Event Title

Automated Adverse Respiratory Event Detection in Volunteers Receiving Opiod Analgesics

Location

Room # EB402

Start Date

5-6-2019 10:00 AM

Description

Introduction: The underlying problem for two of the three most common patterns of unexpected hospital deaths (PUHD) is hypoventilation1. Concern over this opioid-induced respiratory depression has led many experts and consensus guidelines to recommend that all patients receiving opioids be monitored for respiration. We propose an additional metric, ataxic breathing severity, be monitored in patients receiving opioids. Methods: With IRB approval, data were collected from 26 volunteers who were administered target controlled infusions of remifentanil and propofol in order to induce low respiratory rates. Data were collected from a suite of sensors which were analyzed using a single, custom breath detection algorithm. Three domain experts rated the data according to degree of irregularity, or ataxia, in the respiratory waveform. A machine learning algorithm was then trained to reproduce those results. Results: Interrater reliability analysis confirmed that a machine learning algorithm was capable of mimicking scores from domain experts with a high degree of accuracy. Conclusion: Detecting ataxic, or irregular, respiratory rate is possible using an automated detection system.

Comments

Session 1

This document is currently not available here.

Share

COinS
 
May 6th, 10:00 AM

Automated Adverse Respiratory Event Detection in Volunteers Receiving Opiod Analgesics

Room # EB402

Introduction: The underlying problem for two of the three most common patterns of unexpected hospital deaths (PUHD) is hypoventilation1. Concern over this opioid-induced respiratory depression has led many experts and consensus guidelines to recommend that all patients receiving opioids be monitored for respiration. We propose an additional metric, ataxic breathing severity, be monitored in patients receiving opioids. Methods: With IRB approval, data were collected from 26 volunteers who were administered target controlled infusions of remifentanil and propofol in order to induce low respiratory rates. Data were collected from a suite of sensors which were analyzed using a single, custom breath detection algorithm. Three domain experts rated the data according to degree of irregularity, or ataxia, in the respiratory waveform. A machine learning algorithm was then trained to reproduce those results. Results: Interrater reliability analysis confirmed that a machine learning algorithm was capable of mimicking scores from domain experts with a high degree of accuracy. Conclusion: Detecting ataxic, or irregular, respiratory rate is possible using an automated detection system.