197. Development of a Patient Identification System for Sepsis within a Real-Time Clinical Surveillance Tool
Session: Poster Abstract Session: Antimicrobial Stewardship: Current State and Future Opportunities
Thursday, October 8, 2015
Room: Poster Hall
Posters
  • 197_IDWPOSTER.pdf (110.2 kB)
  • Background:  Timely management of sepsis improves patient outcomes, including mortality. Computerized analytics holds promise for early identification of septic patients to decrease time to appropriate therapy.  A Real-Time Clinical Surveillance Tool (RTCST) is a clinical decision support (CDS) tool that analyzes multiple data streams to highlight opportunities in individual patients.  This study aims to describe development of a CDS for early sepsis identification.

    Methods: Five RTCST rules for identification of septic patients were developed and deployed at four large medical centers during December 2014 and April 2015.  Rule #1: alerted to 2+ systemic inflammatory response syndrome (SIRS) criteria within 4 hours.  Rule #2: alerted to 3+ SIRS criteria over 24 hours.  Rule #3: alerted to lactic acidosis PLUS 1+ SIRS criteria, antibiotic order, vasopressor order, or Mean Arterial Pressure (MAP) < 65 mmHg.  Rule #4: alerted for ONE of the following: hyperlactatemia, low oxygen saturation, or hypotension PLUS 2+ SIRS criteria with temperature defined as >101 ºF.   Rule #5: alerted for 2+ SIRS criteria PLUS an antibiotic order, positive blood/urine/sputum culture, lactic acidosis, or Shock Index >/=0.7. Alerts of individual RTCST rules were compared to patients with an ICD-9 diagnosis code for sepsis and analyzed for selectivity, specificity, positive predictive value and negative predictive value.

    Results: A total of 11,534 patients and 5,478 RTCST alerts were analyzed, with 513 patients diagnosed for sepsis.  Diagnosis codes for sepsis were used for benchmarking.  

    Rule

    Sensitivity

    Specificity

    Positive Predictive Value

    Negative Predictive Value

    #1

    72.6%

    78.4%

    9.5%

    98.9%

    #2

    58.9%

    87.3%

    13.9%

    98.4%

    #3

    18.4%

    93.8%

    50.0%

    77.5%

    #4

    49.2%

    93.7%

    27.4%

    97.4%

    #5

    53.3%

    85.9%

    29.5%

    94.3%

    Conclusion:  Current RTCST rules have shown a high negative predictive value for septic patients; however additional revisions are needed to improve the positive predictive value.  The difficulty and responsiveness inherent in the manual process of identifying septic patients highlights the potential for electronic patient identification.

    Mandelin Cooper, Pharm.D., BCPS1, Joel Daniel, MS, Pharm.D.2, Karla Miller, Pharm.D., BCPP2 and Hayley Burgess, Pharm.D.2, (1)Division Pharmacy, HCA Continental Division; Parallon Supply Chain Solutions, Denver, CO, (2)Hospital Corporation of America, Nashville, TN

    Disclosures:

    M. Cooper, None

    J. Daniel, None

    K. Miller, None

    H. Burgess, None

    Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 7th with the exception of research findings presented at the IDWeek press conferences.