Fig. 4

GSEM of the interaction model in relation to Pseudomonas and Candida infection data. Candida col and Pseudomonas col (ovals) are latent variables representing Candida and Pseudomonas colonization, respectively. The variables in rectangles are binary predictor variables representing the group level exposure to the following; patient selection for candidemia risk factors (CRF); trauma ICU setting (trauma50), mean or median length of ICU stay ≥ 7 days (los7), exposure to a topical anti-septic (a_S), exposure to TAP (tap), concurrency of a control group with an antibiotic-based intervention group (CC), exposure to a non-decontamination based prevention method (non-D), use of mechanical ventilation for more than 90% (mvp90) or exposure to PPAP (ppap). Note that the model factorizes exposures from compound regimens (e.g. SDD and SOD, which combine TAP, an antifungal together with or without PPAP) into singleton TAP, PPAP and anti-fungal exposures. The circles contain error terms. The three part boxes represent the binomial data for Candida and Pseudomonas VAP (v_can_n, v_ps_n) and candidemia (b_can_n) or bacteremia (b_ps_n) counts with the number of patients as the denominator which is logit transformed using the logit link function in the generalized model