éthique / anesthésie-réanimation

Analysis code: https://github.com/damiacer/analysis/blob/main/VT/lme4models.R

Model 1 (1 level: patient)

Mixed effects logistic regression: outcome log odds modeled as a linear combination of the predictor variables
Data is clustered: IDTri_or (or random effect), as repondent-level variable
Model is based on Laplace approximation of likelihood
The optimizer argument is specified to avoid a warning nonconvergence

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
 Family: binomial  ( logit )
Formula: Answer01 ~ Time + Sexe + Age + Guidelines + IHScore + IBScore +  
    fa_moral + fa_eth + re2 + (1 | IDTri_or)

Output

   Data: vt
      AIC       BIC    logLik  deviance  df.resid 
 8619.444  8731.779 -4294.722  8589.444     13198 
Random effects:
 Groups   Name        Std.Dev.
 IDTri_or (Intercept) 0.7585  
Number of obs: 13213, groups:  IDTri_or, 1235
Fixed Effects:
(Intercept)         Time        Sexe2          Age   Guidelines      IHScore      IBScore    fa_moral1  
   2.460966    -0.000478    -0.191956    -0.006537     0.252791     0.024055    -0.003681    -0.094921  
  fa_moral2    fa_moral3      fa_eth1      fa_eth2      fa_eth3       re2Yes  
  -0.161025    -0.583345     0.090643     0.101774    -0.093397    -0.158502  
optimizer (bobyqa) convergence code: 0 (OK) ; 0 optimizer warnings; 3 lme4 warnings

OR and 95%CI


                   Est        LL         UL
(Intercept) 11.7161185 7.1575826 19.1779039
Time         0.9995221 0.9992598  0.9997845
Sexe2        0.8253436 0.7092927  0.9603821
Age          0.9934843 0.9865972  1.0004194
Guidelines   1.2876139 1.1089837  1.4950171
IHScore      1.0243469 1.0095637  1.0393466
IBScore      0.9963258 0.9828520  1.0099844
fa_moral1    0.9094444 0.6871448  1.2036606
fa_moral2    0.8512704 0.6449227  1.1236407
fa_moral3    0.5580287 0.3491992  0.8917432
fa_eth1      1.0948776 0.8523786  1.4063667
fa_eth2      1.1071333 0.8655017  1.4162238
fa_eth3      0.9108315 0.5218972  1.5896118
re2Yes       0.8534212 0.7335552  0.9928739

Model 1.1 (exclusion des temps < ou > 0.01 (quantiles))

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
 Family: binomial  ( logit )
Formula: Answer01 ~ Time + Sexe + Age + Guidelines + IHScore + IBScore +  
    fa_moral + fa_eth + re2 + (1 | IDTri_or)

Output

   Data: vt
      AIC       BIC    logLik  deviance  df.resid 
 8201.072  8313.090 -4085.536  8171.072     12922 
Random effects:
 Groups   Name        Std.Dev.
 IDTri_or (Intercept) 0.8037  
Number of obs: 12937, groups:  IDTri_or, 1234
Fixed Effects:
(Intercept)         Time        Sexe2          Age   Guidelines      IHScore      IBScore    fa_moral1  
   2.965611    -0.013206    -0.197301    -0.006142     0.262313     0.022282    -0.004199    -0.100377  
  fa_moral2    fa_moral3      fa_eth1      fa_eth2      fa_eth3       re2Yes  
  -0.179350    -0.628093     0.105370     0.080349     0.018866    -0.132688  
optimizer (bobyqa) convergence code: 0 (OK) ; 0 optimizer warnings; 1 lme4 warnings 


