Skip to main content
. Author manuscript; available in PMC: 2017 Jul 12.
Published in final edited form as: JACC Heart Fail. 2016 Mar 9;4(7):551–558. doi: 10.1016/j.jchf.2016.01.003

Table 4.

Incident Rate Ratios from Generalized Linear Model Regressions of Costs in VA Patients with Heart Failure in FY2010, N=117,882

Model (1) Patient factors only Model (2) Facility only Model (3) Patient and facility factors
R-squared 0.116 0.018 0.132
Root Mean Squared Error 46245 48739 45824
Age in years 0.991 0.990
Gender
 Female Ref Ref
 Male 0.942 0.941
Race/ethnicity
 White Ref Ref
 Black 1.015 1.048
 Hispanic 1.135 1.032
 Other 0.975 0.995
 Unknown 0.761 0.769
Marital status
 Single Ref Ref
 Married 0.801 0.821
 Separated/divorced/widowed 0.933 0.945
Service connected percent
 0 or none Ref Ref
 1–49 0.930 0.933
 50+ 1.163 1.172
Insurance coverage
 None Ref Ref
 Major medical/private/CHAMPUS 0.751 0.755
 Medicare 0.902 0.915
 All other 0.852 0.853
Eligibility for VA
 Above means test Ref Ref
 Below means test, not service connected 1.542 1.524
 Service connected 1.615 1.594
 Other eligibility 1.376 1.357
Chronic conditions
 Diabetes 1.115 1.117
 Hypertension 1.344 1.338
 Ischemic heart disease 1.393 1.385
 Chronic renal failure 1.901 1.868
 Alzheimer’s disease 1.377 1.363
 Depression 1.207 1.207
 Post-traumatic stress disorder 1.030 1.031
 Alcohol use disorder 1.280 1.282
 Drug use disorder 1.266 1.242
 Other psychiatric conditions 1.404 1.390

All parameters in the table were statistically significant at p<0.001

Incident rate ratios were obtained from generalized linear regression models with a Poisson family distribution and log-link function. IRRs represent percent change in costs associated with a one-unit change in continuous variable or the percent change in costs relative to the reference category for categorical variables. Coefficients from facility dummies in models 2 and 3 are not shown in table.