JSTOR

Methods

The Krankenhaus Infektions Surveillance System (KISS) is a Web‐based national surveillance system for nosocomial infection in Germany.⁸ The surveillance method used by KISS is almost identical to that of the National Nosocomial Infections Surveillance System.⁹ The definitions of the Centers for Disease Control and Prevention are used for diagnosing nosocomial infections,¹⁰ device‐associated infections are calculated, and complications such as the development of secondary BSIs or death are recorded. Despite the fact that the main objective of KISS is to provide data for action, this information may also be useful for epidemiological studies.

For this study, all KISS data for pneumonia and primary BSIs occurring in ICUs from January 1997 through June 2003 were used. The following risk factors were considered: type of ICU, size and type of hospital, time from admission to infection, sex, age, mechanical ventilation before the onset of pneumonia, central venous catheter use, total parenteral nutrition before the onset of primary BSI, type of pathogen, and year of infection. Whenever a nosocomial infection occurred, the pathogens identified were communicated to the surveillance system; up to 4 pathogens could be recorded for each infection site. When more than 1 pathogen was reported, all the microorganisms were considered for analysis, because it is almost impossible to identify the single causative agent of a specific nosocomial infection.

Resistance data were collected only for some frequent pathogens and relevant antibiotics. For Staphylococcus aureus, we recorded whether the infection was caused by methicillin‐susceptible S. aureus or methicillin‐resistant S. aureus (MRSA), and for enterococci, resistance to vancomycin was documented. For gram‐negative pathogens, the definitions of multidrug resistance used in this study are specified in Table 1. For this investigation, the 10 most frequent pathogens associated with nosocomial pneumonia and primary BSI were included. If a pathogen was detected in a patient with pneumonia group, it was also included in the analysis of primary BSI.

Table 1.  Definitions of Multidrug Resistance Used for Gram‐Negative Pathogens in the Krankenhaus Infektions Surveillance System

Open New Window

For the analysis, only complete data sets were considered. For univariate analysis (contingency tables), the Fisher exact test was used, with a significance level of .05. Multiple logistic regression analysis was performed with stepwise variable selection using a commercial statistical package (SAS). Because of the large database for pneumonia cases, a significance level for entry into and remaining in the model for pneumonia was set at less than .001. For primary BSIs, a significance level of .05 was used.

Results

By the end of June 2003, we had data for 680,299 patients observed during 9,334 months of surveillance in 289 ICUs. The most frequent nosocomial infection was pneumonia (9,818 cases) and nosocomial primary BSI (3,225 cases). Complete data sets were available (and thus included in the study) for 8,432 cases of pneumonia (85.9%) from 202 ICUs and 2,759 primary BSI cases (85.6%) from 190 ICUs.

A total of 750 patients (8.9%) with nosocomial pneumonia and 302 patients (10.9%) with nosocomial primary BSI died. The most frequent pathogen associated with nosocomial pneumonia was S. aureus (24.4 isolates per 100 cases), followed by Pseudomonas aeruginosa (16.5 per 100 cases), Klebsiella species (11.8 per 100 cases), Escherichia coli (10.0 per 100 cases), and Enterobacter species (8.6 per 100 cases). Of the nosocomial primary BSIs, coagulase‐negative staphylococci (32.7 isolates per 100 cases) was the most frequent pathogen, followed by S. aureus (15.6 per 100 cases), Enterococcus species (11.8 per 100 cases), Enterobacter species (5.1 per 100 cases), and Klebsiella species (4.7 per 100 cases). A total of 21.5% of the S. aureus isolates causing pneumonia were MRSA, and 8.4% of the P. aeruginosa isolates and 11.2% of the Stenotrophomonas maltophilia isolates were also considered multidrug resistant, according to our definition of multidrug resistance. Of the S. aureus isolates that caused primary BSIs, 26.5% were MRSA, and 9.4% of the P. aeruginosa isolates were also multidrug resistant. The prevalence of multidrug resistance among each of the other types of pathogens included in the analysis did not exceed 5% for either nosocomial pneumonia or nosocomial primary BSI. The distribution of patients with nosocomial pneumonia and primary BSI according to risk factor appears in Table 2, and the distribution according to infecting pathogen is specified in Table 3.

