Automated surveillance: Incidence of bloodstream infections related to central venous lines (completed)

The aim of this study is to develop methodology to extract information from EMR to enable reporting of process indicators related to insertion, control and removal of central venous catheter (CVC) and to determine the incidence of CVC-related bloodstream infections (CVC-BSI).

About the project

This project represents one of 7 work packages of the project EviCare.The objective of EviCare is to develop methods and technology for providing «Evidence-Based Medicine» (EBM) at the point of care, integrated with an electronic health record (EHR) or other health information systems directly involved in the clinical process, resulting in higher quality of care and a more detailed, transparent documentation of care processes.  In work package 6 the superior objective is development and evaluation of new guideline-oriented quality indicators.

Outcomes

The first step in this project is development of natural language processing technology to mine electronic health records (EHR) for triggers to detect the numerators and denominators of incidence of CVC-BSI: number of CVC-BSI/CVC days (date of CVC removal – date of CVC insertion). As a step 2 the instrument/tool will be used to identify CVC-BSI in a pilot material of 1000 EHRs. Finally the instrument/tool will be tested in hospital settings, and implementation of real-time reports on incidence of CVC-BSI will be evaluated in a case- control study.

Background

Bloodstream infections (BSI) are frequent and severe complications, and are often related to treatment with central venous catheter (CVC). The use of central lines in medical treatment is indispensable for many patients, but is also exposing them for risk of infection and consequently increased morbidity and mortality. From studies in ICUs we know the potential to prevent CVC-BSI is quite large, and the need of surveillance and control with central venous catheters is pointed out.

Surveillance studies of CVC related BSI (CVC-BSI) is most often performed in intensive care units (ICU). The sparse amount of studies performed in general wards is probably explained by lack of resources and adequate systems to monitor CVC and bloodstream infections. However, nationally and internationally, infection control professionals seem to agree about the need of knowledge on this matter in general wards. From the few studies made, incidence of CVC-BSI in general wards varies from 4 to 8 CVC-BSI per 1000 intravascular device days.

Suspecting that a large amount of CVC-days in a hospital appear outside the ICU, there is a pressing need to decide the extent of catheter-use and related complications in general wards. Applying manual nosocomial infection surveillance is not only time-consuming and expensive, but makes a risk for misclassification that electronic surveillance would handle. Most systems to detect adverse events require clinical data in coded format, but in the EHR physicians usually describe the present situation in a comprehensive and short free text summary. Natural language processing has in several studies been tested as a method to extract information about adverse events from free text. The technique allows automated conversion of free text information into a coded form, and reduces the need of additional documentation in structured fields

Financing

  • The Research Council of Norway

Cooperation

  • Innlandet Health Trust
  • The Norwegian Knowledge Centre for Health Services/Avdeling for kvalitetsmåling og pasientsikkerhet
  • DIPS ASA (Clinical information systems)
  • Datakvalitet AS (Content management systems)
  • Norwegian University of Science and Technology / Department of Computer and Information Science

Start - Finish

01.01.2010 – 15.09.2013

Published May 21, 2011 6:20 PM - Last modified Feb. 17, 2020 10:08 AM

Contact

Project leader

Geir Bukholm

PhD-candidate

Christine Tvedt
 

Participants

Detailed list of participants