Empirical aspects of linking intensive care registry data to hospital discharge data without the use of direct patient identifiers

In the field of intensive care, clinical data registries are commonly used to support clinical audit and develop evidence-based practice. However, they are often restricted to the intensive care unit episode only, limiting their ability to follow long-term patient outcomes and identify patient readmissions. Data linkage can be used to supplement existing data, but a lack of unique patient identifiers may compromise the accuracy of the linkage process. The aim of this study was to assess the quality of linking the Australia/New Zealand critical care registry to a state financial claims database using a method without direct patient identifiers and to identify possible sources of bias from this method. We used a linkage method relying on indirect patient identifiers and compared the accuracy of this method to one that also included the patient medical record number and date of birth. The overall linkage rate using the method with indirect identifiers was 92.3% compared to 94.5% using the method with direct identifiers. Factors most strongly associated with not being a correct link in the first method included patients at one study hospital, admissions in 2002 and 2003 and having a hospital length of stay of 20 days or more. Linking the Australia/New Zealand critical care without direct patient identifiers is a valid linkage method that will enable the measurement of long-term patient survival and readmissions. While some sources of bias have been identified, this method provides sufficient quality linkage that will support broad analyses designed to signal future in-depth research.