November always brings to mind Thanksgiving and turkey, but for those of us in medical informatics it means it’s time for AMIA’s national symposium and this time it is back in Washington DC. With political upheaval in healthcare and the opioid epidemic making headlines across the nation there will be no shortage of talking points. AMIA brings together some of the best and the brightest minds in medical informatics and a great place to engage with each other to highlight opportunities for IT to improve the lives of Americans with technology.
This will be my fifth time going to AMIA, a relative novice compared to many, but I find the scale of this event much more palatable than the behemoth that is HIMSS, with a more forward-looking vibe. Linguamatics have many events planned for AMIA this year, including a pre-symposium talk, a presentation on physician metrics by MUSC and a Learning Showcase presentation. Check out the details below and stop by and see us at booth #205.
Presented by Erin Tavano, Clinical Data Scientist, Linguamatics
8:30AM – 4:30PM, Saturday November 4, 2017, Georgetown East
Linguamatics is presenting at the NLP symposium and presenting a poster on our analysis of opioid risk factors using NLP. Erin Tavano, a clinical data scientist for Linguamatics has taken standard models of risk and incorporated those into a series of queries to give a risk score. These factors include alcohol abuse, sexual abuse and are based on the Opioid Risk Tool (ORT). This shows how at-risk patients can be flagged before addiction is an issue at the population and real-time scale
Presented by Vivienne Zhu, Assistant Professor, MUSC
4:42PM – 5:00PM, Monday November 6, 2017, Jefferson West
Physician quality reporting that relies on coded administrative data alone may not completely and accurately depict providers’ performance. To assess this concern with a test case, MUSC developed and evaluated a natural language processing (NLP) approach to identify falls risk screenings documented in clinical notes of patients without coded falls risk screening data. Extracting information from 1,558 clinical notes (mainly progress notes) from 144 eligible patients, they generated a lexicon of 38 keywords relevant to falls risk screening, 26 terms for pre-negation, and 35 terms for post-negation. The NLP algorithm identified 62 (out of the 144) patients whose falls risk screening were documented only in clinical notes and not coded. Manual review confirmed 59 patients as true positives and 77 patients as true negatives. Their I2E NLP approach scored 0.92 for precision, 0.95 for recall, and 0.93 for F-measure. These results support the concept of utilizing NLP to enhance healthcare quality reporting.
Presented by Simon Beaulah, Director, Healthcare, Linguamatics
1:15PM – 1:35PM, Tuesday November 7, 2017, Columbia Hall
The growing availability of clinical data means there is a wealth of opportunity to improve disease understanding and outcomes, through research and quality improvement methodologies. But with 80% of the richest insights trapped in unstructured text, AI techniques like NLP are a critical piece of functionality in any medical center. This presentation will look at the key capabilities for NLP and how interoperability and democratization of NLP functionality is vital in the following application areas:
- Automated extraction of cancer insights from pathology reports for cancer registries, bio-specimen and research data warehouses
- Exploration of large clinical data sets to support machine learning algorithm development, such as for opioid abuse prediction
- Analysis of phenotypic characteristics in clinical data to support phenotype/genotype analysis
- Identification of clinical trials candidates
- Advanced searching of scientific literature for rare disease associations with genetic variants