stars in the sky
Y1  - 2021/07/29
JF  - Demographic Research
JO  - Demographic Research
SN  - 1435-9871
SP  - 397
EP  - 452
DO  - 10.4054/DemRes.2021.45.13
VL  - 45
IS  - 13
UR  - https://www.demographic-research.org/volumes/vol45/13/
L1  - https://www.demographic-research.org/volumes/vol45/13/45-13.pdf
L2  - https://www.demographic-research.org/volumes/vol45/13/45-13.pdf
L3  - https://www.demographic-research.org/volumes/vol45/13/files/readme.45-13.txt
L3  - https://www.demographic-research.org/volumes/vol45/13/files/demographic-research.45-13.zip
N2  - Background: Healthy longevity (HL) is an important measure of the prospects for quality of life in ageing societies. Incidence-based (cf. prevalence-based) models describe transitions among age classes and health stages. Despite the probabilistic nature of those transitions, analyses of healthy longevity have focused persistently on means (``health expectancy''), neglecting variances and higher moments.   

Objective: Our goal is a comprehensive methodology to analyze HL in terms of any combination of health stages and age classes, or of transitions among health stages, or of values (e.g., quality of life) associated with health stages or transitions.   

Methods: We construct multistate Markov chains for individuals classified by age and health stage and use Markov chains with rewards to compute all moments of HL.   

Results: We present a new and straightforward algorithm to create the multistate reward matrices for occupancy, transitions, or values associated with occupancy or transitions.  As an example, we analyze a published model for colorectal cancer. The possible definitions of HL in this simple model outnumber the stars in the visible universe. Our method can analyze any of them; we show four examples: longevity without abnormal cells, cancer-free longevity, and longevity with cancer before or after a critical age. 
 

Contribution: Our methods make it possible to analyze any incidence-based model,with any number of health stages, any pattern of transitions, and any kind of values assigned to stages. It is easily computable, requires no simulations, provides all the moments of healthy longevity,  and  solves the inhomogeneity problem.    

ER  -