kpozin.net Projects & Musings by Konstantin Pozin

Research for the 2006 Intel Science Talent Search and NYC Science & Engineering Fair
Carried out at Memorial Sloan-Kettering Cancer Center, primarily in the summer of 2005.

Abstract:

Studies have shown that most alarms generated by physiological monitoring systems in modern hospitals are false or clinically irrelevant. This makes it difficult for nurses to pick out and respond to the relevant alarms, reducing efficiency and posing a potential risk to patients.

This project explored methods for reducing alarms at a metropolitan hospital, using data from its twelve-bed Intensive Care Unit. Several programs were written to collect real-time waveform, vital sign, and alarm data over the hospital’s GE Medical monitoring network, and to decode and display it for analysis. An algorithm was created to eliminate redundant notifications, reducing the overall volume of alarm data by 70 percent versus the hospital’s current alarm paging system. Algorithms were created to filter out several sources of false alarms by detecting data artifacts and by comparing vital sign information from multiple devices (ECG, pulse oximeter, arterial pressure probe, etc.). Prospective additional methods for reducing false alarms were also proposed.

This research provided effective methods for gathering medical data for analysis, and for using the data to decrease the flow of redundant and false alarm information. The work serves as a proof-of-concept, with the potential for implementation in future monitoring systems.