TABLE 4.
Propensity-matched cohort: secondary outcomes with tigecyclinea
| Univariate analyses |
Multivariable analyses |
|||||
|---|---|---|---|---|---|---|
| Outcome | No tigecycline (N = 140) | Tigecycline (N = 28) | P value | Odds ratio or estimate (LOS) | 95% CI | P value |
| Colectomy/diverting ileostomy due to CDI | 4 (2.9%) | 3 (10.7%) | 0.167 | 4.45b | 0.59–32.4 | 0.128 |
| Hospital mortality attributable to CDI | 21 (15.0%) | 5 (17.9%) | 0.924 | 1.42 | 0.39–5.14 | 0.594 |
| Subsequent recurrence | 30 (21.4%) | 7 (25.0%) | 0.868 | 1.19 | 0.4 −3.28 | 0.742 |
| Hospital length of stay | ||||||
| Mean days (SD) | 23.1 (27.2) | 37.1 (46.3) | 0.131 | 0.71 | 0.39–1.03 | <0.001 |
| Hospital length of stay after CDI | ||||||
| Mean days (SD) | 15.2 (20.5) | 27.8 (43.5) | 0.146 | 1.08 | 0.69–1.47 | <0.001 |
n (%) unless otherwise specified. P values for univariate analyses calculated using chi-square tests or independent-samples t tests (length of stay). P values in bold-faced type are considered to be significant. For results of multivariable logistic regression (colectomy/ileostomy, attributable mortality, recurrence) and linear regression (total length of stay, length of stay after CDI) with generalized estimating equation method (to adjust for within-subject correlation), all models were adjusted for ATLAS Score, hypotension, time period, and serum lactate. Hospital length of stay calculated between admission/discharge and length of stay after CDI calculated between CDI diagnosis and discharge. Length of stay regression coefficients represent the estimated differences (in days) for the tigecycline group compared with nontigecycline (coefficients significantly greater than zero interpreted as longer length of stay in tigecycline group). LOS, length of stay; CI, confidence interval; SD, standard deviation.
Generalized estimating equation method could not be applied to colectomy/diverting ileostomy due to low event numbers so ordinary multivariable logistic regression was used.