Librarians Connecting EHR Data – a paper presented at MLAQUAD 2012

I had a fantastic time at MLAQUAD in Baltimore.  It was great to see some old friends and meet some new ones.  I had crab cakes every day, and they were all good.  There were a couple of queries about my presentation, plus some interest from a couple of people on #medlibs, so here it is.  It is very much a work in progress so any suggestions are welcome.

This link goes to my slide presentation in SlideShare: ConnectingEHRdatamlaquad12 from Margaret Henderson

And here is my script, also in SlideShare. This presentation was my first paper at a meeting.  I’ve done quite a few posters, and a couple of lightning posters ( 5 minute talks), but this was new.  I also had lost most of my voice due to a cold so I couldn’t actually speak my presentation aloud to practice it.  So I thought the best way to get my timing right was to write it all out.  In future, I will only note the things I really want to remember, like names, and then let the rest just flow, since I generally have no trouble babbling on about things I’m interested in.

Librarians Connecting EHR Data

  1. Good morning everyone.  I would like to share with you a new project I am working on – using the data from our hospital EHR system as part of the curriculum of our medical school.
  2. With all the concern about new librarian roles for the future and how we can embed librarians, I thought I would first tell you about the connections that led me to this work. In the spring of 2010 I attended the NLM/ Wood’s Hole MBL biomedical informatics course.  Here is half of my group after our tour of Martha’s Vineyard.  I loved the course.  On Thursday of that week, Joan Ash, a professor in the Department of Medical Informatics and Clinical Epidemiology at Oregon Health & Sciences University (and also a librarian and member of MLA) did a role playing session on the use of technology in hospitals, but she also mentioned that OHSU had grants (HITECH/ARRA) to get a graduate certificate in biomedical informatics.  I immediately decided to apply, but I couldn’t get my old, Canadian university transcripts in time, so I had to wait a year.  But I did get into the program, starting in Sept. 2011. After taking classes on informatics, evidence based practice, clinical information systems, health information management, statistics, SQL programming, and IT in healthcare, I had to decide on a practicum project.
  3. While reading a blog post , by the chair of the OHSU DMICE dept., Dr. Bill Hersh (who has also taught at Wood’s Hole), I became really interested in the idea of analytics as described in the report that was mentioned (see quote on slide).  I wanted to do a practicum that would use analytics in some way to improve patient outcomes.
  4. After asking around at my institution, I discovered that our Center for Clinical and Translational Research, part of the CTSA consortium, had a Biomedical Informatics Core (BIC) that was working on the analytics I was interested in.  They are actually setting up quite a few different resources, but the main focus has been on REDCap for surveys, i2b2 for cohort discovery, setting up a clinical data warehouse for hospital EHR data from Cerner, and HealthFacts, a large Cerner supplied database with data from 160 hospitals, which can be used for research.
  5. REDCap is used for collecting survey data and analyzing it, but to get the patient sets(which needs IRB approval) the researchers need to use i2b2 to see if there is a large enough cohort in the system.
  6. 12b2 is another open source software for research, but it was specifically designed for translational research and relating patients to genomic information.  At VCU, researchers don’t need IRB approval to use this database and they can play around with the patient characteristics they need for their research to find out if there are enough patients in our community for a study. When I learned about i2b2, I felt there must be more we could do with it.  One Core staff member said he had taught REDCap to a medical informatics class so they could learn about databases, but I thought using i2b2 would be even better.
  7. So I started to learn more about the system and think of ways to use it for teaching.  First off, there are 4 ontologies used for various parts of the system.  ICD9-CM for Diagnoses and Procedures, LOINC for lab tests, NDC for medications, and Snomed-CT for microbiology.  The important thing to realize about this database is that the Diagnosis code is for billing purposes, so there can be other diagnoses which aren’t billed and the numbers may not reflect what a researcher expects based on their experiences.
  8. The actual interface looks like this.  On the left, a box to search for terms to use. a workplace to share searches – especially if you need help, and a previous queries area that stores what I have done.   On the right, a query tool for setting up terms into a search statement, and a query status box to show how your search is running.