                   Est         LL         UL
(Intercept) 19.4065480 11.4775551 32.8130949
Time         0.9868811  0.9846441  0.9891233
Sexe2        0.8209432  0.7009238  0.9615135
Age          0.9938765  0.9866545  1.0011514
Guidelines   1.2999328  1.1126069  1.5187981
IHScore      1.0225320  1.0071551  1.0381438
IBScore      0.9958093  0.9817388  1.0100816
fa_moral1    0.9044968  0.6762637  1.2097566
fa_moral2    0.8358134  0.6268233  1.1144832
fa_moral3    0.5336082  0.3269499  0.8708907
fa_eth1      1.1111216  0.8564412  1.4415365
fa_eth2      1.0836651  0.8390149  1.3996533
fa_eth3      1.0190452  0.5606912  1.8520947
re2Yes       0.8757382  0.7479021  1.0254249

Model 1.2 (exclusion du quantile > 0.01)

Data

17214 answers
13084 obs

Model

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
 Family: binomial  (logit)
 
Formula: Answer01 ~ Time + Sexe + Age + Guidelines + IHScore + IBScore +      re2 + (1 | IDTri_or)

Output

   Data: vt
      AIC       BIC    logLik  deviance  df.resid 
 8362.416  8429.728 -4172.208  8344.416     13075 
Random effects:
 Groups   Name        Std.Dev.
 IDTri_or (Intercept) 0.8304  
Number of obs: 13084, groups:  IDTri_or, 1235
Fixed Effects:
(Intercept)         Time         Sexe          Age   Guidelines      IHScore      IBScore       re2Yes  
   3.082086    -0.012712    -0.182982    -0.007368     0.266074     0.024013    -0.004474    -0.147872  
optimizer (bobyqa) convergence code: 0 (OK) ; 0 optimizer warnings; 1 lme4 warnings

Results


                     Est           LL            UL
(Intercept)  3.082085514  2.571080345  3.5930906828
Time        -0.012712070 -0.014983331 -0.0104408085
Sexe        -0.182981650 -0.341615626 -0.0243476746
Age         -0.007368114 -0.014553974 -0.0001822531
Guidelines   0.266074052  0.109396740  0.4227513641
IHScore      0.024012538  0.008756122  0.0392689533
IBScore     -0.004473799 -0.018719452  0.0097718551
re2Yes      -0.147871542 -0.305928094  0.0101850104


                  Est*         LL         UL
(Intercept) 21.8038272 13.0799477 36.3462372
Time         0.9873684  0.9851284  0.9896135
Sexe         0.8327834  0.7106213  0.9759463
Age          0.9926590  0.9855514  0.9998178
Guidelines   1.3048317  1.1156049  1.5261548
IHScore      1.0243032  1.0087946  1.0400502
IBScore      0.9955362  0.9814547  1.0098198
re2Yes       0.8625419  0.7364396  1.0102371

Est = OR

Model 2 (2 levels: patient and randomization group)

Mixed effects logistic regression: outcome log odds modeled as a linear combination of the predictor variables
Data is clustered: IDTri_or (random effect or random intercept), as repondent-level variable, and Guidelines (random effect or random intecept)
No random spline selected

m7 <- glmer(Answer01 ~ Time +  IHScore + IBScore +  
              fa_moral + fa_eth + re2 + 
              (1|IDTri_or) + (1|Guidelines), data = vt, family = "binomial", 
            control = glmerControl(optimizer = "bobyqa",
                                   optCtrl = list(maxfun = 100000), nAGQ = 0))

Output

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) [
glmerMod]
 Family: binomial  ( logit )
Formula: Answer01 ~ Time + IHScore + IBScore + fa_moral + fa_eth + re2 +  
    (1 | IDTri_or) + (1 | Guidelines)

   Data: vt
      AIC       BIC    logLik  deviance  df.resid 
 8658.152  8755.509 -4316.076  8632.152     13200 
Random effects:
 Groups     Name        Std.Dev.
 IDTri_or   (Intercept) 0.7361  
 Guidelines (Intercept) 1.1863  
Number of obs: 13213, groups:  IDTri_or, 1235; Guidelines, 2
Fixed Effects:
(Intercept)         Time      IHScore      IBScore    fa_moral1    fa_moral2  
  0.1733771   -0.0004707    0.0300095   -0.0048630    0.0465026   -0.0228168  
  fa_moral3      fa_eth1      fa_eth2      fa_eth3       re2Yes  
 -0.3451896    0.0950926    0.0999617   -0.1275309   -0.1855003  
optimizer (bobyqa) convergence code: 0 (OK) ; 0 optimizer warnings; 2 lme4 warnings

