Authors' objectives:

The purpose of this review was to look at methodological development and evaluation of 30-day mortality as quality indicator for Norwegian hospitals

Authors' recommendations: The main issue is whether mortality measures based on administrative data are valid indicators of true, hospital-specific mortality, while accounting for presumed bias, resulting from inaccurate coding, diagnostic variability and less than ideal case-mix adjustment. On the one hand, there remains a possibility that the bias of the indicator is large enough to influence the comparison between hospitals in a significant way. On the other hand, the results indicate that there are unacceptably high differences between hospitals. A review of the literature indicates that these differences seem to agree with those reported internationally. It is the role of the public health authorities to weigh the risk of incorrectly exposing hospitals as having poor quality, against the possibility that large apparent discrepancies in mortality reflect a true situation. The present study is limited by the lack of clinical data and independent validation of diagnoses and codes. Within these limits, we have performed a study of plausible bias magnitudes indicating that unacceptable bias is probably avoided. Still, we feel that the issue is not settled in a satisfactory way. Further study, geared towards resolving the bias question, is recommended. We have identified some less fundamental issues that need to be addressed: the need for more reliable registration of very early deaths, or choosing a strategy to reduce the sensitivity of 30D mortality to these cases, particularly for acute myocardial infarction, as well as the use of correct decision rules to identify hospitals as performance outliers. It is necessary to finalize the decision rules and their parameters, based on discussion with the various users of the indicators. Besides bias, the most important criterion is precision. We have shown that the mortality indicator can be used to identify, with good statistical precision, hospitals where the probability of dying is appreciably different from the average. Further studies should focus on validation of the results using clinical and laboratory data in addition to information from journals and direct communication with hospitals.

Authors' methods: Review

Project Status: Completed

URL for project: http://www.nokc.no/

Year Published: 2005

English language abstract: An English language summary is available

Publication Type: Not Assigned

Country: Norway

Organisation Name: Norwegian Institute of Public Health

Contact Address: Universitetsgata 2, Postbox 7004 St. Olavs plass, NO-0310 Oslo NORWAY. Tel: +47 23 25 50 00; Fax: +47 23 25 50 10;

Contact Name: Berit.Morland@nokc.no, dagny.fredheim@nokc.no

Contact Email: Berit.Morland@nokc.no, dagny.fredheim@nokc.no

Copyright: The Norwegian Knowledge Centre for the Health Services