A multistate model to project elderly disability in case of limited data

BACKGROUND Prevalence of disability depends on when a person becomes disabled (disability incidence) and when he or she dies (mortality). Multistate projection models can take into account both underlying processes of disability prevalence. The application of these models, however, is often hampered by high data requirements. OBJECTIVE This paper describes a generic estimation procedure for calculating disability incidence rates and mortality rates by disability status from data on disability prevalence and overall mortality. The procedure allows for the addition of risk factors. METHODS We estimate disability incidence rates from disability prevalence and mortality rates by disability status using prevalence data on disability from SHARE and mortality data from Eurostat and the Rotterdam Study of Health (ERGO). We use these rates to project future trends of ADL-disability prevalence among the elderly in the Netherlands for the period 2008–2040 using the multistate projection model LIPRO. RESULTS This paper shows that even in the case of limited data, multistate projection models can be applied to project trends in disability prevalence. In a scenario that assumes constant disability incidence rates, disability prevalence among the elderly will increase even though the mortality rates of disabled persons exceed those of non-disabled people. In a 1 Netherlands Interdisciplinary Demographic Institute (NIDI-KNAW)/University of Groningen, the Netherlands. E-Mail: [email protected]. 2 Netherlands Interdisciplinary Demographic Institute (NIDI-KNAW)/University of Groningen, the Netherlands. 3 Nursing home physician, Stichting Groenhuysen, Roosendaal, the Netherlands. Van der Gaag et al.: A multistate model to project elderly disability in case of limited data 76 http://www.demographic-research.org scenario that assumes declining incidence rates at the same pace as declining mortality rates, disability prevalence will be significantly lower. This latter scenario results in an almost similar decline in disability prevalence as the scenario assuming a strong reduction of age-specific obesity among the elderly. One conclusion, therefore, could be that the prevalence of obesity should be seriously reduced to reach a strong reduction of disability incidence. CONCLUSIONS The strength of this method to calculate disability incidence-rates based on disability prevalence-rates is that the relationship between changes in mortality and changes in disability is taken into account, and that the effects of risk factors can be estimated. The improved transparency of the projections, the generic nature of the model and the applicability to all (European) countries with disability prevalence data make it a useful instrument for making plausible projections of future patterns of disability prevalence based on disability incidence.