Data-Analytics Competition: Forecasting Future Preanalytical Errors

Does your institution suffer from a large number of hemolyzed samples? Do you want to learn how to approach this problem while conserving limited resources? This morning’s session, “Data Analytics Competition: Forecasting Future Preanalytical Errors,” will tackle precisely that.

 Mark Zaydman, MD, PhD, assistant professor of pathology and immunology at Washington University School of Medicine, will announce the winners of a unique competition focused on using data science to forecast preanalytical errors caused by improper blood-specimen collection. The session will highlight the lessons that medical laboratory professionals can take away from the competition.

“Members of the laboratory medicine community are well poised to lead in applying data analytics and digital technologies into laboratory practice,” says Zaydman.

The competition was co-hosted by the Section of Pathology Informatics of Washington University in St. Louis and the ADLM Data Analytics Steering Committee. It used a machine-learning and data-science website called Kaggle to crowd-source solutions to problems across disciplines.

This year’s competition, called “Help with Hemolysis,” supplied a real-world, de-identified dataset representing hemolysis in the clinical lab. The participants were asked to use the dataset to assess which blood-specimen collectors would benefit the most from being trained on phlebotomy best practices.

The goal was to identify ways to minimize in vitro hemolysis while making efficient use of laboratory time and resources. The winning solution could guide institutions on how to spend their educational capital more effectively.

Educational interventions are known to help minimize hemolysis, but they are costly and may be temporary, especially if staff turnover is high. The teams had a little over a month to work with the supplied dataset before submitting their solution. They were also required to submit their code to provide insight into how they approached the problem. A total of 18 teams participated in the challenge.

After presenting an overview of the competition and summarizing the strategies used by the teams, Zaydman will present the winner—Team Hemolyers. The winning team will then share their solution so attendees can learn more about who they are and how they approached the problem, as well as ask any questions.

The creative education style of this session deviates from the traditional didactic lecture, and it is designed to appeal to people with all levels of experience. The competition format is an engaging way to learn, build a collaborative community, and identify powerful solutions to real-world problems.

This is the second year this competition has been held. After hosting it twice, Zaydman notes that both times the winner has been an interdisciplinary team that combines clinical laboratory expertise and computer-science knowledge. This is largely because data-science tools and computational resources are increasingly accessible to those without an advanced degree in data science.

Interdisciplinary experts can identify gaps in patient care and work together to produce creative and practical solutions. Zaydman emphasizes that, although this year's task was particularly challenging, the results were impressive. “A valuable model doesn’t have to be perfect, as long as it saves costs and improves patient care.”