Skip to main content

Table 2 Development of GSEM models; model C, model B and model A

From: Structural equation modelling the relationship between anti-fungal prophylaxis and Pseudomonas bacteremia in ICU patients

  Model C Model B Model A
  Additional file 1: Fig S7 Additional file 1: Fig S8 Fig. 4
     95%CI
Factora−i     
b_Ps_n     
 Pseudomonas colonization 1.25*** 1.23*** 1.23*** 0.86 to 1.59
 ppap 0.76* 0.74* 0.73* 0.1 to 1.36
 _cons − 5.92*** − 5.79*** − 5.83*** − 6.37 to − 5.28
v_Ps_n     
 Pseudomonas colonization 1 1 1 (constrained)
 mvp90 0.35 0.25 0.28 − 0.11 to 0.67
 non_D − 0.54*** − 0.52*** − 0.46*** − 0.71 to − 0.22
 _cons − 4.48*** − 4.22*** − 4.31*** − 5.01 to − 3.57
Pseudomonas colonization     
cc    0.37* 0.06 to 0.68
tap − 0.64*** − 0.49*** − 0.44*** − 0.68 to − 0.21
Anti-septic (a_S) − 0.86*** − 0.44 − 0.44 − 0.94 to 0.06
los7 0.82*** 0.69*** 0.69*** 0.38 to 0.99
Trauma (trauma50) − 0.05 − 0.01 − 0.03 − 0.4 to 0.34
crf 0.20 − 0.29 − 0.38 − 0.95 to 0.19
Candida colonization   0.39*** 0.37*** 0.26 to 0.49
b_can_n     
 Candida colonization 0.73*** 0.7*** 0.7*** 0.33 to 1.07
 _cons − 5.01*** − 4.95*** − 4.97*** − 5.34 to − 4.6
v_can_n     
 Candida colonization 1 1 1 (constrained)
 mvp90 − 0.77 − 0.56 − 0.54 − 1.4 to 0.31
non_D − 0.22 − 0.31 − 0.25 − 0.84 to 0.33
_cons − 3.65*** − 4.0*** − 4.05*** − 5.77 to − 2.36
Candida colonization     
cc    0.45 − 0.19 to 1.09
tap 0.96* 0.96* 1.02* 0.11 to 1.93
Anti-septic (a_S) − 1.28** − 1.27** − 1.23** − 2.13 to − 0.32
los7 0.13 0.13 0.1 − 0.45 to 0.65
Trauma (trauma50) 0.11 0.02 0.03 − 0.87 to 0.82
crf 1.43** 1.51** 1.48** 0.45 to 2.5
Amphotericin − 1.73** − 1.77*** − 1.78*** − 2.79 to − 0.78
Nystatin − 0.90 − 1.05 − 1.04 − 2.33 to 0.3
Azoles and other − 1.44** − 1.53** − 1.47** − 2.56 to − 0.38
Error terms     
var (e. Pseudomonas col) 0.52*** 0.33*** 0.32*** 0.23 to 0.44
var (e. Candida col) 1.37*** 1.3*** 1.27*** 0.82 to 1.97
Model fiti     
AIC 3974 3928 3921
Groups (n) 464 464 464
Clusters (n) 279 279 279  
Factors (N) 29 30 32
  1. Shown in this table are models derived with all studies increasing in complexity from left to right. The figures corresponding to models C (Additional file 1: Fig. S7), model B (Additional file 1: Fig. S8) models A (Fig. 4)
  2. *p < 0.05; **p < 0.01; ***p < 0.001
  3. av_ps_n is the count of Pseudomonas VAP; v_can_n is the count of RT Candida; b_ps_n is the count of Pseudomonas bacteremia and b_can_n is the count of Candidemia
  4. bppap is the group wide use of protocolized parenteral antibiotic prophylaxis; tap is topical antibiotic prophylaxis; non-D is a non-decontamination intervention
  5. cmvp90 is use of mechanical ventilation by more than 90% of the group
  6. dcrf is group wide exposure to a candidemia risk factor
  7. eLOS7 is a mean or median length of ICU stay for the group of more than 7 days
  8. fTrauma ICU arbitrarily defined as an ICU for which > 50% of admissions were for trauma
  9. gPseudomonas colonization (Pseudomonas col) is a latent variable
  10. hCandida colonization (Candida col) is a latent variable
  11. iModel fit; AIC is Akaike’s information criteria. This indicates model fit taking into account the statistical goodness of fit and the number of parameters in the model. Lower values of AIC indicate a better model fit. Groups (n) is the number of patient groups; clusters (n) is the number of studies; factors (N) is the number of parameters in the model