NEO-KISS is a patient-based surveillance method for VLBW that includes patients in surveillance until a weight of 1800 g is achieved, if they do not die or are transferred earlier. The detailed surveillance method used in NEO-KISS is described elsewhere[1],[2]. It can also be found underhttp://www.nrz-hygiene.de/en/surveillance/hospital-infection-surveillance-system/neo-kiss/ together with the latest reference data. The following variables were collected for all patients: birth weight, sex, multiple birth, gestational age and type of delivery. Cases of primary bloodstream infections (BSI) and pneumonia were determined using modified CDC definitions[2]. Surveillance persons from each neonatal intensive care unit (NICU) have to attend an introductory course before starting data collection where the definitions are explained and trained with case studies.
Definitions and selection criteria for analyzed NICUs
In order to compare voluntary and mandatory surveillance data, two groups were defined by the surveillance start date of the first patient within a single NICU and the time of surveillance between the first and last patient.
Group 1 consisted of the old participants with voluntary participation. They started patient surveillance between January 2000 and December 2002 and participated continuously for 3 years (1095 days) in NEO-KISS. 26 NICUs met this criterion.
Group 2 consisted of the new participants with mandatory participation. They started with patient surveillance between January and December 2006 and participated continuously for 3 years. 95 NICUs met this criterion.
Influence on infection rates
To answer the first question, data from 2007 were examined. NICU parameters and infection rates of the NICUs in both groups were compared by Chi-square or Wilcoxon rank sum test.
Influence on surveillance effect
To stimulate further infection control measures, all NICUs participating received a biannual feedback report including crude and standardized infection rates of their own NICU compared with the national reference data until 2006. From 2007 on, a web-based data entry with the possibility of immediate data calculation and feedback has been employed. Problems of surveillance were discussed and prevention activities were shared at annual workshops. Because primary BSI is the most frequent HCAI in VLBW infants, the influence of participation in the surveillance system on HCAI rates focused on the incidence density of primary BSI. In an earlier investigation of 24 units during their first three years of continuous participation, a significant decrease in BSI between first and third year of participation was found[3].
For the second question, all VLBW-infants within the first three years of participation in both groups were considered. In the univariable analysis, primary BSI incidence densities in the individual years of participation were determined and compared. The relative risks (RR) and their 95% confidence intervals (CI95) were calculated. In the multivariable analysis, logistic regression and Poisson regression models were performed to identify significant risk factors for the occurrence of BSI. The following parameters were considered: birth weight (5 categories, 250 g steps), gestational age (4 categories, <27/27-28/29-30/>30 weeks), sex, mode of delivery (sectio, emergency sectio), multiple birth, surveillance end point (3 categories: 1800 g/transfer/died), and VLBW-volume year 2007 (≤/>30 VLBW infants). In the Poisson regression model, the log number of patient days was treated as offset parameter for number BSI. All parameters were considered in a full model and model parameters were excluded stepwise by the smallest chi-square value and p ≥ 0.05 in the type III test. The adjusted effect measures for the year of participation were calculated by generalized estimating equation (GEE) models that consider cluster effects within a NICU. In this model, all significant parameters from the first model building step were included. However, for face validity reasons, we added sex into all final models. The quasi-likelihood information criterion (QIC) as a modification of the Akaike information criterion (ACI) was used as goodness-of-fit measure in the GEE model. P values of less than 0.05 were considered significant. All analyses were performed with SPSS (IBM SPSS Statistics; IBM Corporation, Armonk, NY, USA) and SAS (SAS Institute Inc., Cary, NC, USA).
Our study was based on surveillance data. All data were anonymous and collected in accordance with the Germanhttp://recommendations of good epidemiological praxis with respect to data protection. As a federal law, the German Protection against Infection Act (Infektionsschutzgesetz §23) regulates the prevention and management of infectious disease in humans. All hospitals are obliged to collect and analyse continuously nosocomial infections and resistant pathogens. These routine data were reported to the National Reference Centre of the Surveillance of Nosocomial Infections. Ethical approval and informed consent were thus not required.