Table 1.

Dementia Policy Model Structure and Inputs

Step 1: Disease Progression
Model Parameter	Description	Data Source
Cognition Function Behavioral and psychological symptoms	Linear mixed effects models directly estimated from data (Supplementary Table S4)	National Alzheimer’s Coordinating Center6,13
Step 2: Care Transitions
Model Parameter	Description	Data Source
Community to long-term care facilitya	Weibull survival model directly estimated from data (Supplementary Table S7)	National Alzheimer’s Coordinating Center6
Facility to communityb	0–90 days = 0.13 90–180 days = 0.009 180–365 days = 0.003	Arling et al.17
Medicare to Medicaidc	community-dwelling = 0.00206residing in long-term care facility = 0.01056	Lim et al. and Spillman et al.18,19
Step 3: Time Spent Caregiving and Expenditures
Model Parameter	Description	Data Source
Time receiving informal caregiving	Regression model from literature	Jutkowitz et al.14
Time receiving formal caregiving	Regression model directly estimated from data (Supplementary Table S8)	Aging, Demographics and Memory Study5
Medicare expenditures	Regression model from literature	Jutkowitz et al.15
Out-of-pocket medical expenditures	Regression model from literature	Jutkowitz et al.14
Long-term care facility expenditures	private pay = $7,270/month Medicaid pay = $6,236/month	MetLife and American Health Care Association22,24
Medicaid community expendituresd	$900/month	Garfield et al. and Bharmal et al.25,26
Step 4: Mortality
Model Parameter	Model Estimate/Estimation Method	Data Source
Mortality rates	Age-, sex-, and race-mortality rates and dementia specific hazard rate	US life tables and Brookmeyer et al.7,20

Notes: Persons with dementia were individually simulated. At point of entry (i.e., diagnosis) the model generated the characteristics of the person with dementia (age, gender, education, race, marital status, region of residence, insurance status, household income, number of children, comorbidities) and characteristics of the primary caregiver (if the caregiver lives with the person with dementia, and the relationship between the person with dementia and caregiver). During each monthly cycle an individual’s cognitive and functional abilities and number of behavioral and psychological symptoms were determined (Step 1). The clinical features and personal characteristics were used to determine transitions (Step 2). Personal characteristics, the clinical features, place of residence, and insurance status, were used to estimate cost of care (Step 3). If an individual was predicted to survive (Step 4) the cycle, then they repeated Steps 1–4. If they were predicted to die, then they exited the model.

^aModels the risk of long-term care facility admissions excluding admissions for Medicare covered skilled nursing care.

^bIf a person with dementia did not leave the long-term care facility within a year it was assumed they remained in the facility for life.

^cOnce an individual was dual-eligible it was assumed they would enroll in Medicaid and remain on Medicaid for life. Background Medicaid transition risk for community-dwelling individuals without dementia was 0.0008. Individuals with dementia had an excess transition risk (hazard ratio 2.575). All individuals residing in a facility had a 0.0085 added Medicaid transition risk.

^dMedicaid expenditures for those with dementia residing in the community. In counterfactual analyses Medicaid expenditures for those without dementia were $810.