Monday, November 15, 2010

AMIA 2010: Sunday


My first AMIA meeting and I'm finding it filled with all sorts of goodies. There's lots of tweeting going on, too. (Go to TweetChat with hashtag #AMIA2010 to keep up.)

I typed some notes and thought I'd put them here (1) so I can find them again easy and (2) some folks may find them useful (thought maybe somewhat cryptic). (The odd formatting is from OmniOutliner; I've given up on fixing the font colors--the red means nothing--and the bullets don't line up right.)

AMIA 2010 Susan Dentzer, Editor Health Affairs

2010.11.14

"Golden Age" of PPACA (?)

compared Greek myth of Ae____ to Obama

Open up goldmine of data to public

CHDI website www.hhs.gov—open

use data to shape community responses to childhood obesity... Children's Optimal Health

[] l/u Health Affairs 2010 Nov 29:2047

rapid-cycle improvement of medical home care @Geisnger

reductions in risk-adjusted chronically ill hospital admission and readmission rates

Three Aims

1. Better Health

2. Better health care

3. Better value

IOM study on variations

look at var in HC spending, pt diversity, MD decisions on what care to give and evidence

Shared decision making by pts

ACA provision

even informed pts don't exercise their options well

preference-sensitive care

Quotations

Quotes by George Carlin, Churchill on American People

Quote by Jerry Garcia re Somebody has to do something and it is pathetic that it must be us



Informatics Issues in HIEs

20101114 Rob Kolodner, Session chair

Development and use of a Medication history Service Associated with a health information exchange: Architecture and Preliminary findings

M. Frisse, Vanderbilt Center for Better Health; L. Tang; A. Belsito, M. Overhage, Regenstrief Institute

Memphis HIE in use 4.5 yrs... still a pilot going to full force

48% had full med hx; getting retail pharmacies online made a big different

36% pt not located

$4 generics good for pts, bad for accurate HIE data



Emergency Medical Services: the frontier in health information exchange

JT Finnell, M. Overhage, Regenstrief Institute

Project: push HIE data out to EMS staff in 2 counties in Indiana (Siren by Medusa), using ruggedized tablet PC

Uses LN, FN DOB Gender, can add zip and SSN optional

Pushes back a pdf with a bunch of data: med list, allergies, DNR status, NOK notification, PMH

Started at 15% requests to 26% of field pts seen

14% of medics never used

connectivity a problem

41% medics nearly always queried system

66% said data was important for providing care

"truth serum" effect (we can look it up if you don't tell us the truth)

Value with heavy utilizers and pts "found down"


Private Medical Record Linkage with Approximate Matching

E. Durham, Y. Xue; M. Kantarcioglu; B. Malin, Vanderbilt University

Vanderbilt -- Emory: Flaw in current model for sharing de-identified data

If pts don't match, they can be overcounted

prob with fragmented data (spelling, missing data, etc)

Private Record Matching can improve this using a hash function to maintain privacy but compare individuals

Steps:

blocking (toss out some)

field comparison

similarity function measures degree of similarity using a comparison vector and into a record pair similarity score

draw a line where scores above a number are considered a match

record pair comparison and classification

Fellegi-Sunter probability vectors: uses log function of an agreement weight and a disagreement weight for each field (eg, FN, LN, MOB)

Sum the similarity scores across all fields

Used a data corrupter function to mess up experimental data to see if they get rematched

looked at accuracy vs runtime

Main thing they did was to use approximate matching rather than binary matching, leading to increase in accurate matches

Note: they used a centralized approach but hope to extend ot decentralized model.


Continuity of Care Document (CCD) enables Delivery of Medication histories to the Primary Care Clinician

L. Simonaitis, A. Belsito, G. Cravens, C. Shen, J. Over- hage, Regenstrief Institute

Used INPC (Indiana)... x1995, 6M pts, 3B data items

Workflow

pt to front desk... ADT clinic trigger sends CCD request

printer prints med hx (CCD)

MD reviews meds

CCD is generated, pulling in data from SureScripts and MA, and put rx codes from dif stds in same CCD

Results: used in 1 clinic over 9 docs and 4500 pts

90% of cases took less than 2 min

med hx helped doc discover drugs they didn't know pt taking, controlled rx overuse, and med underuse

docs usually did not show med hx to pts

no change to MD workflow

accommodates paper-based clinic

[] XSLT transformation of CCD data very promising