QPID Featured at the AMIA Annual Meeting – Studies Show Impact of QPID

We were pleased to rub shoulders with our colleagues and customers at last week’s AMIA (American Medical Informatics Association) annual meeting. Of note were posters presented by two teams, one led by Gaurav Singal and the other by Shann-Chirag Gandhi. The teams presented case studies of QPID at work to support clinicians with patient information extracted from medical records.

In QPIDMed: A Search-Driven Automated Chart Biopsy Dashboard, Dr. Gaurav Singal and his research partners at the Massachusetts General Hospital (MGH) described the QPID EHR and Medicine Portal applications.  The problem with the modern-day EHR, simply put, is that critical information is hard to find. As the poster aptly summarizes:

Critical information is frequently buried beneath volumes of data, frequently leading to errors in patient management.

The authors provide several clinical scenarios in which simply storing data in an EHR has failed, including “a patient who presented with a fracture bleeding profusely because of chronic anticoagulation – with notes documenting warfarin should have been stopped years ago.”  The problem is not only the volume of data, we believe, but that notes such as this one are found in unstructured fields that are laborious to comb through without a smart NLP-based search engine such as QPID.

Queries can be developed within QPID that can identify clinical concepts, not just keywords. QPID queries showed on average over 90% sensitivity and 99% specificity, according to the authors, exceeding performance of simple keyword searches.

The Medicine Portal (or what the authors’ cleverly call an automated “chart biopsy” dashboard) was deployed to the MGH Department of Medicine in September 2012. Usage is heaviest at patient admission, but extends into a “long tail” of many days into hospitalization.

The authors plan to study nuances of usage to guide development of additional “contextually aware” dashboards such as those optimized for rounds, floors, the ICU and so forth.  Further, they plan to study impact on clinical outcomes and test ordering.

Shann-Chirag Gandhi of the MGH and his co-authors including Dr. Arun Krishnaraj who is now at the University of Virginia, presented their findings in the Design and Validation of Automated, Customized Clinical History Searches for Imaging Interpretation.  The problem is that radiologists often receive only a brief clinical summary with an imaging request, making interpretation less than optimal.

Detailed knowledge of a patient’s past medical history can be crucial for optimal study interpretation. Manual electronic record searching is time-consuming and can lead to potentially lower-quality, less-efficient interpretations.

The authors used QPID to extract information pertinent to the interpretation of liver, prostate, and rectal MRIs and display that information in an organized graphical interface which included interpretation guidelines.

Patient historical data that is pertinent to the radiologist can be “distilled and converted” to fast, accurate search queries using QPID. For example, in this case, queries designed for magnetic resonance (MR) imaging studies encompassed relevant imaging studies, laboratory/pathology results, medication and allergy lists; admission, progress, operative/procedural and discharge notes; as well as other unstructured/structured data.

For the study, nine MR search algorithms related to liver, prostate and rectum exams were run on a test sample of 20 patient records.  High levels of Positive Predictive Value (0.86 +/- 0.02) and Negative Predictive Value (0.91 +/- 0.01 were reported.

We are looking forward to follow-up research from our esteemed customers.