Increasing vigilance on the medical/surgical floor to improve patient safety


  • Joshua L. Jacobs,

    1. Joshua L. Jacobs MD Associate Professor Division of Medical Informatics, Department of Medicine, John A. Burns School of Medicine, University of Hawaii, Honolulu, Hawaii, USA
    Search for more papers by this author
  • Nathaniel Apatov,

    1. Nathaniel Apatov PhD CRNA Director US Army Graduate Program in Anesthesia Nursing (MCHK-HE) Phase II, Tripler Army Medical Center, Honolulu, Hawaii, USA
    Search for more papers by this author
  • Matthew Glei

    1. Matthew Glei Executive Vice President Hoana Medical, Inc., Honolulu, Hawaii, USA
    Search for more papers by this author

Joshua L. Jacobs:


Aim.  This paper reports a study designed to assess an automated non-invasive, patient vigilance system, the LG1TM system, for determining heart rate and respiration rate. The study uses collected data to optimize the LG1TM’s alert management scheme for medical/surgical wards.

Background.  Thousands of patients die unnecessarily each year because of compromised patient safety in hospitals. Economic pressures to reduce hospitalization costs, exacerbated by increasing nursing shortages, have created a need for new approaches to patient vigilance. Advanced technologies may help nurses to provide high-quality care while controlling costs and improving patient safety.

Methods.  Heart and respiration waveforms from 287 patients were captured by sensor arrays embedded in the mattress coverlets of their beds. No real-time monitoring was performed. Raw data were processed by proprietary algorithms and compared with data captured by a standard reference device. Alert performance was verified by hand-scoring the signal data and matching it against clinical events observed through a systematic review of each patient's medical record. The data were collected between June 2004 and February 2005.

Results.  Experimental algorithms for heart rate had an accuracy of −1·47 (sd 1·90) and a precision of 4·60 (sd 2·46). Respiration rate algorithms showed an accuracy of −0·94 (sd 1·26) and a precision of 4·02 (sd 1·17). Algorithms identified 178 true-positive physiological alerts on 15 patients. None of the events was deemed clinically significant at chart review. The combined false-positive alert rate for the algorithms was 0·007 events per hour.

Conclusion.  This study demonstrates the accuracy and precision of the signal processing algorithms in the LG1TM system. Future work will focus on assessing the system's impact on patient outcomes and its integration into the nursing workflow.