Development of a survival model for unexpectedly long hospital stays

CBS has investigated the possibility of developing a new model for the indicator for unexpectedly long hospital stays.
The current indicator calculates the percentage of clinical hospital stays with a duration 50% longer than expected. This indicator is used to monitor the quality and safety of hospital care in the Netherlands. Hospitals use the indicator to investigate possible improvements in care. With the current indicator, that corrects for a limited set of properties of hospital stays, there are differences in outcome between types of hospitals that are possibly explained by differences in hospital populations not corrected for. Therefore, DHD, the organisation that collects the hospital discharge data and calculates this indicator, has asked Statistics Netherlands to develop a new model taking into account a larger number of patient properties. This report describes this research. It appears that, irrespective of the method used, for a large part the same hospital stays are classified as of an unexpectedly long duration. Furthermore, with the new model there is still a correlation between type and size of the hospital and the outcome of the indicator. Further research is necessary into the causes of the big differences in outcome between (type of) hospitals.