  9. In the navigate terms area, the ontology folder opens up to show all the areas that can be searched.
  10. And you can keep opening the various hierarchies to get to what you need. While browsing is fine, it may take a while.
  11. So, you can switch to the Find Terms tab and do a name or code search. You’ll notice though that things become a lot more complicated and specific. Luckily, you can just choose the base term to get the whole group. But if you needed secondary diabetes mellitus, you’d need to add more codes.  And the terms list includes things from all ontologies.
  12. Once the search criteria have been determined you can add date limits and run the search.  The number of patients is +_ 3 to keep the set deidentified.
  13. Once you have a set, you can switch to Analysis Tools and do basic Demographics on the set.
  14. The basic demographics can give you some idea of the type of community you will be working with.  So you can see, it is a not too hard a search process, but you do need to think about the ontologies – which is a good skill for medical students who will be using EHRs.
  15. It is easy to say medical students should learn how to use these analytic databases, but another thing to convince faculty. So I’ve been combing through various objectives reports and finding relevant reasons to learn to use an EHR database.  The Medical School Objectives Project of the Association of American Medical Colleges has put out several reports, with many objectives.  This one, Physicians must be dutiful, specifically mentions retrieving biomedical information – not just literature.
  16. This MSOP report looks at the medical informatics skills needed for various roles – Life-Long Learner, Clinician, Educator/Communicator, Manager, and in this objective, the role of Researcher.
  17. And in this report, a newer one, which is nice since the MSOP reports are from 1998, Utilizing Informatics is a core competency for working in interprofessional teams.
  18. Using i2b2 to find information to educate students about their patient populations can help direct their studies.  Also, learning to use databases can help with future quality analysis efforts.
  19. I2b2 can only do so much and the Provider information, especially on the hospital side as opposed to the clinic side, is based on the admitting provider, not necessarily the provider/resident who treated the patient. So I am learning about the Data Warehouse that is being set up at VCU.  Right now Bob, the BIC specialist, is using a Business Objects program to run searches, similar to the i2b2 interface where you move folders and terms to different boxes.  He almost has the more functional Data Warehouse ready.  It will smooth out some of the variations in the EHR record, and allow searching of the terms entered, not just codes.  And it can be accessed through any hospital terminal. We will create a template search for the residents so they just need to put in their name and then they can find out their patient loads for self-assessments.
  20. I’ve touched on the problems a couple of times, but like any research work, there are problems with the data.  As long as you are aware of the issues, you can temper your conclusions properly.
  21. You can actually search PubMed for i2b2, clinical data warehouse, and Cerner Health Facts and find articles that use the databases for research.  As you search, you will find that there are also all sorts of specialized health databases that are also being used.  EHRs are generating huge amounts of data.
  22. Which leads me to a suggestion for those of you who don’t have access to hospital data, learn about health data – not just health statistics.   This relates well to our speaker this morning who discussed open data (link to a blog post about this keynote  There are large amounts of open data on the web and some of it is medical.  Learn about these resources and the tools needed to use the data.  According to David Stuart, data is the new book.  In his book  Stuart paraphrases Ranganathan’s laws starting with “Every user his data”
  23. So look for data. is a good starting point.
  24. has quite a few categories of health data.
  25. And of course, don’t forget all the genetic data in NCBI.
  26. Yesterday, when Bart Ragon, , spoke about the BioConnector room they have created at UVa, he ended with a slide of flowers growing as an analogy for their growing program.  I’ve ended with flowers but for a slightly different reason.  This piece of embroidery is a reproduction of an Elizabethan coif.  It will be made up next year and used by interpreters at Agecroft Hall, an historical house in Richmond, VA.  I think this coif is like the projects that I have been working on.  It requires many people working as a team, lots of different skills, and lots of time.  But in the end it will be a wonderful thing.
This entry was posted in EHR data, MLAQUAD 2012, Papers and presentations. Bookmark the permalink.