Definitions HCWs were grouped according to the change in their HH compliance between baseline and intervention. ‘Improving HCW’ were defined by having improved compliance by at least 20%, ‘Worsening HCW’ decreased by at least 20%, and ‘Non‐changing HCW’ changed less than 20%, if at all. The 20%‐threshold was chosen retrospectively based on the rounded pooled mean change among all HCWs. An ‚activity index‘, defined by number of hand hygiene opportunities per hour of observation, was defined as a proxy for the intensity of care [25]. Analysis plan and statistics To meet our study scope, we chose four analytical models. In Model 1, we evaluated the extent to which changes in the individual HH compliance between study periods were associated with ICU characteristics. We calculated Spearman rank correlation coefficients of the proportion of Improving HCWs with a) the nurse‐to‐patient ratio and b) the pooled mean baseline HH compliance of all HCWs. In Model 2, we assessed a potential association between the intervention effect on the individual HCW and the overall ICU. We calculated the Spearman’ rank correlation coefficients for the proportion and median improvement of Improving HCWs with the pooled change in HH compliance of all HCWs. In Model 3, we tested the potential association of the change in HH compliance for each individual HCW (measured as change in percentage points (pp); outcome variable) with HCW characteristics (i.e., professional category, baseline compliance) and contextual factors (i.e., activity index, ICU type, ICU nurse‐ to‐patient ratio, and proportion of improving HCWs). We used a generalized linear mixed model (GLMM) with a normal distribution, allowing for clustering at the ICU level. Variables with a P‐value <0.25 in the univariable analysis were included in the multivariable model using manual backward selection. The proportion of explained variation (R2) was calculated for this model. All changes in HH compliance were calculated as relative proportions (%) or differences in percentage points (pp), using mid‐P exact tests to test for significance. We used SAS 9.4 (SAS Institute Inc., Cary, United States) for all statistical analyses. In Model 4, we created a swarm plot with individual HH compliance at baseline and intervention for each HCW, in each category (improving, non‐improving, worsening HCWs), and a bar diagram of the range in HH between HH sessions for each HCW, during the baseline and intervention phase, for visual display of individual HCW compliance patterns in the seven study ICUs.
RkJQdWJsaXNoZXIy MTk4NDMw