CHICAGO – In as little as 15 seconds, an automated electronic medical record search application can retrieve vital past medical history for patients presenting to the emergency department.
"We believe this kind of innovative interface allows for a more efficient view of the entire patient history in the electronic health record," Dr. Arun Krishnaraj said at the annual meeting of the Radiological Society of North America.
The researchers also hope that the novel search tool, known as QPID (Queriable Patient Inference Dossier), will help reduce inappropriate imaging in the emergency department (ED), where incomplete awareness of past imaging studies often leads to duplication.
Dr. Arun Krishnaraj
QPID is a programmable, ontology-driven semantic search application that extracts data from multiple data repositories and then indexes or prepares that information for a search, explained Dr. Krishnaraj of Massachusetts General Hospital and Harvard Medical School in Boston. It gathers data through Web services available in Harvard’s network and can be automated to run against a service schedule or care unit census. It is not a data repository nor does it store personal health information.
QPID goes beyond the simple key word search by allowing a clinician to use a variety of natural language expressions to perform Boolean searches for two terms such as "hepatitis and cirrhosis"; negate a target phrase to exclude unwanted hits, for example, "the patient does not have hepatitis"; and to find exact or partial matches for acronyms, synonyms, or misspellings such as "hepatitis/heputytus," he said.
The three distinct advantages of QPID over a traditional manual record search is that it can search for a concept such as "Does the patient have a malignancy?"; it enables complex structured queries to be run automatically; and it integrates the search output into a web browser or office application, Dr. Krishnaraj said.
The automated queries can be retrieved with a simple search string such as "*malignancy." QPID would then search for evidence of malignancy including synonyms such as tumor, mass, and neoplasm. It will exclude the term "mass" when used in other ways, as in "Mass." for Massachusetts, while also matching exact acronyms for types of malignancy such as NSCLC (non–small cell lung cancer).
QPID has been live for about 6 months at the Massachusetts General ED and about 1 month at the Brigham and Women’s Hospital ED, Dr. Krishnaraj said in an interview. It is available as a search platform to any clinician within Massachusetts General, with custom interfaces developed for several divisions including radiology and gastroenterology, among others.
The researchers validated QPID by performing an automated search for each of the 74 topics included in the application for 500 consecutive patients who presented to the hospital’s ED in 2010. The automated results were then compared with those from two clinicians who performed an untimed manual review for the same 74 search topics on 30 randomly selected patients in the cohort.
"We believe this kind of innovative interface allows for a more efficient view of the entire patient history in the electronic health record."
The average search time for QPID to research all 74 topics was 15 seconds, plus or minus 5 seconds, Dr. Krishnaraj said. To complete a thorough review of all available data, the manual review averaged 5-10 minutes per patient.
For finding laboratory results, QPID demonstrated a sensitivity of 97% and specificity of 99%, with a positive predictive value (PPV) of 99% and negative predictive value (NPV) of 96%.
Excellent results were also seen for free text searches, such as "is there a history of PE in the last 10 years?" For these, QPID had a sensitivity of 98%, specificity of 93%, PPV of 90%, and NPV of 98%.
Dr. Krishnaraj acknowledged that the overall results could be affected by data in the EMR not detected by either QPID or manual review. Other limitations include the potential for false positives, overestimation of low-prevalence conditions, and missing data for care received outside the hospital network and not recorded electronically.
The researchers are currently evaluating how best to measure the effect of QPID use on imaging use in the two hospitals’ large and busy EDs.
Dr. Krishnaraj and a coauthor reported research support/grants from General Electric.