Impact of availability of guidelines and active surveillance in reducing the incidence of ventilator-associated pneumonia in Europe and worldwide

The present work is based on a comprehensive internet-based survey of 1730 ICU physicians. The questionnaire was available online from April 1 to September 1, 2012 and was distributed through various channels by the European Society of Intensive Care Medicine, by national ICU societies in Austria, France, Belgium, the Netherlands, Italy, and Greece as well as through the Revista Electronica de Medicina Intensiva, an electronic newsletter on intensive care medicine in Spanish distributed in Spain, Portugal, and Latin America [28]. Accordingly, the underlying response rate of the survey was unknown. Overall, we received replies from 77 different countries worldwide, of which the majority i.e. 1351 replies came from 36 European countries (including Cyprus, Georgia, Turkey and the Russian Federation). Participation in the survey was anonymous and, as no patient-related information but only process indicators were asked, ethical approval was not necessary [29, 30]. See Table 1 for details of the countries with most responses (n > 20). The exact details of the overall survey are already described elsewhere [28, 31, 32]. The present work applies a cross-country comparison of selected survey results in order to analyse factors influencing compliance with a number of well-established VAP prevention measures. Essentially, the survey generated information about (1) country of abode, (2) some aspects about the hospital in which the physician works, (3) a large number of parameters relating to the ICU in which the physician works and (4) a number of person-specific parameters. With respect to the ICU level, two aspects of the survey are of particular interest for this analysis: Firstly, the physicians were asked to answer the following question (yes/no): “In my ICU, we have written clinical guidelines for the prevention of ventilator-associated pneumonia” (in the following referred to as existence of VAP-standard). Secondly, the physicians answered the question (yes/no) “In my ICU, we count and record VAP on a routine basis” (in the following referred to as existence of VAP surveillance system). Thirdly, the physicians were asked whether, care for intubated patients in their ICU included the following measures: (1) Head of bed elevation; (2) Daily sedation and weaning protocol; (3) Oral care with chlorhexidine; (4) No ventilator circuit change unless indicated; (5) Cuff pressure control at least every 24 hrs; (6) Strict hand hygiene using alcohol, especially before managing the airways (in the following referred to as compliance with prevention measure (1) – (6)). In addition to these 6 specific VAP prevention measures, we determined a variable of overall compliance (compliance-score) by simply adding up self-reported compliance with the 6 specific measures. If for instance all 6 measures are applied in the particular ICU, this variable is equal to 6; if however only two measures are applied the variable is equal to 2.

Although there is no universally accepted gold standard for prevention of VAP, a recent study defined a European care bundle for prevention of VAP and ranked VAP prevention measures by combining criteria such as the strength of the supporting evidence, ease of implementation and expected impact on VAP incidence [33]. The top five clinical interventions of this ranking were included in our questionnaire (VAP prevention measure 2–6). As a control, the most commonly recommended clinical practice of head of bed elevation was included in the survey and analyzed as VAP prevention measure 1. Accordingly, self-reported compliance with the target VAP prevention measures were interpreted as compliance with bundle-like VAP prevention measures under routine conditions.

The statistical analysis consisted of three steps. In a first step, country-specific averages of stated compliance with specific VAP prevention measures univariately regressed against the country-average responses to the questions on whether there were written clinical guidelines in place on ICU level for prevention of VAP (and whether there were VAP surveillance systems in place). The results provide a first insight into the relationship between described variables on a cross-country level. Next, within-country averages of stated compliance with the VAP prevention measures were calculated separately for physicians stating whether their ICU had written VAP-standards or not (and whether they had VAP surveillance systems in place or not). Thus, the evidence from these within-country differences in average compliance with the VAP prevention measures was summarized by a p-value from a paired Student’s t-test. In a further step, we endeavoured to utilize the information on inter-ICU differences, i.e. on whether associations exist at ICU-level within each country. Unfortunately, the ICU of each participant is not known. Because our main outcome was defined at ICU level – i.e. compliance with 6 specific measures, we were forced to define pseudo-ICUs by analysing patterns in the ICU characteristics reported by the participants (country of abode, ICU type, number of beds in ICU and number of beds in hospital) and then grouping participants with similar patterns into pseudo-ICUs. This was done in a rather liberal manner to ensure that each existing ICU was covered by a pseudo-ICU, at the same time allowing for pseudo-ICUs to cover several responses. For ICU type and number of beds in hospital only three answers were possible (ICU type: medical, surgical, mixed; number of beds in hospital: < 300, 300–1000, > 1000) while number of beds in an ICU was a continuous variable. The number of ICUs we were able to define this way in the different countries is shown in Table 1. Within each pseudo-ICU we have some variation in the individual responses, which both reflects the flaws in the pseudo-ICUs and the differences in the individual responses within an ICU. We therefore decided to perform an analysis at the individual level, taking however both country and pseudo-ICU levels into account to avoid an overoptimistic assessment of the statistical significance since we had ignored the high correlations between the outcomes within each ICU. Moreover, using pseudo ICUs instead of the true ICUs results in more valid p-values here because possible and to some extent unknown clustering is taken into account. We used linear regression/logistic regression models with the covariates of interest, adjusting for country (as a categorical covariate) as well as number of beds, gender and experience as individual measurements, and took the clustering within pseudo ICUs into account by using robust standard errors (application of the cluster option in Stata 12). Using pseudo ICUs instead of ICUs is a valid approach here because the central assumption of independence of the outcomes between clusters is still valid, even if the clusters are bigger than necessary. As covariates of interest we included presence of written clinical standards on ICU level and presence of VAP surveillance systems.

