Project Details - Ended
- Grant Number:R01 HS022728
- Funding Mechanism:
- AHRQ Funded Amount:$1,959,741
- Principal Investigator:
- Project Dates:9/30/2013 to 9/29/2018
- Care Setting:
- Medical Condition:
- Type of Care:
- Health Care Theme:
Allergies are linked to an important group of chronic and serious illnesses such as asthma, sinusitis, hay fever, and atopic eczema. For some people, allergic reactions can be severe and even fatal. As such, obtaining allergy information is critical to safe prescribing, preventing adverse drug events, and reducing the cost of care. Clinicians routinely elicit allergy information during the medical interview; however, allergies are often poorly documented in the electronic health record (EHR). The resulting information is neither interoperable across clinical information systems nor easily reusable for other applications.
The research team developed a natural language processing (NLP) system for allergy information and evaluated whether it improved patients’ allergy lists. Free-text allergy entries in the EHR were mapped to standard terminologies, with qualitative metrics used to assess mapping accuracy. Next, they developed an allergy information extraction and encoding module that was integrated with the Medical Text Extraction Reasoning and Mapping System (MTERMS), an NLP tool previously created by the team. The NLP module was then used to identify allergy information from free-text clinical notes.
The specific aims of the research were as follows:
- Build a comprehensive knowledge base for allergy information.
- Develop and evaluate an NLP system for processing allergy information.
- Use NLP output to facilitate allergy reconciliation.
- Disseminate the methods used and the resulting tool.
The evaluation found that 96.7 percent of allergens in free-text notes were identified by the NLP module, compared to a manual chart review. A standalone web application was then developed to batch-process notes, display allergy information, and compare the extracted allergies with those in patients’ structured allergy lists. Finally, data from the NLP module were used to describe the epidemiology of drug allergies, food allergies, and allergic reactions using longitudinal EHR data. The team widely disseminated their findings, including descriptive studies about allergies in 28 publications.