FORECASTING MODELS FOR ACUTE AND LONG TERM CARE

点击翻译按钮获取中文摘要

THIS IS A SHANNON AWARD PROVIDING PARTIAL SUPPORT FOR THE RESEARCH PROJECTS THAT FALL SHORT OF THE ASSIGNED INSTITUTE'S FUNDING RANGE BUT ARE IN THE MARGIN OF EXCELLENCE. THE SHANNON AWARD IS INTENDED TO PROVIDE SUPPORT TO TEST THE FEASIBILITY OF THE APPROACH; DEVELOP FURTHER TESTS AND REFINE RESEARCH TECHNIQUES; PERFORM SECONDARY ANALYSIS OR AVAILABLE DATA SETS; OR CONDUCT DISCRETE PROJECTS THAT CAN DEMONSTRATE THE PI'S RESEARCH CAPABILITIES OR LEND ADDITIONAL WEIGHT TO AN ALREADY MERITORIOUS APPLICATION. THE ABSTRACT BELOW IS TAKEN FROM THE ORIGINAL DOCUMENT SUBMITTED BY THE PRINCIPAL INVESTIGATOR. DESCRIPTION: This application aims at the creation of yearly projections of elderly persons in the United States by age, sex, chronic physical and cognitive disability intensities and durations and levels of use of acute and long-term care services. Transition rates will be based on a variety of national surveys, including the National Long-Term Care Survey (NLTCS) (1982,84,89, and 94), the Medicare Current Beneficiary Survey (1991 and later) and the National Longitudinal Study of Aging. Data on non-elderly populations will come from the Survey of Income and Program Participation (1984 and 1990 supplements on disability) and other sources. The starting state for the projections will be based on Public Use Microdata Sample files from the 1990 Census, which will also supply some targets for validation exercises. Six types of models or forecasting techniques will be used for projections of increasing structural complexity. These are described on pages 41 to 43. They are: 1) Application of cross-sectional prevalence rates to existing official population projections for the U.S. Census Bureau and Social Security Administration. 2) Cohort component Markov chain projections with unchanging transition rates defined for a small number of disability states. 3) Cohort component Markov projections with parameterized multiplicative Poisson submodels, smoothed by empirical Bayes mixing. 4) A Grade of Membership Model extension of the projections under Model 3, which involves a complex form of pooling and remixing of transition matrices (still, apparently, constant over time). 5) Two-component stochastic process projection models for GoM scores for individuals. 6) A trajectory model that replaces the GoM score with an age and sex specific matrix multiplying a time-invariant individual- specific trajectory model.