Background
Multimorbidity affects the majority of elderly adults and is
associated with higher health costs and utilization, but how specific
patterns of morbidity influence resource use is less understood.
Objective
To identify specific combinations of chronic conditions, functional
limitations, and geriatric syndromes associated with direct medical costs
and inpatient utilization.
Design
Retrospective cohort study using the Health and Retirement Study
(2008â2010) linked to Medicare claims. Analysis used machine
learning techniques: classification and regression trees (CART) and random
forest.
Subjects
A population-based sample of 5,771 Medicare-enrolled adults age 65
and older in the United States.
Measures
Main covariates: self-reported chronic conditions (measured as none,
mild, or severe), geriatric syndromes, and functional limitations. Secondary
covariates: demographic, social, economic, behavioral, and health status
measures. Outcomes: Medicare expenditures in the top quartile and inpatient
utilization.
Results
Median annual expenditures were $4,354, and 41% were
hospitalized within two-years. The tree model shows some notable
combinations: 64% of those with self-rated poor health plus ADL and
IADL disabilities had expenditures in the top quartile. Inpatient
utilization was highest (70%) in those age 77 â 83 with mild
to severe heart disease plus mild to severe diabetes. Functional limitations
were more important than many chronic diseases in explaining resource
use.
Conclusions
The multimorbid population is heterogeneous and there is
considerable variation in how specific combinations of morbidity influence
resource use. Modeling the conjoint effects of chronic conditions,
functional limitations, and geriatric syndromes can advance understanding of
groups at greatest risk and inform targeted tailored interventions aimed at
cost-containment.