Table 2.  Distribution of Potential Risk Factors for Death Among Intensive Care Unit Patients With Pneumonia or Primary Bloodstream Infection (BSI)

Open New Window

Table 3.  Distribution of Pathogens Recovered From Intensive Care Unit Patients With Pneumonia or Primary Bloodstream Infection (BSI)

Open New Window

According to findings of univariate analysis, the mortality rate associated with nosocomial pneumonia was significantly higher for patients treated in a medical or surgical ICU, when the hospital had more than 1,000 beds, when the hospital was a major teaching hospital (but not a university hospital), when the median length of time to infection was greater than 6 days, and when the patients’ age was older than the median of 65 years. The presence of MRSA (but not all S. aureus strains), P. aeruginosa (in general and if resistant to multiple drugs), and S. maltophilia was also associated with increased mortality from nosocomial pneumonia. For nosocomial primary BSI, a significantly higher mortality rate was found when the patients were cared for in a major teaching hospital (not a university hospital) and if the patient age was older than the median of 63 years. No individual pathogens were significantly associated with increased mortality from nosocomial primary BSI in the univariate analysis (Table 4).

Table 4.  Risk Factors for Death Among Intensive Care Unit Patients With Pneumonia or Primary Bloodstream Infection (BSI)

Open New Window

Multivariate logistic regression analysis confirmed that treatment in a medical or surgical ICU, hospital size of more than 1,000 beds, and age older than the median value were risk factors for increased mortality from nosocomial pneumonia. Both MRSA and multidrug‐resistant P. aeruginosa were also associated with increased mortality from pneumonia (Table 5). Affiliation with a university hospital was the only factor associated with lower mortality risk from pneumonia. Age older than the median value and MRSA were also associated with increased mortality from nosocomial primary BSI, whereas university affiliation and S. aureus or coagulase‐negative staphylococci as causative pathogens were significant protective factors.

Table 5.  Findings of Multiple Logistic Regression Analysis of Risk Factors for Death Among Intensive Care Unit (ICU) Patients With Pneumonia or Primary Bloodstream Infection (BSI)

Open New Window

Discussion

To the best of our knowledge, this is the only study to date to use surveillance data for investigating risk factors for death due to the 2 most important nosocomial infections in ICUs: pneumonia and primary BSI. The overall mortality of 8.9% for pneumonia and 10.9% for primary BSI calculated by us is relatively low, compared with findings from other studies that investigated the mortality rate for nosocomial infection.^7,11‐14 This discrepancy may be associated with the fact that, in Germany, because of the relatively small number of intermediate care units, many patients with less severe illness, who in other countries are treated in such units, are routinely treated in ICUs.

Furthermore, we used the sensitive but relatively less specific Centers for Disease Control and Prevention criteria for diagnosing nosocomial infection.¹⁰ This approach may lead to classifying a patient as having pneumonia when in reality no pneumonia is present. The diagnostic quality of the individual ICUs should also be considered when interpreting the data. Identification of a pathogen from blood culture (after the exclusion of any skin contaminants) shows beyond any doubt that BSI is present; therefore, careful ruling out of secondary BSI is necessary. However, only a subgroup of ICUs routinely perform bronchoalveolar lavage if nosocomial pneumonia is suspected, whereas others use clinical criteria, in addition to identification of pathogens from tracheal secretions, for diagnosing pneumonia. Some uncertainty therefore persists regarding whether some cases of pneumonia were recorded only because a microorganism was found in the tracheal secretions.

As mentioned in Methods, up to 4 pathogens can be recorded and referred to the surveillance system for each infection site whenever a nosocomial infection has occurred. Because it is almost impossible to identify the causative agent for each nosocomial infection, all microorganisms were considered in our analysis. This approach, as well as the interaction of pathogens, may lead to a bias with regard to the risks associated with individual pathogens.

Furthermore, we did not record any information in the national surveillance system on the antibiotics used or the timing schedules of antibiotic therapy, so we are unable to ascertain the role of antibiotics on outcome. To account for the influence of new antibiotics introduced during the study period on the mortality rate, we accordingly adjusted the data for each year of infection.

The most important limitation of our study was the absence of a generally applied severity of illness scoring system in German ICUs. We, therefore, were unable to adjust for patients’ severity of illness. The only possibility was to adjust for age, sex, device use, and time in the ICU before infection as a surrogate parameter for severity of illness, but adjusting for the median time to infection cannot account for a prior colonization with multidrug‐resistant organisms that result from previous treatment in other hospitals or departments.

