Bayesian exploration of dependencies of laboratory tests and evaluation of test redundancy
Zeyneb Guenfoud, Péter Antal · 2018
Laboratory tests are frequently used in exploratory diagnostics when medical experts use a collection of tests to find the causes of symptoms. Two common types of errors in this scenario are (1) the omission and subsequent overlook of informative tests, frequently with predictably abnormal values, and (2) the requests of unnecessary measurements, frequently predictably with normal values. We present results confirming the applicability of clinical laboratory data to diminish these errors, both related to multivariate, systems-level predictability of tests. We explore the dependency structure of laboratory tests using Bayesian networks, and we quantified the shared information content of individual laboratory tests with all the other tests.