Project Details
Projekt
Data mining of medical information systems to improve pediatric decision making guided by laboratory test results
Applicant
Privatdozent Dr. Jakob Zierk
Subject Area
Pediatric and Adolescent Medicine
Medical Informatics and Medical Bioinformatics
Medical Informatics and Medical Bioinformatics
Term
since 2021
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 462065145
Laboratory tests are an important tool in modern medicine with diagnostic and therapeutic implications for most diseases. Reference intervals reflect the distribution of laboratory test results in a healthy population, and are essential when using laboratory test results to guide clinical decisions. During childhood and adolescence, laboratory test results are subject to considerable changes due to physiological growth, however, currently available reference intervals do not adequately reflect these unique dynamics. This restricts the value of laboratory diagnostics in all pediatric age groups, particularly in premature and newborn infants, although precise age-adjusted reference intervals are especially important in these children due to their substantial contribution to pediatric morbidity and mortality. Furthermore, currently used strategies for the interpretation of laboratory findings (sequential classification of separate test results with a subjective assessment of the importance of a test result’s aberrance, and limited application of systematic approaches to the recognition of all disease-relevant patterns) can be supported by modern machine learning algorithms.In extensive preliminary work we have established data-driven methods and a multicenter study to determine reference intervals using laboratory test results obtained during patient care. This has enabled us to determine pediatric reference intervals with unprecedented accuracy and to close important gaps in pediatric laboratory medicine. However, due to limited data standardization and privacy-related restrictions, the available data set did not contain clinical information (e.g. diagnoses or outcome data). Our previous analyses were therefore limited to frequently performed laboratory tests in term neonates and children, and the accuracy of reference intervals in young infants was limited. Recently, the German Medical Informatics Initiative (MII) has made the "core data set" available, enabling the secondary data use of laboratory test results, diagnoses, interventions, and outcomes in all German university hospitals while respecting privacy regulations, thus enabling a substantial expansion of our previous investigations.In the proposed project, we will perform multi-center analyses of the MII core data set to determine reference intervals for premature infants, establish reference intervals for a comprehensive panel of laboratory tests (i.e. including analyses performed only under specific clinical suspicions), and improve the (especially age-related) accuracy of pediatric reference intervals. We will then use the established reference intervals to implement modern strategies for the interpretation of laboratory tests in children, which increase the clinical benefit of laboratory diagnostics by integrating clinical information and multiple laboratory test results (multidimensional classification), and use machine-learning algorithms to predict clinically relevant endpoints.
DFG Programme
Research Grants
