EpiPortal: An Electronic Decision Support System for Infection Control

Abstract

A comprehensive, electronic hospital epidemiology decision support system serves diverse users but its primary user is the infection control professional (ICP). Utilizing off-the-shelf components and accepted standards enables the system to be open, vendor-independent and ICP-controlled. Its development can flexibly respond to the evolving nature of infection control practice.

OBJECTIVES

New York Presbyterian Hospital (NYPH) has developed an epidemiology decision support system we call “EpiPortal”. It is a framework for electronic, data-driven functionalities that contains a continually-evolving collection of secure, interrelated, web-linked modules geared towards enhancing the workflow of diverse user groups, but especially that of ICPs. Its design requires that it be (1) adaptable to the continually changing science of infection control and varying local practices, (2) built with standardized and widely-available components and (3) able to process electronic source evidence for all microorganism-related disease, not just culture and sensitivity procedures. As a corollary of (1), all definitions, rules, and workflow representations are formulated and thoroughly vetted by ICPs themselves. “Black-box” or inadequately validated vendor-supplied algorithms are not acceptable.

METHODS

Observations of daily ICP workflow were performed. Rules and user interfaces are examined with informatics staff for computational feasibility. Informatics representatives participate in infection control working meetings as well as the Hospital’s Infection Control Committee. Specially trained administrators serve as project managers to facilitate communication between Infection Control and developers. EpiPortal’s primary data source is the Hospital’s clinical data warehouse, an HL7-based database that integrates multiple transactional source systems. Online analytic processing (OLAP) using Microsoft’s Analysis Services and Panorama is the primary method for analyses/reports of incidence rates, microbial susceptibility patterns, and antibiotic utilization. The core architecture generates infection episodes, active phases and worklists (Figure 1a), and is extensible to OLAP analyses (Figure 1b) and other user-requested functionalities (Figure 2).

Figure 1. EpiPortal Architecture.

Boxes: program modules; Italics: user-supplied knowledge; Unboxed Text: database tables and views.

Figure 1a: Core Architecture; Figure 1b: Architecture Extension. For incidence/prevakence rate views.

Figure 2. EpiPortal User Options.

Options customizable for different user groups

RESULTS

Initially implemented on NYPH’s Columbia University campus, the system is being deployed to NYPH’s Weill-Cornell campus. Real-time alerts are generated with a separate HL7-based clinical alert monitor¹. Natural language processing rapidly delivers selected radiology results². CLSI M39-A compliant antibiograms³ that previously took 2 person-months to develop are now interactive, up-to-date and can be generated in hours. Daily worklists facilitate ICP management of infections caused by selected organisms such as VRE, MRSA, C. difficile. Worklists have increased detection in readmitted patients of VRE by 27% and of MRSA by 25%. System development continues.

CONCLUSIONS

By requiring ICP workflow be the central focus and the technical architecture use standardized components, the design objectives of EpiPortal demand considerable up-front investment. However, the investment enables a broad, flexible, and robust infection control decision support system.