This poster details an unsupervised learning algorithm for the automated calculation of reference intervals for laboratory tests. The pipeline was validated for pediatric clinical chemistry testing using Canadian CALIPER as gold standard reference database. The results of the pipeline can inform hospital systems on how to fill in population-specific gaps in their reference intervals, a common problem for pediatric health care delivery.

Learning Objective 1: Understand the current issues surrounding population reference intervals particularly in the pediatric population. How to sample reference patients using clinical data to identify "healthy" individuals upon which to base a population reference.

Learning Objective 2: Appreciate the scale and pipeline throughput for informing clinical decision making in determining population specific reference ranges in clinical chemistry.


Nathan Patel (Presenter)
University of Michigan

Lee Schroeder, University of Michigan

Presentation Materials: