Using Voogle to Search Within Patient Records in the VA Corporate Data Warehouse

Search

The introduction of search engines can be considered one of the major technologic disruptors of the 21st century.⁸ However, this advance has not yet made significant inroads into health care, despite advances in other domains. As of 2019, EHR users are still required to be familiar with the system’s data and menu structure in order to find needed information (or enter orders, code visits, or any of a number of tasks). Anecdotally, one of the authors (David Eibling) observed that the most common question from his trainees is “How do you . . .?” referring not to the care of the patient but rather to interaction with the EHR.

What is needed is a simple query-based application that finds the data on request. In addition to Voogle, other advances are being made in this arena such as the EMERSE, medical record search engine (project-emerse.org). Voogle was released to VA providers in 2017 and is available through the Internet Explorer browser on VA computers with VA intranet access. The goal of Voogle is to reduce HCP cognitive load by reducing the time and effort needed to seek relevant information for the care of a specific patient.

Natural Language Processing

Linguistic analysis of text seeking to understand its meaning constitutes a rapidly expanding field, with current heavy emphasis on the role of artificial intelligence and machine learning.¹ Advances in processing both structured data and free-text notes in the health care domain is in its infancy, despite the investment of considerable resources. Undoubtedly, advances in this arena will dramatically change provider cognitive work in the next decades.

VistA is coded in MUMPS (Massachusetts General Hospital Utility Multi-Programming System, also known as M), which has been in use for more than 50 years. Voogle employs iKnow, a novel natural language processing (NLP) application that resides in Caché (Intersystems, Boston, MA), the vendor-supported MUMPS infrastructure VistA uses to perform text analysis. iKnow does not attempt to interpret the meaning of text as do other common NLP applications, but instead relies on the expert user to interpret the meaning of the analyzed text. iKnow initially divides sentences into relations (usually verbs) and concepts, and then generates an index of these entities. The efficiency of iKnow results in very rapid indexing—often several thousand notes (not an uncommon number) can be indexed in 20 to 30 seconds. iKnow responds to a user query by searching for specific terms or similar terms within the indexed text, and then displays these terms within the original source documents, similar to well-known commercial search engines. Structured data are indexed by the iFind program simultaneously with free-text indexing (Figure 1).

Security

Maintaining high levels of security of Health Insurance Portability and Accountability (HIPAA)-compliant information in an online application such as Voogle is critical to ensure trust of veterans and HCPs. All patient data accessed by Voogle reside within the secure firewall-protected VINCI environment. All moving information is protected with high-level encryption protocols (transport layer security [TLS]), and data at rest are also encrypted. As the application is online, no data are stored on the accessing device. Voogle uses a secure Microsoft Windows logon using VA Active Directory coupled with VistA authorization to regulate who can see the data and use the application. All access is audited, not only for “sensitive patients,” but also for specific data types. Users are reminded of this Voogle attribute on the home screen.