Skip to main content
. 2016 Mar 1;193(5):495–503. doi: 10.1164/rccm.201507-1381OC

Table 3.

Patient Characteristics According to Model 1 Propensity Score Quintile

Characteristics Quintile 5 (Highest Risk of Early HFOV) (n = 213) Quintile 4 (n = 213) Quintile 3 (n = 213) Quintiles 1-2 (Lowest Risk of Early HFOV) (n = 425) P Value*
Early HFOV, n (%) 99 (46) 50 (23) 26 (12) 17 (4) <0.001
Age at PICU admission          
 Median (IQR), yr 4.3 (1.2–11.5) 6.7 (2.4–13.1) 3.3 (0.5–10.8) 4.0 (1.0–9.8) 0.07
 No. (%)         0.18
  6.00–17.99 yr 87 (41) 113 (53) 84 (39) 173 (41)  
  2.00–5.99 yr 49 (23) 52 (24) 31 (15) 100 (24)  
  2 wk–1.99 yr 77 (36) 48 (23) 98 (46) 152 (36)  
PRISM III-12 score, median (IQR) 15 (9–22) 12 (6–19) 10 (5–14) 7 (3–11) <0.001
Primary diagnosis, n (%)         <0.001
 Pneumonia 72 (34) 95 (45) 82 (39) 156 (37)  
 Bronchiolitis 18 (8) 18 (8) 32 (15) 81 (19)  
 Acute respiratory failure r/t sepsis 75 (35) 42 (20) 47 (22) 55 (13)  
 Asthma or reactive airway disease 6 (3) 13 (6) 20 (9) 57 (13)  
 Aspiration pneumonia 10 (5) 12 (6) 15 (7) 24 (6)  
 Other 32 (15) 33 (15) 17 (8) 52 (12)  
Any past medical history, n (%) 106 (50) 118 (55) 110 (52) 246 (58) 0.07
Organ dysfunction on Days 0–1          
 Median (IQR), number of organ dysfunctions 3 (2–4) 3 (2–3) 3 (2–3) 2 (1–2) <0.001
 Number (%)          
  Cardiovascular 184 (86) 162 (76) 146 (69) 111 (26) <0.001
  Neurologic 136 (64) 110 (52) 104 (49) 118 (28) <0.001
  Hematologic 68 (32) 56 (26) 41 (19) 61 (14) <0.001
  Renal 16 (8) 22 (10) 13 (6) 14 (3) 0.002
  Hepatic 63 (30) 52 (24) 40 (19) 52 (12) <0.001
Worst OI on Days 0–1, n (%)         <0.001
 8.0–15.9 0 5 (2) 116 (54) 380 (89)  
 16.0–24.9 14 (7) 134 (63) 85 (40) 45 (11)  
 25.0–39.9 100 (47) 74 (35) 12 (6) 0  
 ≥40.0 99 (46) 0 0 0  

Definition of abbreviations: HFOV = high-frequency oscillatory ventilation; IQR = interquartile range; OI = oxygenation index; PICU = pediatric intensive care unit; PRISM III-12 = Pediatric Risk of Mortality III score from first 12 hours in the PICU; r/t = related to.

*

P values for the comparison between groups were calculated using logistic, linear, cumulative logit, and multinomial logistic regression accounting for PICU as a cluster variable using generalized estimating equations for binary, log-transformed continuous, ordinal, and nominal variables, respectively.