Transcript AAVLD Informatics Committee: Data Standardization in the

EMR overview:
Opportunities, Realities and the
Practice/Research Divide
Rockville, MD
2009 07 13
Clement J. McDonald MD
Lister Hill National Center for Biomedical Communications
National Library of Medicine
1
Disclaimer

These remarks are based on my opinion and
experience and do not necessarily represent
an official opinion or position of the National
Library of Medicine or NIH
20009 06 03
Clem McDonald - Lister Hill Center
2
BACKGROUND
OPPORTUNITIES
Where are the opportunities
 Repositories
, EMRs
 Personal health records
 Networked devices and instruments
Clem McDonald - Lister Hill Center
5
REPOSITORIES, EMRS
LOTS OF DATA – MORE
TO COME
Current state of world
 Clinical




repository
Everyone has or is close to having them
If they are only halfway decent - everyone loves
them -doctors and administrators
EMR (with CPOE +- note entry)
Slower uptake


Takes doctors time –
Primary care doctors have NO time
Clem McDonald - Lister Hill Center
7
Typical clinical repository content
 Any
information generated via computer
systems

ORDERS /Prescriptions

Dictation (op notes, visit notes discharge summaries)

Laboratory results

Radiology reports

Radiology images

Endoscopy reports

Admission and discharge DX’s =and more
Clem McDonald - Lister Hill Center
8
EKG flow sheet- click to get the tracing
Clem McDonald - Lister Hill Center
Clem McDonald - Lister Hill Center
Radiology studies- click to get the study
Radiology images - thumbnail
Clem McDonald - Lister Hill Center
BIGGER
Clem McDonald - Lister Hill Center
Research uses
 Find
numbers and statistics needed to plan
studies and write grants.
 Help to recruit study patients with providers
consent and involvement
E.g. Regenstrief Institute, Columbia, Harvard
 De-identified
studies and statistical analysis
(many examples)
 Provide follow up data for longitudinal
studies.
 At IU , 80% of human studies used the
Regenstrief Data during some point of their
evolution.
Clem McDonald
Lister Hill Center NLM
Huge repositories with much
research use at many institutions
 Broad
spectrum of structured data - billions of
data points






Partners HealthCare (Sean Murphy)
Regenstrief Institute (Marc Overhage)
Kaiser Permanente
Veterans Administration
Columbia University
Mayo Clinic
Clem McDonald
Lister Hill Center NLM
Tools available for querying EHRs at
many institutions
 They
have invested in the hard work of
regularizing / standardizing the data - at many
levels.
20009 06 03
Clem McDonald - Lister Hill Center
16
Harvard-Partner’s Query tool
Query items
Person who is using tool
Query construction
Results - brokenClem
down
by number distinct of patients
McDonald
Lister Hill Center NLM
Vanderblit’s clinical and
specimen search
Search requirements specify to
return only records with
biological samples available over
a certain volume amount
•Researcher selects samples
•Researcher executes search using defined parameters
Select
Researcher
selects most
appropriate
records
Keywords in context provide information
for evaluating records
Clem McDonald
Lister Hill Center NLM
Regenstrief SPIN tool
Clem McDonald
Lister Hill Center NLM
What use for drug and alcohol abuse
research
 Care
institutions can and do capture info about
smoking, do CAGE screening- etc
 Should be a part of routine data gathering at intake
 Can be the basis of interventions
 Treatment center data is never available, by law.
 But repositories do carry related data obtained in
course of routine care- e.g. Ethanol levels.
Medications – prescribed during acute care
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
20
RHIOS –the next step –even
more data, population based
 Examples

INPC central Indiana (2 Billion results- (Marc
Overhage)




(McDonald 2006 Health Affairs)
Memphis – (Marc Frisse -Vanderbilt)
The Ontario Children's network (all test results from
all pediatric hospitals made available to all
pediatricians ) Gill Hill
Massachusetts e-Health project - five practices,
many sites (David Bates)
Clem McDonald
Lister Hill Center NLM
More RHIOs on the way
2007 12 14
Clem McDonald
Lister Hill Center NLM
A national infrastructure, RHIOs and
standards
 Big
federal push for clinical data and
message standards
 This will facilitate and foster RHIOs (EHIs).
And they will ink blot across the country
 And permit links between these regional
networks.
 Huge research opportunities
 Will return to the subject
Clem McDonald - Lister Hill Center
23
PERSONAL HEALTH
RECORDS (PHRS)
20009 06 03
Clem McDonald - Lister Hill
Center
24
PHRs - who is providing them
 Everyone

