Table 3.
| Parameter | IRR (95% CI) c | P Value |
|---|---|---|
| COB rate per 100 admissions | ||
| 1st quartile | Reference | |
| 2nd quartile | 1.25 (1.17–1.33) | <.0001 |
| 3rd quartile | 1.44 (1.35–1.54) | <.0001 |
| 4th quartile | 1.52 (1.41–1.63) | <.0001 |
| HO test intensity | ||
| 1st quartile | Reference | |
| 2nd quartile | 1.58 (1.41–1.76) | <.0001 |
| 3rd quartile | 1.93 (1.72–2.16) | <.0001 |
| 4th quartile | 2.39 (2.12–2.69) | <.0001 |
| CO test intensity | ||
| 1st quartile | Reference | |
| 2nd quartile | 0.83 (0.79–0.87) | <.0001 |
| 3rd quartile | 0.76 (0.72–0.81) | <.0001 |
| 4th quartile | 0.74 (0.70–0.79) | <.0001 |
| HO test prevalence | ||
| <2nd quartile | Reference | |
| 3rd quartile | 1.25 (1.18–1.33) | <.0001 |
| 4th quartile | 1.39 (1.30–1.48) | <.0001 |
| ICU admissions, % | ||
| ≤3rd quartile | Reference | |
| 4th quartile | 1.11 (1.05–1.16) | <.0001 |
| Not reported | 1.61 (1.48–1.74) | <.0001 |
| Mean LOS | ||
| ≤3rd quartile | Reference | |
| 4th quartile | 1.15 (1.09–1.21) | <.0001 |
| Bed size | ||
| 01–100 beds | Reference | |
| 101–200 beds | 1.25 (1.12–1.39) | <.0001 |
| 201–300 beds | 1.49 (1.34–1.65) | <.0001 |
| 301–500 beds | 1.43 (1.29–1.59) | <.0001 |
| ≥500 beds | 1.42 (1.27–1.59) | <.0001 |
| Patients aged 41–64 y, % | ||
| 1st quartile | Reference | |
| 2nd quartile | 1.12 (1.04–1.20) | 0.0015 |
| 3rd quartile | 1.15 (1.07–1.24) | <.0001 |
| 4th quartile | 1.38 (1.28–1.49) | <.0001 |
| Patients aged >80 y, % | ||
| 1st quartile | Reference | |
| 2nd quartile | 0.84 (0.79–0.89) | <.0001 |
| 3rd quartile | 0.79 (0.74–0.84) | <.0001 |
| 4th quartile | 0.83 (0.77–0.90) | <.0001 |
Note. AIC, Akaike information criteria; BIC, Bayesian information criteria; CI, confidence interval; CO, community onset; COB, community-onset bacteremia; HO, hospital onset; HOB, hospital-onset bacteremia; ICU, intensive care unit; IRR, incidence rate ratio; LOS, length of stay.
Goodness-of-fit statistics: AIC = 12,882, BIC = 13,042. For the complex model we calculated a 4% reduction in AIC and a 3.4% reduction in BIC compared with the simple model, which indicates an improvement in model fit.
For model replication purposes, regression coefficients and standard errors are presented in Supplementary Table S2 (online).
Estimated increase in HOB relative to the reference. For example, compared to those hospitals in the first quartile of HO test intensity, hospitals in the second quartile were 1.58 times higher (58% higher) in HOB rate per 100 admissions.