As a growing problem in patient care, hospital-acquired infections require particular attention in intensive care patients since they are significantly associated with worse outcomes [1] and higher costs [2–4]. Pneumonia (LRTI) is the most prevalent infection among ICU-patients [5–7]. The incidence of VAP ranges between 9.3 and 13.6 cases per 1,000 ventilation days. Early adequate antimicrobial therapy is an essential determinant of clinical outcome. Choosing the appropriate agent, however, remains challenging since in most cases no data on the identity and susceptibility of the pathogen is available at the time of treatment initiation [8–12].

Several guidelines on the initial therapy of HAP and VAP were developed in the last years. Recently, updated versions of the recommendations issued by the ATS/IDSA (USA) and PEG (Germany) were published [13–15]. These guidelines acknowledge the risk of difficult-to-treat pathogens as causative agents of nosocomial pneumonia. The recommended regimens include monotherapy with acylaminopenicillines + beta-lactamase inhibitors, 3rd generation cephalosporins, quinolones, carbapenemes and various combination regimens.

So far only a few published studies investigated the impact of clinician adherence to therapeutic guidelines on clinical outcomes or at least the usefulness of guidelines in clinical care [16–18]. We investigated the effects of guideline adherence on clinical and economical outcomes of patients with HAP/VAP in the ICU.

German hospitals collect and submit vast amounts of administrative, clinical data. A subset of approximately 250 hospitals additionally provides cost data. These data are primarily used for the annual evaluation and cost calculation within the German DRG-System (GDRG). These datasets are frequently used in clinical-economical analyses as they include all primary and secondary diagnoses (ICD-10 encoded) [19–24].

We investigated HAP/VAP outcomes and therefore focussed on ICD codes representing pneumonia as a secondary diagnosis. In our study we collected the above-specified data from 5 hospitals (2 university hospitals and 3 major urban tertiary hospitals). These hospitals were selected because we assumed that guidelines adherence was highest in tertiary care and university hospitals. Moreover the selected hospitals participate in the German DRG annual evaluation program. The German Institute for the Hospitals Remuneration System (InEK) that performs these evaluations uses high quality standards regarding cost data.

Together with clinical advisors we selected the patient population of interest. As HAP/VAP are specifically relevant on ICU [5, 9, 25], we used the recommendations of the PEG guidelines for adults to determine whether a given initial intravenous antibiotic therapy (IIAT) was adherent to guidelines. Thus we only included patients who where at least 18 years old at the time of treatment initiation.

Identifying patients for clinical economic analyses from DRG-related data has certain limitations and therefore needs special attention. The diagnoses are encoded in ICD-10 GM (German modification). A patient has a primary diagnosis (PDX) which is the retrospectively identified cause for hospitalization. Each dataset also contains a number of secondary diagnoses (SDX). A SDX must be documented if the disease caused significant resource consumption during the hospital stay. Some SDXs, the so-called complications and comorbidities (CC), are cost-relevant within the DRG system. In the German DRG dataset an SDX has no date and no flag whether it is already present on admission (POA) or not. Thus the mere existence of a pneumonia code among the SDXs does not unequivocally indicate a HAP/VAP. However, since 2009/10 the ICD-10 GM includes the code U69.00 as an indicator SDX for HAP/VAP, making HAP/VAP identification much easier.

In our dataset we excluded all patients with pneumonia or exacerbated COPD as PDX. All patients having a pneumonia code as SDX and no such code as PDX were included.

The PEG guidelines (2004/2005 version) for IIAT of HAP are similar to the ATS or IDSA guidelines in terms of the therapeutic options. We chose the PEG recommendations since they provide a convenient scoring system that defines three different patient groups based on the risk of the involvement of complicated pathogens. [26–31].

Using the initial data filter, we identified patients with probable HAP/VAP. We then performed the following analyses via medical chart review (if available):

In total, we received 211,084 datasets from the 5 hospitals for the year 2007. Through the initial data filter we identified 895 patients with probable HAP/VAP. 647 patients were excluded during chart reviews. For 223 patients the medical record was missing (could not be retrieved) or not complete (e.g laboratory or microbiology data). 166 patients turned out to be cases with CAP. This is due to the fact that pneumonia and another relevant condition were already present on admission but the other diagnosis was chosen as primary diagnosis for DRG purposes. In most cases this was acute left-heart failure combined with pneumonia. In 127 cases, pneumonia was coded as SDX but could not be validated via chart review. 45 cases had aspiration during or shortly after admission, 41 cases had been transferred from another hospital (i.e. IIAT started there) and 36 cases had pneumonia onset earlier than 48 hours after admission. Therefore, these three groups did not meet the definition of HAP, i.e. pneumonia occurring 48 h or later after hospital admission [25]. 33 patients had an ICU stay of more than 20 days, 2 cases had cystic fibrosis as underlying chronic illness which is not covered by the general recommendations for HAP/VAP. finally 1 patient was excluded since he was readmitted within less than 14 days and both stays were consolidated due to DRG regulations. The cost data therefore do not longer reflect the hospital stay in which the pneumonia occurred.

