Table 2.

Optimization of regression models for lung ultrasound score

Linear model				Mixed-effects mode
Predictors	Estimates	CI	p	Estimates	CI	p
(Intercept)	285.18	222.06 to 348.30	< 0.001	264.97	179.41 to 350.53	< 0.001
LUSs	− 4.82	− 6.84 to − 2.80	< 0.001	− 3.66	− 6.15 to − 1.17	0.004
Age	− 1.43	− 2.43 to − 0.43	0.006	− 1.39	− 2.61 to − 0.17	0.025
Random effects
σ2					2457.29
τ00					10,042.68 Subject
τ11					13.62 Subject.LUS
ρ01					− 1.00 Subject
N					25 Subject
Observations		95			95
R2/R2 adjusted		0.316/0.301			0.375/NA
AIC		1078.783			1051.101

The optimal models were found to be those including age as a fixed term, but not the hospital admission day. A linear mixed effects model accounting for intersubject variation in intercept and value of the LUS score parameter estimate, as a random effect, was found to be superior in terms of R2 and AIC. AIC Akaike’s information criterion, CI 95% confidence intervals