Objectives: Indirect data mining methods have been proposed for review of published reference intervals (RIs), but methods for identifying patients with a low likelihood of disease are needed. Many indirect methods extract test results on patients with a low frequency blood sampling history to identify putative healthy individuals. Although it is implied there has been no attempt to validate if patients with a low frequency blood sampling history are healthy and if test results from these patients are suitable for RI review.
Methods: Danish nationwide health registers were linked with a blood sample database, recording a population of 316,337 adults over a ten-year period. Comorbidity indexes were defined from registrations of hospital diagnoses and redeemed prescriptions of drugs. Test results from patients identified as having a low disease burden were used for review of RIs from the Nordic Reference Interval Project (NORIP).
Results: Blood sampling frequency correlated with comorbidity Indexes and the proportion of patients without disease conditions were enriched among patients with a low number of blood samples. RIs based on test results from patients with only 1-3 blood samples per decade were for many analytes identical compared to NORIP RIs. Some analytes showed expected incongruences and gave conclusive insights into how well RIs from a more than 10 years old multi-center study (NORIP) performed on current pre-analytical and analytical methods.
Conclusions: Blood sampling frequency enhance the selection of healthy individuals for reviewing reference intervals, providing a simple method solely based on laboratory data without the addition of clinical information.
Keywords: Nordic reference interval project (NORIP); data mining; indirect reference interval.
© 2021 Walter de Gruyter GmbH, Berlin/Boston.