After this process, 221 patients were included in the final analysis set. Shapiro-Wilkes test of the subpopulations showed normal distribution of values in both subpopulations. T tests on showed no significant differences.

Figure 1 provides an overview of the inclusion process.

The association to clinical improvement and survival of possible cofactors such as age, gender, presence of neurological, cardiologic disease or polytrauma was tested by logistic regression.

The rates of clinical improvement were 64% for all patients, 82% (85/104) in the GA-group and 47% (48/103) in the NGA group, respectively (p < 0.001). The odds ratio of therapeutic success with GA versus NGA treatment was 5.821 (p < 0.001, [95% CI: 2.71212.497]). The odds ratio for age (elderly vs. non-elderly) was 0.983 (p = 0.215, [95%% CI: 0.957-1.010]), the OR for gender (female vs. male) was 1.030 (p = 0.940 [95% CI: 0.475-2.232]) and the OR for the disease groups (present vs. not present) was 1.048 (p = 0.902, [95% CI: 0.4962.214]).

The survival rates were 80% for all patients, 86% in the GA-group (92/107) and 74% in the NGA-group, respectively (0 = 0.021). The odds ratio of mortality for GA vs. NGA treatment was 0.565 (p = 0.117, [95% CI: 0.276-1.155]). Having neurological disease, cardiologic disease or polytrauma was found to be a strong negative predictor of survival with an OR of 0.241 (p = 0.012, [95% CI: 0.096-0.609]). The odds ratios for age (elderly vs. non-elderly) and gender (female vs. male) were 1.026 (p = 0.79, [95% CI: 0.9971.056]) and 1.246 (p = 0.545, [95% CI: 0.612-2.536]), respectively.

For the subsequent economical analyses the patients with lethal outcome were excluded. All outcome measures were compared via Student T test.

Patients in the subpopulation with adequate IIAT show a significantly shorter length of stay (23.9 vs. 28.3 days; p = 0.022), have significantly less hours of mechanical ventilation (175 vs. 274; p = 0.001), incurred lower total costs (EUR 28,033 vs. EUR 36,139; p = 0.006) and lower ICU costs (EUR 13,308 vs. EUR 18,666, p = 0.003). Drug costs for the hospital stay were also lower (EUR 4,069 vs. EUR 4,833) yet not significant.

We then analyzed the specific IIAT used and ranked the type of inadequacy of IIAT.

The most frequent type of inadequacy was monotherapy in patients with a CPRS score of 6 or higher (Group III) that should receive combination therapy according to PEG recommendations. The second most common type of inadequacy was the use of aminopenicillin/BLI (e.g. amoxicillin/clavulanic acid) instead of an acylaminopenicilin/BLI (e.g. piperacilline/tazobactam) in cases having a CPRS score of 3-5 (Group II). Another common deviation from the PEG recommendation was the use of a cephalosporin group 2/3a (e.g. cefazoline) instead of group 3b/4 (e.g. ceftazidime) in cases having a CPRS score of 3-5 (Group II).

The results of this study clearly show the clinical and economical superiority of adequate initial antibiotic treatment according to PEG guidelines. Although this study is unique in combining analyses of clinical and economical outcomes in HAP/VAP according to IIAT adequacy, other studies reveal similar results. Orrick et al. performed an analysis with a very similar design for community-acquired pneumonia (CAP). In their dataset, the majority of patients (75.8%) received preferred antibiotics according to the IDSA guidelines. The group treated with preferred antibiotics had a shorter mean length of hospital stay (4.5 vs 6.8 days, p = 0.002), a lower total cost of hospitalization (median USD 2,047 vs. 3,805, p = 0.021) and lower antibacterial costs (median USD 97 vs. 171, p = 0.038) compared with patients who did not receive the preferred therapy [33]. They concluded that introducing and intensifying adherence to IDSA guidelines reduces the economic burden associated with CAP.

