Tjallie van der Kooi

harder you look the more you find’) [17, 18]. Surveillance definitions sometimes differ from clinical definitions, and some clinicians therefore challenge the cases recorded by the surveillance team. Frequent changes of infection control staff, as observed in Dutch hospitals, may additionally affect the consistency of recording.  Analysis and interpretation of data. To better understand trends in time and differences between centres case mix adjustment is important. To avoid interference with the infection, the collected patient and procedure/device characteristics are preferably determined at admission or at the start of the medical intervention that poses the risk. These characteristics should also be uniformly available in the electronic patient files of different hospitals. In practice, therefore, surveillance must often seek a compromise with limited data on patient and procedure/devices [19‐21]. The choice of these potential risk factors is based on clinical knowledge and on earlier studies. However, studies may differ in patient population and health care setting and consequently in their findings. Apart from this, the HAI incidence in these earlier studies will most often be higher, with therefore more discriminatory risk factors than at the time a surveillance protocol is drawn up. For example, the analysis in chapter 2 and 3 demonstrated that some of the recorded risk factors, which were selected based on earlier studies, e.g. the APACHE II score, did not appear to be associated with CRBSI risk in the Dutch surveillance data. A longer ICU stay before CVC insertion, which can serve as a proxy for disease severity, was associated with a higher CRBSI risk. Again, with further decreasing CRBSI risk in the ICU, it was no longer associated (data not shown). Another possible limitation, following from the requirement that data for surveillance should be easily retrievable, is the lack of detail. For example, in the Dutch CRBSI surveillance, catheter applications are not recorded on a daily basis nor per lumen. The resulting data do not allow a very accurate assessment of e.g. the CRBSI risk of TPN. Daily recording of a relevant patient or device characteristic, as practiced in the VAP surveillance programme, enables better evaluation of the associated risk, including as a function of time (chapter 4), but proved too laborious for routine surveillance.  Timely feedback of surveillance results to HCWs and hospital management is necessary to be able to act upon the results and improve the care in question. Hospitals that participated in the Dutch surveillance programme received feedback reports annually or more frequently if requested. For CRBSI and VAP, these included the predicted incidence given the distribution of the hospital’s patient population over the major ‘uncontrollable’ risk factors (use of the CVC for CRBSI; specialty for VAP). Hospital participating in prevalence surveys receive a report with the prevalence stratified by a number of patient characteristics and by clinical specialty. 10 239 General Discussion

RkJQdWJsaXNoZXIy MTk4NDMw