OR and 95%CI

                  Est        LL        UL
(Intercept) 1.1893145 0.2200529 6.4278586
Time        0.9995294 0.9992763 0.9997825 *
IHScore     1.0304643 1.0163874 1.0447362 *
IBScore     0.9951488 0.9822436 1.0082236
fa_moral1   1.0476008 0.8023437 1.3678271
fa_moral2   0.9774416 0.7506913 1.2726830
fa_moral3   0.7080861 0.4498424 1.1145812
fa_eth1     1.0997607 0.8638120 1.4001583
fa_eth2     1.1051286 0.8720356 1.4005268
fa_eth3     0.8802662 0.5165240 1.5001600
re2Yes      0.8306886 0.7184983 0.9603968 *

Model 3: nested random effects

m8 <- glmer(Answer01 ~ Time +  IHScore + IBScore +  
              fa_moral + fa_eth + re2 + 
              (1|Guidelines/IDTri_or), data = vt, family = "binomial", 
            control = glmerControl(optimizer = "bobyqa",
                                   optCtrl = list(maxfun = 100000), nAGQ = 0
            ))

Model

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) [
glmerMod]
 Family: binomial  ( logit )
Formula: Answer01 ~ Time + IHScore + IBScore + fa_moral + fa_eth + re2 +  
    (1 | Guidelines/IDTri_or)

Output

   Data: vt
      AIC       BIC    logLik  deviance  df.resid 
 8658.152  8755.509 -4316.076  8632.152     13200 
Random effects:
 Groups              Name        Std.Dev.
 IDTri_or:Guidelines (Intercept) 0.7361  
 Guidelines          (Intercept) 1.1863  
Number of obs: 13213, groups:  IDTri_or:Guidelines, 1235; Guidelines, 2
Fixed Effects:
(Intercept)         Time      IHScore      IBScore    fa_moral1    fa_moral2  
  0.1733771   -0.0004707    0.0300095   -0.0048630    0.0465026   -0.0228168  
  fa_moral3      fa_eth1      fa_eth2      fa_eth3       re2Yes  
 -0.3451896    0.0950926    0.0999617   -0.1275309   -0.1855003  
optimizer (bobyqa) convergence code: 0 (OK) ; 0 optimizer warnings; 2 lme4 warnings

Warnings

In vcov.merMod(m8) :
  variance-covariance matrix computed from finite-difference Hessian is
not positive definite or contains NA values: falling back to var-cov estimated from RX

OR and 95%CI

                  Est        LL        UL
(Intercept) 1.1893145 0.2200529 6.4278586
Time        0.9995294 0.9992763 0.9997825 *
IHScore     1.0304643 1.0163874 1.0447362 *
IBScore     0.9951488 0.9822436 1.0082236
fa_moral1   1.0476008 0.8023437 1.3678271
fa_moral2   0.9774416 0.7506913 1.2726830
fa_moral3   0.7080861 0.4498424 1.1145812
fa_eth1     1.0997607 0.8638120 1.4001583
fa_eth2     1.1051286 0.8720356 1.4005268
fa_eth3     0.8802662 0.5165240 1.5001600
re2Yes      0.8306886 0.7184983 0.9603968 *

éthique / anesthésie-réanimation

damiano.cerasuolo@unicaen.fr

2023-06-08

Model 1 (1 level: patient)

Output

OR and 95%CI

Model 1.1 (exclusion des temps < ou > 0.01 (quantiles))

Output

Model 1.2 (exclusion du quantile > 0.01)

Data

Model

Output

Results

Model 2 (2 levels: patient and randomization group)

Output

OR and 95%CI

Model 3: nested random effects

Model

Output

Warnings

OR and 95%CI