Table 3. Risk Factors for NEC and Mortality Using Multivariable Competing-Risks Cox Regression Modela.
| Risk Factors | NEC | Mortality | ||||
|---|---|---|---|---|---|---|
| Cause-Specific HR (95% CI)b | P Value | % Reliabilityc | Cause-Specific HR (95% CI)b | P Value | % Reliabilityc | |
| Model 1—Primary Analysis (N = 598)d | ||||||
| Birth weight, per 100-g increase | 0.72 (0.62-0.84) | <.001 | 98 | 0.63 (0.48-0.82) | <.001 | 99 |
| Received RBC transfusion in a given weeke | 0.44 (0.17-1.12) | .09 | 45 | 1.36 (0.27-6.82) | .71 | 19 |
| Severe anemia in a given week (hemoglobin ≤8 g/dL)e | 5.99 (2.00-18.0) | .001 | 70 | 1.66 (0.40-6.85) | .48 | 25 |
| Days of breast milk feeding in first 10 days of life, per 1-day increase | 1.10 (1.01-1.21) | .04 | 37 | 0.87 (0.77-0.98) | .02 | 47 |
| SNAP on day of birth, per 1-point increase | 1.00 (0.93-1.07) | .99 | 8 | 1.12 (1.03-1.23) | .01 | 59 |
| Days of antibiotic treatment in first 10 days of life, per 1-day increase | 1.04 (0.93-1.16) | .50 | 8 | 1.01 (0.83-1.21) | .96 | 9 |
| Model 2—Study Exposures + Confounders With Reliability ≥50% (N = 598)d | ||||||
| Birth weight, per 100-g increase | 0.72 (0.62-0.84) | 99 | 0.57 (0.44-0.75) | 99 | ||
| Received RBC transfusion in a given week | 0.43 (0.18-1.04) | 47 | 2.02 (0.45-8.97) | 25 | ||
| Severe anemia in a given week | 5.49 (1.81-16.6) | 69 | 2.25 (0.52-9.79) | 35 | ||
| Model 3—Model 2 + Adjustment for Early Respiratory Illness Severity (N = 598)d | ||||||
| Birth weight, per 100-g increase | 0.72 (0.61-0.83) | 99 | 0.57 (0.43-0.75) | 100 | ||
| Received RBC transfusion in a given week | 0.45 (0.18-1.14) | 44 | 1.72 (0.30-9.72) | 26 | ||
| Severe anemia in a given week | 5.48 (1.78-16.8) | 68 | 2.36 (0.53-10.4) | 35 | ||
| Received surfactant therapy in first week of life | 1.95 (0.86-4.30) | 11 | 0.46 (0.16-1.38) | 14 | ||
| Mechanical ventilation in first week of life | 0.45 (0.22-0.94) | 28 | 3.23 (0.58-18.0) | 14 | ||
| Model 4—Propensity Score Adjusted With Propensity Score Specified as a Continuous Covariate (n = 596)f | ||||||
| Received RBC transfusion in a given week | 0.57 (0.23-1.40) | 25 | 1.92 (0.40-9.26) | 49 | ||
| Severe anemia in a given week | 4.52 (1.51-13.6) | 64 | 2.95 (0.77-11.2) | 40 | ||
| Model 5—Propensity Score Adjusted With Inverse Probability of Treatment Weighting (n = 596)f | ||||||
| Birth weight, per 100-g increase | 0.69 (0.62-0.77) | 100 | 0.55 (0.43-0.72) | 100 | ||
| Received RBC transfusion in a given week | 0.37 (0.11-1.16) | 60 | 3.35 (0.74-15.1) | 45 | ||
| Severe anemia in a given week | 3.42 (1.05-11.1) | 52 | 1.78 (0.41-7.78) | 23 | ||
Abbreviations: HR, hazard ratio; NEC, necrotizing enterocolitis; RBC, red blood cell; SNAP, Score for Neonatal Acute Physiology.
Competing risks were 44 infants with NEC and 32 deaths total; 13 infants with NEC died. All models include adjustment for center (not shown).
To estimate cause-specific HR for NEC, 44 events were used, and 19 deaths were used to estimate cause-specific HR for mortality.
Percentage of time risk factor appears in 1000 bootstrap models. Risk factors appearing in at least 50% of models are reliable.
Models 1, 2, and 3 include center as a covariate.
All models specify RBC transfusion and severe anemia as time-dependent covariates evaluated in 1-week intervals.
The following covariates were used to model the propensity score: mechanical ventilation on the day of birth, receipt of at least 1 dose of antenatal steroids, birth weight, gestational age, SNAP, 5-minute Apgar score, hemoglobin level at birth, and clinical center. Models exclude 2 infants due to missing data on 5-minute Apgar score (additional information on propensity score modeling reported in eMethods, eTables 11-14, and eFigures 3-4 in the Supplement). Models 4 and 5 include center in the propensity score model estimating probability of RBC exposure.