Google Health
https://www.google.com/health/html/faq.html

My Medicare (and 2 other pilots)
http://www.mymedicare.gov/

Microsoft- HealthVault
http://www.healthvault.com/

Intuit – Quicken Health
http://quickenhealth.intuit.com/

Many other institutions
Clem McDonald , Lister Hill Center, NLM
25
Special features of PHRs
 Some


E.g. weight, glucometers, exercise (esp Microsoft)
Continua – a consortium of device vendors working
on standards for capturing home instrument data
 Many




connect to home instruments
research and treatment opportunities
Collect day-to-day status of patients in research
studies
With patient consent- options for patient recruitment
Rule rules can provide he behavioral intervention
Could be the core of a drug treatment record – with
patient in charge
26
Some examples of active and
research use – (of PHR Portals)
 Active





research projects and patient usage
Partners HealthCare
Children's Hospital of Boston
Vanderbilt University
Department of Veterans Affairs
Kaiser Permanent
20009 06 03
Clem McDonald - Lister Hill Center
27
NLM’s PHR
Overview
 Uses




NLM’s standard vocabularies
LOINC for observations
Rx.Terms (subset of Rx.Norm) for drugs
SNOMED CT for conditions and findings
HL7 data types
 Links
to information sources from NLM
(MedlinePlus, ClinicalTrials.gov), CDC, etc
 Open Source
NLM-PHR Highlights
page data entry form – no jumping
around to complete an entry
 Codes key information - e.g. drugs, problems,
etc
 One click links to educational info
 Rule-based reminders about prevention and
healthy behavior ( Behavioral interventions)
 Automatic computation of derived values and
defaults
 Rule-based form morphing
 One

What shows on the form changes
Gender, female, Pap and mammogram
Change to male and immediately
Discontinue, Revise, Delete to Generate
Dynamic Accurate Medication Tracking
33
Delete the beclomethazone
(permanently)
34
Form can morph to ask any set of
questions- e.g. Phq-9
35
Could (will) use exact same method
for NIDA’s abuse screener
20009 06 03
Clem McDonald - Lister Hill Center
37
NLM PHR Demo
 https://phr.nlm.nih.gov/

not yet publicly available
Clem McDonald - Lister Hill Center
38
HOME MONITORING
DEVICES
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
39
Opportunities for NIDA treatment
programs
getting cheap – and options for direct
delivery to PHRs or treatment centers are coming
 Conceivable that a package of physiologic
measures –pulse, skin conductance, activity level,
could provide info for titering drugs during initial
treatment
 We have an epidemic of methadone related
deaths
 Would continuous O2 monitoring and email
alarms to treatment centers prevent ?
 Devices
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
40
THE DIVIDE
20009 06 03
Clem McDonald - Lister Hill
Center
41
Different directions in standards
 Clinical
world has been using the HL7 message
standard for 15 years
 Federal government has defined very specific
standards that leverage these existing standards



