EHR 'search engine' could optimize emergency department imaging

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - EHR Round Up

Targeted summaries of data contained in the EHR generated by queries written for a programmable health record intelligence system are a rapid and accurate way for clinicians to gather relevant medical histories on patients, according to an article published in the June issue of the Journal of the American College of Radiology.

Given the urgent nature of the emergency department (ED) setting, having all relevant medical information about a patient is necessary, particularly in complex scenarios. Although EHR systems include information such as laboratory tests, pathology reports, diagnostic studies, clinical notes and demographic information, much of this data is stored as free-text documents. This makes efficient access and review difficult for care providers. As a response to this issue, a team of clinicians and software developers in the informatics division for the department of radiology at Massachusetts General Hospital in Boston developed a search engine for the EHR known as Queriable Patient Inference Dossier (QPID).

“If one considers an essential component of an EHR to be its data repository, then QPID is a programmable system that facilitates the extraction of information from that repository,” wrote the article’s lead author, Arun Krishnaraj, MD, MPH, of the University of Virginia in Charlottesville, and colleagues. The authors sought to validate the queries from the QPID application that expedites summarization of medical history relevant to initial screening and management decisions and was developed specifically for the ED.

The custom QPID ED application included a set of 74 query topics important for screening and management by ED physicians. The study cohort included 500 consecutive adult patients whose clinical documents containing both structured data and unstructured data in the EHR were reviewed. A tool that performed an automated QPID search on all 74 queries allowed for performance validation of each.

Results revealed a mean search time of 15 seconds for the application to complete all 74 searches on each patient. Notable searches that directly impact imaging utilization in the ED included deep vein thrombosis in the past five years, pulmonary embolus in the past ten years, evidence of prior ectopic pregnancy, low ejection fraction, mechanical valve, coagulation parameters and the presence of an automatic implantable cardioverter-defibrillator.

QPID queries for structured data exhibited a pooled positive predictive value (PPV) of 87 percent and a negative predictive value (NPV) of 86 percent. For unstructured data, the QPID queries demonstrated a pooled calculated PPV of 75 percent and an NPV of 88 percent.

The calculated free marginalκcoefficient for interrater reliability between the two clinicians whose search results served as the gold standard was 0.87.

“The present work is a necessary, preliminary step toward demonstrating that the combination of concept specification by clinical users, and the capability to handle heterogeneous (structured and unstructured) input EHR data may facilitate review of salient clinical data on the often complex patients who present to the ED,” wrote Krishnaraj and colleagues. “We hope to demonstrate in future studies that this tool will influence the rate and appropriateness of imaging utilization as well as other health care resources.”