Table A4.

Covariates Within Each Regression Model

ACOs
Maine	Minnesota	Vermont
Common covariates
Age and age‐squared Gender Disability Beneficiary's classification on the Chronic Illness and Disability Payment System If the beneficiary was continuously enrolled (with less than 1‐month gap) during the year Number of months the beneficiary was enrolled in Medicaid during the year County‐level federal poverty level, median age, and uninsured rate Metropolitan status of the beneficiary's county County‐level hospital beds, physicians per capita, and community mental health centers per capita	Age and age‐squared Gender Disability by age group (adult/child disabled) Beneficiary's classification on the Chronic Illness and Disability Payment System If the beneficiary was continuously enrolled (with less than 1‐month gap) during the year Whether the beneficiary was enrolled for at least 9 months in the previous year County‐level federal poverty level and median age Metropolitan status of the beneficiary's county County‐level hospital beds	Age category (<1, 1 to 18, 19 to 64, 65‐plus) Gender Disability Beneficiary's classification on the Chronic Illness and Disability Payment System Number of months beneficiary was Medicaid eligible during the year (minimum of 10) If beneficiary was continuously enrolled 10 or more months in previous year County‐level federal poverty level
Model‐specific covariatesa
•Race (nonwhite and missing race) •Medicare/Medicaid enrollee •If the beneficiary had full Medicaid benefits during the year •If the beneficiary was enrolled in the ACO at some point in both test years •Method of attribution to the ACO or the comparison group, ie, whether the beneficiary was enrolled in the ACO or the comparison group because he or she had a majority of visits to a primary care provider, because he or she was enrolled in an HH, or because he or she had a majority of visits to an emergency department		•Beneficiary's participation in the Chronic Care Initiativeb •Attribution method of beneficiary (claims‐based or choice/autoassigned) •If beneficiary was attributed to a Vermont Blueprint for Health medical home •If beneficiary was eligible through Medicaid expansion
Maine BHH
Common covariates
•Age category (<1, 1 to 18, 19 to 64, 65‐plus)+ •Gender •Disability •Beneficiary's classification on the Chronic Illness and Disability Payment System •If the beneficiary was continuously enrolled (with less than 1‐month gap) during the year •Number of months the beneficiary was enrolled in Medicaid during the year •County‐level federal poverty level, median age, and uninsured rate •Metropolitan status of the beneficiary's county •County‐level hospital beds, physicians, and community mental health centers per capita
Model‐specific covariates
•Race (nonwhite and missing race) •Medicare/Medicaid enrollee •If the beneficiary had full Medicaid benefits during the year •If the beneficiary was enrolled in the BHH at some point in both test years

^aCovariates vary slightly by model due to differences in data availability and appropriateness for the model. Specifically, race is included only in Maine because it was not available for the other states; Medicare/Medicaid status and full benefits status applied only to Maine as these populations were not included in the other states; and beneficiary attribution methods or participation in a specific state program applied only to the given state (Maine and Vermont). We do not include the same number of county‐level variables for Vermont because there was not enough variation by county. Some covariates (such as age) were specified slightly differently in each state based on which form of the variable produced the best balance when applying the propensity score weights.

^bA Vermont Medicaid program that targets members at risk for adverse health outcomes. It provides case management and social support services to improve their health and reduce costs.