Table 5.
Study Outcomes in Propensity-Score Matched Sample
| Prolonged antibiotics | No prolonged antibiotics | RR (95% CI)a | p-valuea | |
|---|---|---|---|---|
| Infants, n | 2,181 | 2,181 | ||
| Death/NDI, n (%) - with adjustment for key covariatesb | 1,091 (50.0) | 1,047 (48.0) | 1.04 (0.98-1.11) 1.05 (0.99-1.11) | 0.18 0.09 |
| Death before follow-up, n (%) - with adjustment for key covariatesb | 512 (23.5) | 469 (21.5) | 1.09 (0.98-1.22) 1.07 (0.97-1.19) | 0.12 0.17 |
| Survived to follow-up and evaluated, n | 1,669 | 1,712 | ||
| NDI, n (%) - with adjustment for key covariatesb | 579 (34.7) | 578 (33.8) | 1.03 (0.94-1.13) 1.05 (0.96-1.16) | 0.57 0.25 |
| Bayley 3 cognitive composite score <85 - with adjustment for key covariatesb | 530 (32.0) | 533 (31.4) | 1.02 (0.92-1.13) 1.05 (0.96-1.16) | 0.70 0.29 |
| GMFCS level ≥2 - with adjustment for key covariatesb | 162 (9.7) | 155 (9.1) | 1.07 (0.87-1.33) 1.17 (0.95-1.45) | 0.50 0.14 |
| Bilateral blindness | 27 (1.6) | 24 (1.4) | 1.15 (0.67-1.97) | 0.61 |
| Hearing impairment | 58 (3.5) | 49 (2.9) | 1.22 (0.85-1.76) | 0.29 |
| Blindness or hearing impairment - with adjustment for key covariatesb,c | 79 (4.8) | 71 (4.2) | 1.15 (0.84-1.56) 1.16 (0.85-1.58) | 0.38 0.36 |
Relative risks (RR) and confidence intervals (CI) from log binomial (for primary comparisons) or Poisson regression models (for models with post-match adjustment) using methods that account for the matched pair design with statistical significance determined by the Wald chi-square test. Poisson models were used due to convergence problems with log binomial models when covariates were added.
Covariates used in post-matching adjustment included important confounders (study center, maternal education, infant GA, BW, male sex, severe IVH diagnosed within 7 days of birth), as well as highest respiratory support at 24 hours of life due to some evidence of residual imbalance after matching based on a standardized mean difference between groups >0.10 for one level (High-Frequency ventilation, 0.112).
Study center could not be included in the model fit to the combined outcome blindness or hearing impairment as some centers had no children with either outcome.