IcuARM-An ICU Clinical Decision Support System Using Association Rule Mining

From Clinfowiki
Revision as of 03:41, 15 October 2015 by Teddisque (Talk | contribs)

Jump to: navigation, search

This is an article review of the 2013 article by Chih-Wen Cheng et. al. entitled icuARM - An ICU Clinical Decision Support System Using Association Rule Mining [1]

Introduction

To meet the specific Clinical Decision Support needs of critical care clinicians, the authors developed a ICU CDSS which incorporated a GUI to assist physicians and other clinicians with data mining in real time in a patient specific way.

Methods

Authors used the MIMIC-II database to build their rules. MIMIC-II is a database containing clinical and physiologic data from tens of thousands of ICU patients. They then used Association Rule mining to develop the CDSS logic and algorithms.

Results

The outcome was a rule mining application which can be accessed in real time in the ICU, physicians can select antecedents and consecuents, find an existing rule in the database and determine the confidence level of the set of antecedents. This is essentially creating a cause and effect database in real time, and applying it to current patients to determine likely outcomes or consequences of a condition or set of conditions. As users continue to browse and create new rules, the rule data bank will grow.

Conclusion

This stand-alone CDSS access a huge amount of raw data, uses the data to develop evidence in real time and applies that evidence to guide a clinicians decisions.

Comments

Being able to apply actual physiologic and clinical data to a decision in real time is an amazing use of biomedical informatics. I can see this being most useful in guiding drug choices, for example. If a physician wants to know if Epinephrine or Levophed use is associated to better outcomes in a hypotensive patient post gastric bypass surgery who has a history of Sjogren's syndrome he or she has no time to go review literature. Assuming any such studies exist. I can see the next step for a system like this would be to integrate it into an EMR and make the interface contextual and able to pull real time patient data into the interface without an end-user having to manually enter it.

References

  1. Chih-Wen, C., Chanani, N., Venugopalan, J., Maher, K., & Wang, M. D. (2013). icuARM-An ICU Clinical Decision Support System Using Association Rule Mining. Translational Engineering in Health and Medicine, IEEE Journal of