# Table 3 Random effect negative binomial regression on days in hospital

IRR SE p 95% CI
Race/ethnicity
Non-Latino Whitea
Black 0.862 0.043 .003 0.782, 0.951
Latino 0.913 0.039 .032 0.840, 0.992
Age 1.004 0.002 .038 1.000, 1.007
Male 0.904 0.034 .007 0.841, 0.972
Education 1.000 0.007 .968 0.987, 1.013
Homeless 1.400 0.055 < .001 1.297, 1.512
History of mental health issues 0.623 0.023 < .001 0.579, 0.671
Days of mental health counseling 1.022 0.004 < .001 1.013, 1.031
Days of psychiatric care 1.064 0.004 < .001 1.056, 1.071
Days of physical problems 1.074 0.002 < .001 1.071, 1.077
Age at first drug use 1.003 0.002 .258 0.998, 1.007
Days of primary drug use 0.994 0.002 < .001 0.991, 0.997
Primary drug problem
Alcohola
Cocaine 1.695 0.099 < .001 1.512, 1.901
Heroin 0.791 0.052 < .001 0.695, 0.901
Marijuana 0.929 0.064 .281 0.812, 1.062
Methamphetamine 0.831 0.078 .047 0.692, 0.998
Other 1.222 0.100 .014 1.041, 1.434
Children younger than 18 1.008 0.006 .143 0.997, 1.020
Program modality
Outpatienta
Methadone 1.034 0.130 .789 0.809, 1.323
Residential 1.821 0.092 < .001 1.650, 2.010
1. Note: IRR, incidence rate ratio. IRRs can be interpreted as the estimated rate ratio for a 1-unit increase in the independent variable, given the other variables are held constant in the model. For example, if days of mental health counseling increased by 1 point, the ratio for number of ER visits would be expected to increase by a factor of IRR = 1.022, while holding all other variables in the model constant.
2. Wald chi-square with 20 degrees of freedom = 5313.21. The corresponding p-value is less than 0.0001.
3. aReference category.