Bird et al. performed an analysis comparing data before vs. after introduction of a guideline for a VAP prevention bundle in surgical ICUs (SICU). Prior to the introduction of the bundle, VAP was recorded at a rate of 10.2 cases per 1000 ventilator days. Compliance with the VAP bundle increased over the study period from 53% and 63% to 91% and 81% in the participating SICUs, respectively. The rate of VAP decreased to 3.4 cases per 1000 ventilator days. Estimated cost savinggs amounted to USD 1.08 million [34]. The authors concluded that the initiation and introduction of the bundle reduced the incidence of VAP and associated cost.

Improved survival was also shown in a retrospective analysis of CAP cases by Frei et al. They found that guideline-discordant therapy was associated with an increase in inpatient mortality (25% vs. 11%; odds ratio = 2.99 [95% CI, 1.08-9.54])[35].

Ott et al. (2) investigated patients with HAP/VAP and MRSA vs. MSSA as the underlying pathogen. For MRSA patients, the cost differences were much higher than in prior studies (€ 60,684 vs. € 38,731; p = 0.01). Moreover, more patients died in the MRSA-group. (13 vs. 1 deaths, p < 0.001). We could not do a subgroup analysis for the MRSA patients as we did not collect microbiology results in our study.

In our study, the most frequent type of inadequate IIAT was monotherapy when the guideline recommends a combination therapy. Particularly in patients at high risk of involvement of complicated pathogens (CPRS score of ≥ 6), the outcomes were significantly better when an adequate IIAT was chosen. This correlation was well established in analyses performed by Kollef et al., Kolenti et al. and other groups [8, 36–38].

Possible limitations of our study include the following. While the clinical and economical resource consumption was not significantly different between the subpopulation with adequate vs. inadequate IIAT and the overall morbidity was very well comparable, we found a significant difference in the CPRS score groups (3.5 +/-2.2 vs. 6.0 +/-2.5). It appears that patients at higher risk of complicated pathogen involvement are more likely to receive inadequate IIAT. Another possible explanation is that experience-based local standards for IIAT are well sufficient for patients with lower CPRS score but inadequate for patients with higher CPRS score.

If a standard therapy regimen is applied to all HAP/VAP patients without applying the risk score, the number of adequately treated patients would be lower in patients with high CPRS. This is underlined by an analysis of IIAT adequacy according to the PEG therapy groups: more group I patients received adequate IIAT compared to group III patients who were more likely to receive inadequate IIAT.

However other cost drivers could have influenced the differences between the subpopulations with adequate vs. inadequate IIAT. Therefore we analysed costrelevant CPRS parameters with a significant difference of occurrence in the GA vs. NGA groups, the most prominent being 'extrapulmonary organ failure' (1 vs. 12 patients) and 'late onset' (7 vs. 19 patients). Logistic regression did not reveal any significant associations with cost, LOS or HMV.

Adequate initial intravenous antibiotic therapy of HAP/VAP according to PEG guidelines was clinically superior and cost-saving vs. inadequate IIAT. Rates of survival rates and clinical improvement were significantly higher in patients treated according to guideline recommendations. Length of stay, duration of mechanical ventilation, total costs and ICU costs were significantly lower as well for the subpopulation receiving adequate IIAT.

The most frequent type of inadequate IIAT was monotherapy in patients where the PEG guidelines recommend a combination therapy.

Moreover it appears that experience-based therapeutic strategies which are most likely in place in every hospital are mostly sufficient for patients with lower risk for complicated pathogens but not for those with a higher CPRS score. Educating physicians on guideline recommendations in patients with higher CPRS scores - ideally adapted to the individual hospital's resistance patterns -- may improve treatment adequacy and thus outcomes in HAP/VAP with the add-on benefit of cost reductions. In particular, the use of combination therapies as recommended by the guidelines may help to achieve a higher rate of adequate IIAT and thus improve clinical outcomes and economic parameters.

Using the data from the German KISS surveillance program [5] one could extrapolate an yearly number of 30,090 HAP/VAP cases in German ICUs. Assuming that the observed rate of adequate IIAT of less than 50% was generalizable to all German ICUs treating HAP/VAP patients, the consistent use of IIAT according to PEG guidelines may save 2,042 lives and total costs of € 125,819,000.

M. H. Wilke received consultancy fees and honoraria from Pfizer Pharma GmbH, Novartis Pharma Gmbh, and Wyeth Pharma GmbH (until 2009).

R. F. Grube received consultancy honoraria from Pfizer Pharma GmbH, Novartis Pharma Gmbh, and Wyeth Pharma GmbH (until 2009).

K. F. Bodmann received honoraria for lectures from Pfizer Pharma GmbH, Novartis Pharma GmbH, Bayer Pharma GmbH, Infectopharm GmbH, Janssen-Cilag GmbH, B.R.A.H.M.S GmbH, MSD Merck, Sharp & Dome GmbH.