HL7 V2.5 and HL7 CDA for messages
LOINC and Rx.Norm and SNOMED for codes
http://www.himss.org/ASP/topics_FocusDynamic.asp?f
aid=211
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
42
HITSPI
 ONC
has $2 billion to push and complete this
trend.
 Research interest in standards is newish – and
mostly going in its own direction
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
43
The two worlds have different ways of
thinking about data structures
 Makes
it harder to even talk about the differences
 Researchers uses a flat data structure (often a
spread sheet) where the variables are defined as
column headers. So you see Cholesterol as the a
column name and a value 210 (below it ) The units
are assumed.
 Clinical systems use a stacked data structure. So
reading across a row you see Cholesterol within
one cell, 170 in another cell and mg/dl in yet
another cell.
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
44
Flat structure
Patient
ID
Name
1234-5
Surgery
date
Hgb
DBP
Doe Jane 12May95
13
95
3i
80
180
9999-3
Jones T
12.5
88
2
90
230
8888-3
Doe Sam 4June95
16
78
0
80
205
1Aug95
Clem McDonald - Lister Hill Center
# of
BPU
Bypass Cholest
Minute
Stacked structure
Operational Data Base: One Record Per Observation
Pt ID Relevant
Date
Observation
ID
Value
Units
Normal
Rang
Place
Doe J 12-May95
Hemoglobin
13
mg/dl
12.5-15
St Francis Dr Smith
Doe J 12-May95
Hemoglobin
11.5
mg/dl
12.5-15
St Francis Dr Smith
Doe J 12-May95
Dias BP
95
mm/H
g
80-140
St Francis Dr Smith
Doe J 12-May95
Dias BP
110
mm/H
g
80-140
St Francis Dr Smith
Doe J 13-May95
Bypass
minutes
80
min
Doe J 12-May95
Cholesterol
180
Clem McDonald - Lister Hill Center
Observer
St Francis Dr Sleepwell
St Francis Dr Bloodbank
Research World Uses Flat Structure
 When
number of variables (questions) is
small , they are easier to manage and
analyze as a flat structure
 But – for longitudinal studies, the changes
that creep into study protocols generate many
different flat files that are difficult to integrate

Woman’s Health study
 Easier
to combine data from across studies if
stacked structures were used -Clem McDonald - Lister Hill Center
47
Clinical World Uses Stacked Structure
 More
general structure
 Can accommodate the thousands (or tens of
thousands ) of variables
 Allows repeat measures at varying time
intervals
 Allows storage of additional attributes per
result – like who recorded per variable, normal
ranges, etc
 Allows rich definitions of variables in a master
file tied directly to the variable in the database
Clem McDonald - Lister Hill Center
48
Crux of HL7 V2.x





The messages have a stacked data
structure (
Contain only printable characters (ASCII
text) ––V2.5 and uses delimiters to define
the fields and sub fields within its “records”
Vertical bars (|) separate fields
Hat ( ^ ) separate subfields
(There is also an XML version of HL7 V2.
Clem McDonald - Lister Hill Center
49
Fields have data types.
An
espescially important one

CWE = Coded
»Code1 ^ print text ^ Code system
Clem McDonald - Lister Hill Center
50
Here it is
– yellow is variable. Orange is value
Patient level
PID|||0999999^6^M10||TEST^PATIENT^||1992022
5|F||B|4050 SW WAYWARD BLVD |
Order/report level
OBR|||H9759-0^REG_LAB|24358-4^Hemogram^LOINC
 Discrete Results
OBX|2|NM||789- 8^RBC^LOINC||4.9|M/mm3| 4.0-5.4
OBX|3|NM|718-7^HGB^LOINC||12.4|g/dL|12.0 5.0||||F|
OBX|4|NM||20570-8^HCT^LOINC||50|%|35-49|H|||F|
OBX|5|NM||30428-7^MCV^LOINC||81|fL|80-94||||F|b
20009 06 03
Clem McDonald - Lister Hill Center
51
Some things to notice
 Most
clinical measurements . Test results, survey
instruments data can be delivered by this structure
or a nested version (supported in HL7)
 For cross institutional communication need a
universal code for the variables
 The columns of HL7 are “fixed” so little need to
worry about individual data base elements.
 The free LOINC data base is a source for such
codes

As well as form defining packages
2009-07-13
Clem McDonald - Lister Hill Center,
National Library of Medicine
52
LOINC Web site – http://loinc.org/
-can download everything
20009 06 03
Clem McDonald - Lister Hill
Center
Our plea to the research world
 Look

carefully at the coming HTSP standards
If nothing else it will provide some access to most
of the data in clinical systems
 Use
universal codes (LOINC ) for defining
variables
 Use SNOMED CT for coding variables with
categorical values, e.g., anatomic sites,
findings, diagnoses, etc
 Use Rx.terms (RX.norm) to identify drugs
Clem McDonald - Lister Hill Center
54
THE END
Clem McDonald - Lister Hill Center
55