I am grateful for Jim DeLeo and Carl Leonard for giving me a slot to run a brainstorming session at the National Institutes of Health’s Biomedical Computing Interest Group. I called the meeting because I wanted NIH patients to have access to their medical records in services like Google Health and Microsoft HealthVault.
We had a great turnout, over 20 people, all of whom had deep knowledge about different aspects of NIH’s electronic medical records infrastructure, and we began going through the challenges and opportunities for such an initiative. I took some notes but am leaving out the names of the contributors as I had forgotten to ask for their consent. If you read this and want some extra information about the event, please contact me directly at email@example.com.
One outcome I was particularly pleased with was that the few of us who were familiar with the Continuity of Care Record successfully pushed for adoption of this data standard. One of the EMR policy administrators said that she would include the standard in the next roll-out of the software, allowing export of patient data on an NIH computer in CCR format.
Below is the list of challenges we identified:
1. Policy problems (of course)
- built with NIH’s EMR vendor, which does have patient portal product, but this is not yet activated
- medical executive committee raised this possibility 5 years ago but back then did not have a way of pulling out the data
- NIH is exempt entity from HIPAA because it does not move data outside of NIH, but as soon as the data is transferred outside to PHRs, the legal questions begin
2. Need willingness of clinical investigator to hand over the data before it is published in papers (everyone in the room agreed that this should not be a concern, but that it is in reality).
3. Translation of SQL identifying data is optimized for transactions and not aspects
4. Must have data model that takes research needs into account, something the CCR is clearly deficient in
- need right metadata, e.g. were you taking beta blockers at the same time?
- e.g. analyte data will very over years if change machines – the rescaling of the data is done in the PI’s data warehouse, but not back in the original CRIS
- e.g. Genechip analysis data requires information about how te sample was handled, chemical hybridization information, etc