- From: David Booth <david@dbooth.org>
- Date: Wed, 30 Sep 2015 10:49:37 -0400
- To: Peter.Hendler@kp.org
- Cc: its@lists.HL7.org, public-semweb-lifesci@w3.org
Excellent! Let's talk offline about scheduling. Also, a question: what benefits did you experience in using an RDF/SPARQL approach as opposed to a relational/SQL approach? (Playing devil's advocate) Wouldn't the query that you described have been relatively simple in SQL? If not, why not? Thanks, David Booth On 09/29/2015 11:30 PM, Peter.Hendler@kp.org wrote: > Even better. I can just present it to you on a webex or another kind of > call. > > From: David Booth <david@dbooth.org> > To: Peter Hendler/CA/KAIPERM@KAIPERM > Cc: its@lists.HL7.org, public-semweb-lifesci@w3.org > Date: 09/29/2015 03:15 PM > Subject: Re: Using RDF, Datalog, OWL-RL and a "RIM lite ERPA Ontology" > to calculate HEDIS Quality measures > ------------------------------------------------------------------------ > > Excellent! Unfortunately I will miss your presentation. Can we get > your slides? Even better, if your presentation is recorded that would > be awesome. > > Thanks, > David Booth > > On 09/29/2015 05:38 PM, Peter.Hendler@kp.org wrote: > > At KP, and working with Ian Horrock's group at Oxford, we have been > > experimenting with their new RDF, Datalog, OWL-RL triple store called > > "RDFox". > > > > We have calculated the HEDIS Diabetes quality measure on a population of > > over 400,000 patients real data. > > > > We still have to compare our numerators and denominators to results > > calculated with SQL and traditional DB tables. > > > > I will be presenting a very simple version of this at HL7 at the AID > > work group in Atlanta on Monday Q3. > > > > I believe this is the first time a complex HEDIS quality measure has > > been calculated with RDF, OWL and Datalog and SPARQL on a large > > population of real patients. > > > > I will not be presenting the complete complex HEDIS measure (which would > > take days), but a smaller example to explain how it all works. > > > > We used SNOMED subsumption to generate a small value set of SNOMED codes > > that are "kinds of Diabetes". Using that SNOMED VS, we found all the > > patients who had a visit coded for Diabetes. Then we searched all of > > their HgBA1C values and then found the "last value". We could then look > > at the numerical results of the HgBA1C and find how many of them were > > below 7% (good control). > > > > In order to do this we had previously created an OWL ontology based on > > Entities in Roles that Participate in Acts. It is not the full HL7 V3 > > RIM, but only what was needed for this exercise. > > This "KCOM" model is what we presented before at HL7 AID meetings. > > This entire project would not have been possible to do without first > > mapping the raw clinical data to this ERPA OWL backbone ontology. All > > of our queries were based on this ERPA (Entities in Roles Participating > > in Acts). > > > > RDFox is multi threaded and we were able to run the data materialization > > on 8 threads on an 8 core machine with 64 Gig RAM. It ran in only a few > > hours and we have already found ways to speed it up further. > > > > Hope to see you at HL7 Atlanta. > > > > *NOTICE TO RECIPIENT:* If you are not the intended recipient of this > > e-mail, you are prohibited from sharing, copying, or otherwise using or > > disclosing its contents. If you have received this e-mail in error, > > please notify the sender immediately by reply e-mail and permanently > > delete this e-mail and any attachments without reading, forwarding or > > saving them. Thank you. > > > > > > >
Received on Wednesday, 30 September 2015 14:50:05 UTC