Using information theory to identify redundancy in common laboratory tests in the intensive care unit.

Joon Lee,David M Maslove

doi:10.1186/s12911-015-0187-x

Abstract

BackgroundClinical workflow is infused with large quantities of data, particularly in areas with enhanced monitoring such as the Intensive Care Unit (ICU). Information theory can quantify the expected amounts of total and redundant information contained in a given clinical data type, and as such has the potential to inform clinicians on how to manage the vast volumes of data they are required to analyze in their daily practice. The objective of this proof-of-concept study was to quantify the amounts of redundant information associated with common ICU lab tests.MethodsWe analyzed the information content of 11 laboratory test results from 29,149 adult ICU admissions in the MIMIC II database. Information theory was applied to quantify the expected amount of redundant information both between lab values from the same ICU day, and between consecutive ICU days.ResultsMost lab values showed a decreasing trend over time in the expected amount of novel information they contained. Platelet, blood urea nitrogen (BUN), and creatinine measurements exhibited the most amount of redundant information on days 2 and 3 compared to the previous day. The creatinine-BUN and sodium-chloride pairs had the most redundancy.ConclusionsInformation theory can help identify and discourage unnecessary testing and bloodwork, and can in general be a useful data analytic technique for many medical specialties that deal with information overload.

Highlights

Clinical workflow is infused with large quantities of data, in areas with enhanced monitoring such as the Intensive Care Unit (ICU)
We explored an alternative approach in which the science of information theory is used to identifying areas of overlap or redundancy between clinical tests
Almost all ICU admissions had at least one measurement of each lab test on day 1, progressively fewer ICU admissions with lab test results were available in MIMIC II on days 2 and 3

Summary

Introduction

Clinical workflow is infused with large quantities of data, in areas with enhanced monitoring such as the Intensive Care Unit (ICU). Information theory can quantify the expected amounts of total and redundant information contained in a given clinical data type, and as such has the potential to inform clinicians on how to manage the vast volumes of data they are required to analyze in their daily practice The objective of this proof-of-concept study was to quantify the amounts of redundant information associated with common ICU lab tests. Conventional data based on clinical observation and blood tests are collected in large quantities, with newer modalities such as high-resolution imaging studies and genome-wide assays proliferating rapidly. Strategies are needed both to limit the use of unhelpful tests, and to identify from among the resulting data those that are relevant, and those that are merely distractions.

Methods

Results

Conclusion