As shown in Figure 1, country averages for countries where there were more than 20 responses of self-reported compliance with the different VAP prevention measures are positively associated with the country-averages of self-reported presence of written clinical standards for prevention of VAP on the ICU level (p < 0.01).

Furthermore, Figure 2 shows that in most countries where there were more than 20 responses, average self-reported compliance with the different VAP prevention measures is higher when the respondents also confirmed presence of written VAP prevention standards on ICU level (p < 0.01). Correspondingly, we can postulate a connection between compliance with VAP prevention measures and the presence of VAP prevention standards on the between-country (Figure 1) and within-country (Figure 2) level.

As shown in Additional file 1: Figure S1 (available as supplemental material), the results displayed by individual prevention measures provide mixed results. VAP prevention measure 1 - head of bed elevation - is commonly recommended clinical practice with high rates of compliance; however, when looking at the between-country or within-country averages there does not seem to be an obvious connection with the presence of VAP prevention standards. Regarding VAP prevention measure 2 and 3, which refer to daily sedation vacation and weaning protocol and oral care with chlorhexidine, there seems to be a connection between the respective prevention measure and the existence of VAP prevention standards when regarding the between- country and within- country averages. Regarding VAP prevention measure 4 and 5, which refer to no ventilator circuit change unless indicated and cuff pressure control at least every 24 hrs, the country-averages show no clear connection. VAP prevention measure 6, which represents strict hand hygiene using alcohol, especially before managing the airways, shows quite a strong association when looking at the within-country averages, but absolutely no association when looking at the between-country averages.

As shown in Figure 3, country averages of self-reported compliance with the different VAP prevention measures are positively associated with the country averages of self-reported existence of a VAP surveillance system at the 10 percent level only (p = 0.08).

On the other hand, the connection between prevention measure compliance and existence of a surveillance system appears to be more obvious on the within-country level (See Figure 4).

The regression analyses applied take both the interrelation within and between countries into account and show a significant independent association between most of the VAP prevention measures and the existence of written VAP prevention standards on ICU level (see regression (1) to (6) in Table 2). Please note that the non-significance with respect to VAP prevention measure 1 may be a result of the ceiling effect since compliance with the commonly recommended clinical practice of head of bed elevation is very high (see Table 1). The existence of a VAP surveillance system, however, has a slightly smaller impact on compliance with the specific measures. Column (7) in Table 2 shows the results of the analysis in which the variable of total compliance (compliance-score) is placed on the left-hand side of the regression analysis. In detail, the analyses show that existence of standard and VAP surveillance systems impact substantially on total compliance with the VAP prevention measures (both p < 0.001). The country-specific implications of regression (7) may be interpreted as follows: According to our model, a female physician with 10 years of experience in critical care working in a 15-bed ICU in France for instance has a predicted baseline level of VAP prevention measure compliance (compliance-score) of 4.309-0.0329 + (10*0.00594) + (15*0.00316)-0.598 = 3.7849, which equals 63%. This baseline level of compliance with VAP prevention measures increases by 0.573 (p < 0.001) if a written clinical VAP standard is available on ICU level, and by 0.244 (p < 0.001) if a VAP surveillance system is in place, which is equivalent to a 15% and 6% increase in the level of compliance with VAP prevention measures, respectively.