StarTracker: An Integrated, Web-based Clinical Search Engine

Abstract

This poster details the design and use of the StarTracker clinical search engine. This program is fully integrated within our electronic medical record system and allows users to enter simple rules that direct formatted searches of multiple legacy databases.

Background

Over the past several years, the development of a comprehensive electronic clinical record and communication system (StarPanel) at VUMC has largely removed paper based processes in nearly 95% of our clinics. One of the central concepts of this system is the “panel”, a population level view of patient care that enables clinicians to perform actions at a higher level than the individual patient. This view of patient information allows unique approaches to common information problems.

VUMC is a busy, tertiary care medical center with more than 700,000 annual outpatient visits. In this setting, there is a great demand for population level data for the clinician interested in outcomes and the clinical researcher developing study cohorts. The ultimate goal is an integrated searching capability that allows users to define patient groups based upon demographic, clinical and diagnostic data. When possible, the development of an enterprise-level data warehouse is an ideal solution¹, but requires substantial resources and time. An alternative solution that works with multiple legacy systems, without requiring prior integration of those databases, may simply and inexpensively provide significant added value.

Methods/Design

Our solution is a web-based search engine fully embedded in our existing electronic medical record system. This engine allows clinicians and/or researchers to search panels (populations) of patients for specific diagnostic, demographic and clinical criteria. Privacy concerns are addressed through limitations on panel access. The search engine, part of a disease management system known as StarTracker, accesses data from three separate databases, submitting specific queries to each, but requiring the user to enter only a single, simple query.

Summary of search engine use:

1. The user selects a panel to search and enters search rules based upon demographic, diagnostic (ICD-9 codes) and/or specific laboratory criteria (e.g. value and date ranges).

2. The user creates a simple query using these previously defined rules in a format similar to bibliographic tools. This allows an arbitrary level of complexity and does not require training for the user beyond simple instructions.

   *Example query format for five arbitrary rules: (1 or 2 or 3) AND (4 or 2) AND (5)

3. The system parses this query and, as needed, sends separate, formatted queries to the primary clinical database, the administrative database of diagnosis codes, and the patient demographics database.

4. The results of these individual queries are assembled and the list of patients meeting the criteria is returned to the user.

RESULTS

The search engine is currently available to approximately 150 pilot users at VUMC and has been used successfully in the primary care and sub-specialty settings for two primary purposes:

1. Establishing cohorts for IRB approved studies.

2. Physician directed evaluation of clinical outcomes.

CONCLUSION

The development of an easily accessible enterprise data warehouse is usually a long-term goal. The needs of clinicians and researchers are often more immediate. The StarTracker search engine offers a comprehensive, population level search capability using existing hardware and database systems.