Real-Time Automated Surveillance for Ventilator Associated Events Using Streaming Electronic Health Data
Ryan, Erin E
Valdery, Moura Junior
Westover, M BrandonNote: Order does not necessarily reflect citation order of authors.
MetadataShow full item record
CitationShenoy, E. S., E. S. Rosenthal, S. Biswal, M. Ghanta, E. E. Ryan, Y. Shao, D. Suslak, et al. 2017. “Real-Time Automated Surveillance for Ventilator Associated Events Using Streaming Electronic Health Data.” Open Forum Infectious Diseases 4 (Suppl 1): S633. doi:10.1093/ofid/ofx163.1681. http://dx.doi.org/10.1093/ofid/ofx163.1681.
AbstractAbstract Background: Criteria defining Ventilator Associated Events (VAEs) are objective and often available in the electronic health record (EHR) data. The use of ventilation data extracted directly from the patient’s bedside monitor to allow for real-time surveillance, however, has not been previously incorporated into electronic surveillance approaches. Here we describe validation of a system that can detect and report on VAEs hospital-wide autonomously and in real-time. Methods: We developed a secure informatics hardware and software platform to identify VAEs autonomously using streaming data. The automated process included 1) archiving and analysis of bedside physiologic monitor data to detect increases in positive end-expiratory pressure (PEEP) or FiO2 settings; 2) real-time querying of EHR data for leukopenia or leukocytosis and concurrent antibiotic initiation; and 3) retrieval and interpretation of microbiology reports for the presence of respiratory pathogens. The algorithm was validated on two 3-month periods in 2015 and 2016 as follows: 1) autonomous surveillance (AS) generated detections of three VAE subclasses: VAC, IVAC, and PVAP; 2) manual surveillance (MS) by Infection Control (IC) staff independently performed standard surveillance based on chart review, 3) senior IC staff adjudicated the gold standard for cases of AS-MS discordance. The sensitivity (Se), specificity (Sp), and positive predictive value (PPV) of the algorithm are reported. Results: The number of ventilated patients, ventilator days, and events were: 1,591/9,407/3,014. In cases with complete data, AS detected 66 VAE events identified by MS; AS detected 32 VAEs missed by MS; no MS-identified events were missed by AS. The Se, Sp, and PPV of AS and MS were: 91%/100%/100%, and 61%/100%/83%, respectively. Clinical surveillance case reports generated by AS enabled visual interpretation (figure). Conclusion: We developed a surveillance tool directly streaming bedside physiologic monitor and EHR data including ventilator settings, laboratory results, and microbiology reports, to apply the CDC’s VAE definitions on source data. This resulted in an accurate, objective, and efficient method for real-time hospital-wide surveillance. Disclosures All authors: No reported disclosures.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:34493170
- HMS Scholarly Articles