Previous studies have not described treatment in a medical or surgical ICU as an independent risk factor for a worsening prognosis in patients with nosocomial pneumonia. Two smaller studies^15,16 and the EPIC (European Prevalence of Nosocomial Infections in Intensive Care Units) study¹² did not find that the mortality rate differed according to the type of ICU. The increased risk of death associated with nosocomial pneumonia in hospitals of more than 1,000 beds should be considered in light of the observation of a reduced risk for patients treated in a university hospital in Germany with more than 1,000 beds. However, whether the protective effect of the university status of the hospital in which the ICU is located is due to better treatment or other imponderable factors remains a open question. For instance, one may speculate whether understaffing and overcrowding, as well as other risk factors described elsewhere,¹⁷ are less prevalent in university hospitals than in other teaching hospitals. For this study, we were also not able to consider the mortality rates for patients without nosocomial pneumonia or primary BSI. Thus, it remains unclear whether these findings merely reflect the generally higher mortality in these ICU patient groups. Luna et al.¹⁸ found that early initiation of empirical antibiotic therapy was associated with a decreased mortality rate if treatment was adequate, compared with inadequate therapy or no treatment at all. The rates of resistance in university hospitals may be higher because many patients with multiresistant pathogens have already been admitted. University physicians are considering this possible explanation when deciding which empirical therapy to use. (This may be a possible explanation for this finding.) Similar to findings in other studies, age was identified as a risk factor for death among patients with pneumonia and patients with primary BSI.^12,14,19,20

The identification of MRSA as a risk factor for death from nosocomial primary BSI is not surprising, because this has been shown in a number of studies, as well as in meta‐analyses by Cosgrove et al.²¹ and Whitby et al.²² The protective effect when S. aureus or coagulase‐negative staphylococci were identified as the causative organisms for primary BSI can be explained by the fact that empirical therapy for BSI takes into account the presence of these pathogens in almost all cases. P. aeruginosa was not identified as a risk factor for death in the hospital due to primary bacteremia in a recent study by Osmon et al.,⁶ who included all cases of bacteremia rather than only cases of primary bacteremia.

Several studies have identified gram‐negative pathogens, in particular, P. aeruginosa, Acinetobacter species, and Stenotrophomonas species as significant risk factors for death in patients with nosocomial pneumonia.^4,23 Other studies were not able to confirm these findings, however. For instance, the results of a sensitivity analysis by Heyland et al.³ did not show that the mortality rate among patients with high‐risk organisms (eg, P. aeruginosa, Acinetobacter species, Stenotrophomonas species, and MRSA) was greater than that among patients with low‐risk pathogens. Blot et al.²⁴ were also not able to identify detection of P. aeruginosa as an independent predictor of mortality. Our data showed increased mortality from pneumonia due to P. aeruginosa and S. maltophilia, but this finding was not confirmed by multivariate testing. This increased mortality may be associated with the high significance level of <.001 used for this analysis. Interestingly, detection of multidrug‐resistant P. aeruginosa and MRSA remain significant risk factors, even at this high level. Bercault and Boulain,¹³ using a prospective matched risk‐adjusted cohort study, also showed that nosocomial pneumonia caused by multidrug‐resistant microorganisms (including P. aeruginosa strains resistant to imipenem or ceftazidime and MRSA) was significantly associated with death. However, in contrast to the number of studies that demonstrated the significantly higher mortality rates associated with MRSA in patients with primary BSI, information about the consequences of pneumonia in patients with MRSA, compared with the consequences in patients with methicillin‐susceptible S. aureus, is still rare.^25‐27

Conclusion

In addition to age and other probable severity‐of‐illness factors (which were not possible to investigate in this study), the ICU structure (ie, the type and size of the ICU, whether it is affiliated with a university hospital, and the size of the affiliated hospital) may also influence the outcome of patients with nosocomial pneumonia and primary BSI, although it is unclear whether this is only the result of specific conditions within the German health system. The type of pathogen, particularly multidrug‐resistant pathogens, was also a strong predictor of the outcome of ICU patients. However, on the basis of this study, it is impossible to conclude whether resistance to multiple drugs was the crucial factor for patient outcome or whether such resistance was a surrogate parameter for the patients’ severity of illness. The results of this study underline the need for further investigation of the role of antimicrobial resistance in the outcome of patients with nosocomial pneumonia and patients with primary BSI.