Table 3.
Zero‐inflated negative binomial (ZINB) regression model a
| Parameter | Logit | Negative Binomial | ||
|---|---|---|---|---|
| Estimate | P * | IRR | 95% CI | |
| Intercept | −0.87 | .0007 | 3.38 | 2.91, 3.92 |
| Dual care (VA and non‐VA health care) in prior 12 mo | −0.7556 | <.0001 | 1.15 | 1.07, 1.25 |
| Female Sex | 0.18 | .5040 | 1.06 | 0.93, 1.20 |
| <65 y of age | −0.54 | .0018 | 1.18 | 1.09, 1.28 |
| Black or African American | 0.24 | .3008 | 1.00 | 0.88, 1.13 |
| Hispanic | 0.32 | .1494 | 1.17 | 1.04, 1.31 |
| Asian or Asian American | −0.19 | .6621 | 1.26 | 1.06, 1.49 |
| Other minority | −1.96 | .3193 | 1.27 | 0.98, 1.65 |
| Multirace/ethnicity | −0.14 | .7139 | 1.10 | 0.93, 1.28 |
| Unknown race/ethnicity | 0.15 | .6574 | 0.99 | 0.83, 1.17 |
| Not a high school graduate or with GED equivalent | 0.72 | .0321 | 1.08 | 0.85, 1.37 |
| Not married, in a civil union or domestic partnership, or living with a partner | 0.00 | .9845 | 1.00 | 0.92, 1.08 |
| Self‐rated physical health fair or poor | −0.64 | .0002 | 1.27 | 1.17, 1.38 |
| Self‐rated mental health fair or poor | −0.57 | .0034 | 1.25 | 1.16, 1.36 |
Abbreviations: CI, confidence interval; IRR, incidence rate ratios.
a
The ZINB model (n = 2353) also adjusted for clustering of patients within clinics; parameter estimates and IRRs for clinics are not displayed. Parameter estimates generated by the logit component predict the likelihood of being an excess zero (ie, of always reporting no hassles). The count component includes the full range of values for the outcome (including zero) assuming a negative binomial distribution and yields parameter estimates expressed in natural log units. Incidence rate ratios (IRRs) are the exponentiated parameter estimates.
*
Values significant at P < .05 are bolded.