Setting and patient population

In January 2000, NEO-KISS was started as a prospective national surveillance system for the most relevant nosocomial infections in VLBW infants (birthweight < 1500 g) in Germany [7]. By January 2008, 213 neonatology departments participated in KISS surveillance. In NEO-KISS, all VLBW infants admitted to the participating NICUs are kept under surveillance from admission to discharge, transfer or death, and all healthcare-associated infections are recorded.

Data collection

We collected data on the structure characteristics of the neonatological departments and on the working processes of their NICUs with two separate point questionnaires. The first enquiry took place in August 2008, the second in April 2009. Information requested about NICU structure included data on number of wards and beds of the concerning departments, VLBW patient days in 2007, the existence of handover and/or infectious diseases rounds, and existence of microbiological screening, among other items. The NICU processes assessed included e.g. infusion preparation, CVC and incubator hygiene management, unit-level hand hygiene performance and infusion system use. The “realized staffing” for each unit was defined as realized number of nurses at the time of the questionnaire enquiry divided by the (individual hospital based) planned number of nurses. This ratio was multiplied by 100 and is displayed in the results section as “percentage of realized staffing”. Patient data was obtained from the NEO-KISS database for the departments that took part in our enquiry. The data from January 1^st 2008 to June 31^st 2009 were analyzed. In the NEO-KISS surveillance database, basic demographic data on patients (e.g. birthweight, sex, way of childbirth, device-days) is collected. Length of stay is defined as the number of days between birth or admission and until a weight of 1,800 g has been reached, the patient’s discharge or death. The pooled CVC utilization ratio was obtained as number of device-days per 100 patient-days. The pooled incidence density was defined as number of BSI cases per 1,000 patient days.

Definitions of healthcare-associated infections

NEO-KISS focuses on bloodstream infections, pneumonia and necrotizing enterocolitis. Infection is classified as healthcare associated if arised in hospital after the first 72 h of life or 72 h after admission. NEO-KISS uses modified definitions from the Center for Disease Control and Prevention (CDC) for healthcare-associated infections. They are already thoroughly described in former publications [1, 8]. A healthcare-associated BSI was considered central venous catheter-associated if the catheter was present within 48 hours before the infection occurred or if the catheter was still present at infection onset. According to the modified definitions of the CDC for primary sepsis, we stratified the cases in clinically-diagnosed BSI and laboratory-confirmed diagnosis (LCD) of BSI [1, 8]. The LCD BSI cases were further classified by proven pathogens in the two groups coagulase negative staphylococci only (CoNS) or other than CoNS.

Statistical analysis

In the descriptive analysis, we calculated numbers and percentages and/or median and inter-quartile range (IQR; 25% percentile - 75% percentile). In the multivariable analysis, we used logistic regression models to investigate the association between occurrence of CVC BSI and various patient, structure and process parameters. We investigated two different endpoints: CVC BSI and CVC BSI cases with laboratory confirmed (LCD) diagnosis other than CoNS. Since observations within one neonatology department are not statistically independent due to department-dependent policies, adjusted incidence rate ratios with 95% confidence intervals (CI) were estimated based on generalized estimating equation (GEE) models which account for this clustering effect by using an exchangeable correlation structure [9, 10]. The log number of patient days was treated as an offset in the model. For the occurrence of bloodstream infection, the multivariable model building strategy was performed in 3 steps. First step: All patient-based parameters were considered in a logistic regression model by stepwise forward variable selection with the significance level p=0.09 for including a parameter in the model and p=0.10 for excluding a parameter. Second step: All process and structure parameters were considered in the resulting model from step 1 by stepwise forward variable selection using the same significance levels for including and excluding. Third step: With all parameters included by step 1 and 2, a GEE model was calculated that took cluster effects within a department into account. By stepwise backward selection only significant parameters remained in the final model. Excluding criteria were the smallest Chi-square value and p≥0.05 in the Type III score statistic. The quasi-likelihood information criterion (QIC) as a modification of the Akaike information criterion was used as goodness-of-fit measure in the GEE model. P-values less than 0.05 were considered significant. All analyses were performed using SAS (SAS Institute, Cary, NC, USA) [1–3]. In Germany, anonymised secondary data research does not require human research committee review.

Basic characteristics of patients and departments

Frequency of different BSI diagnoses

The pooled incidence of BSI was 15.2 per 100 patients, in median 11.7 (IQR 6.3-19.2). The pooled incidence density of BSI was 4.6 per 1000 patient days (median 3.5; IQR 2.1-5.7). Of the 954 (100%) cases of healthcare-associated BSI, 482 (49%) were laboratory confirmed. Two hundred fifty-eight cases (27%) were due to coagulase negative staphylococci only, 214 (22%) were due to pathogens other than CoNS. Four-hundred fourteen BSIs (43%) were CVC associated. The pooled incidence density was 8.3 per 1,000 CVC days (median 6.8, IQR 1.8-12.2). The pooled CVC utilization ratio was 24.3 per 100 patient days.

Risk factors for CVC BSI

Risk factors for LCD BSI other than CoNS