Skip to main content
. 2020 Nov 14;230:23–31.e10. doi: 10.1016/j.jpeds.2020.11.016

Table XI.

Sensitivity analysis: logistic regression models for severe outcome using multiple imputation for missing data

Case description Multivariable model 1
Multivariable model 2
aOR (95% CI) P value aOR (95% CI) P value
Respiratory illness: 56 events, N = 141
 Age (per 1-year decrease) 1.09 (1.02-1.16) .01 1.09 (1.02-1.15) .01
 Obesity (reference = no) 3.48 (1.22- 9.96) .02 2.84 (1.09-7.39) .03
 Admission WBC count (per unit increase in 109/L) 1.11 (1.02-1.21) .01 1.10 (1.02-1.19) .02
 Hispanic (reference = non-Hispanic white/other) 0.93 (0.30-2.90) .90
 Non-Hispanic black (reference = non-Hispanic white/other) 1.64 (0.34-7.90) .54
 Oxygen saturation <90% (reference = no) 3.56 (0.98-12.95) .05 4.08 (1.17-14.27) .03
 Medical complexity (reference = no) 2.34 (0.94-5.81) .07
 Bilateral infiltrates on radiograph (reference = no)
4.05 (1.59-10.30)
.003
3.88 (1.59-9.48)
.003
Multivariable model 1


aOR (95% CI)
P value


MIS-C: 38 events, N = 68
 Age (per 1-year decrease) 0.98 (0.84, 1.16) .85
 Admission absolute lymphocyte count (per unit decrease in U/L) 6.25 (1.41, 25.0) .02
 Admission CRP level (per unit increase in mg/dL) 1.03 (0.97, 1.09) .33
 Hispanic (reference = non-Hispanic white/other) 1.20 (0.14, 10.09) .87
 Non-Hispanic black (reference = non-Hispanic white/other) 2.86 (0.24, 33.71) .40
 Medicaid/Medicare (reference = private/other) 0.40 (0.09, 1.68) .21

P values for aOR estimated using logistic regression. All estimates are adjusted for site with fixed effects.

For respiratory illness, excluded 2 respiratory cases (one who was chronically ventilator-dependent and discharged on hospital day 1, and one who died on the day of admission). Multiple imputation model included outcome, all predictor variables listed in Table X and hospital site. Results based on 40 imputed data sets.

For MIS-C, excluded 1 MIS-C case (transferred on the day of admission). Multiple imputation model included outcome, all predictor variables listed in Table X and hospital site. Results based on 40 imputed data sets. Model 1 includes variables with P < .25 in bivariate analysis, age and race. Model 2 had nearly complete data (2 missing values) so multiple imputation not performed.

See definitions in Table II.