IPDRA: Intelligent Patient Data Review Assistant

1
IPDRA Intelligent Patient Data Review Assistant
Phase I Final Briefing SBIR Topic
OSD01-DHP11 Jim Ong Stottler Henke 17 June 2002
Stottler Henke
Smarter Software Solutions
2
Stottler Henke

• 40 person applied research and software
  consulting company founded in 1988
• Core competency artificial intelligence (AI)

San Mateo, CA Seattle, WA Boston, MA Education
and training Information access Data mining
Knowledge management Decision-support Planning
and scheduling
3
IPDRA Mission
Help clinicians review patient data more
effectively before and during patient encounters
Current state-of-the-art Limitations
Web application server Relational
database Source-oriented queries Minimal use of
graphics Time-consuming to select relevant
patient data Hard to integrate and data spanning
multiple pages and data sources
.
4
IPDRA Project Team
5
IPDRA Design Goals and Assumptions

• Clinicians are busy before, during, and after
  visits
• Organizations and clinicians vary by their
• preferences for data content and presentation
  method
• willingness to delegate / trust
• proficiency (using software and visualizing data)
  
• IPDRA should
• adapt to the user, patient, type of visit,
  problem, etc.
• be predictable and understandable
• be extensible and maintainable
• encourage best clinical practices
• give precedence to user-supplied reporting
  decisions
• provide default reporting using systems knowledge

6
IPDRA Phase I Project Activities
Reviewed research literature and relevant
technologies (software and data) Gathered
Clinician User Requirements
Graphical Patient Data Summaries
Powsner/Tufte, CrossGraphs lab graphs
Plaisant et al Lifeline, PPD Patient
Profile Current Generation Patient DBs (ICDB,
Logicon) Medical Concept Queries (Zeng) Web-based
reporting systems Java graph objects Apelon
Terminology Server Unified Medical Language
System (UMLS) Arden Syntax (medical
logic) Project team discussions Clinician
questionnaires Requested descriptions of data
views for hypothetical patients Abstracted user
needs from concrete examples
7
IPDRA Phase I Project Activities (2)
Developed High-Level Design Prototyped Java
applet display objects Created mock-ups of
patient data views Developed Phase II Plan
Views show clinically-meaningful data
subsets Supports best clinical practices by
displaying computed targets, benchmarks,
checklists Authoring tool lets non-programmers
create view Re-usable display objects for
clinical data review Patient data object provides
simpler data model and potentially improves
performance via caching Timeline
summary Clinical time-series graph
array Combination timeline and time-series
graphs Cardiac, Liver Tests, Hypertension
View posted on www.shai.com/projects/ipdra Develo
p and evaluate operational prototype with initial
set of views using the Integrated Clinical
Database (ICDB) as the patient database
8
Example Clinician Data Review Goals
current health, recent treatments, and
response patient data related to current
problem notice other patient problems requiring
attention health maintenance identify
compliance issues identify patient conditions
affecting treatment selection
graph of most recent vitals, labs, medications,
If current problem headache, review
cardiovascular disease risk factors if
BPgt135/85, see BP meds and/or other BP readings
to determine if patient not Dx with HTN or
inadequately treated immunizations up to date?
last pap, mammogram/breast exam, psa, etc.
compare data about when patient came for refill
(stored in pharmacy DB) with treatment
plan allergies, other meds, contraindications
9
Example routine visit
Current concerns Health state Last regular
check up Chronic conditions Health
maintenance Current medications Current
allergies Current vitals Current weight
Is anything bothering this individual? Patients
perception of own health status? When was it?
Any issues that require monitoring? Known Dx?
Related symptoms and health state. Recommended
testing based on age and gender. Results of
previous health maintenance studies. Prescription
s? Over the counters? Alternatives? To what
meds, environmental/food elements.
Reactions? Heart rate and BP compared to normal
limits. Compared with 1 year ago and with norms
10
Example problem visit / known issue
Current concerns Health state Last problem
visit Health maintenance Current
medications Current vitals Current weight

Is anything bothering this individual? Patients
perception of own health status? Of
illness-specific health status? Change since last
assessment? What were the active issues? What
was the intervening plan of action? (testing,
treatment, etc.) Test results? What testing is
required for health maintenance on this
problem? Have meds been altered from last visit?
Patient response to any change in regimen?
Refills required? Any blood tests required based
on current meds?. Heart rate and BP compared to
normal limits. Compared with 1 year ago and with
norms
11
IPDRA Patient Data View Concept

• IPDRA maintains a library of views
• Each view facilitates rapid data review via
  effective integration and presentation of a
  clinically-meaningful subset of the patients
  data
• IPDRA automatically selects relevant views for
  visit
• Views support best clinical practices by
  presenting patient data and guideline data
  side-by-side
• Views provide drill-down and inter-view
  navigation
• Users can define views via GUI authoring tool by
  selecting and configuring variable calculations
  and text/graphics subviews and display objects.
• Users can create and save ad hoc views

12

13

14

15
View Logic Requirements
Database retrievals (time stamped data) Data
transformations and computations Conditional
logic (if-then) Iterations (for-each
loops) Actions
Display BP measurements (all, last, within n
years) Compute target BP for patient on HTN drug
according to guidelines If BPgt135/85 ... For
each possible AE for this drug Set variable
value(s) or add/modify display object used in view
16
View Logic Implementation
Borrow concepts from the Arden Syntax for Medical
Logic Systems Augment SQL access with
patient data object
View variables can be scalars or lists and tables
of time-stamped values Assign values to view
variables via database mapping and calculation
statements, specified via GUI Use variable
values in logic and action statements to control
the content or appearance of the view. A display
object property can be associated with a view
variable value. Provides simple object-oriented
access to raw/computed patient data
17
Example KB-Driven Problem View
For a problem For each risk factor For each
related problem For each prescribed, indicated
drug For each reported AE For each
possible AE
Show relevant clinical data in timeline with
treatments Show risk factor and related
data Show problem and related clinical
data Show drug dose and period Show target drug
levels Show AE Show clinical data for the
AE
18
Medical Knowledge Base Supports Views
unique medical concept Ids relationships among
concepts knowledge base (or ontology)

Related comorbid
indication contraindication
has risk factor
Treatment
Risk factor
Problem
possible AE
shown by
shown by
Adverse Event
shown by
Clinical Data
19
Automated View Selection Based on
Type of visit Clinician specialty Current
problems Individual user Patient data
Routine, problem (initial), problem
(follow-up) Primary care, cardiology,
etc. Shortness of breath, chest pain Dr.
Jones 50 year old male chronic problems
hypertension, diabetes current meds Atenolol
Example view selection rules If chronic problems
include hypertension then add Hypertension
View If specialty is cardiology then add
Cardiology View
20
Adaptive View Selection, View Definitions
Page A
Page B
Page C
Panel 1
Panel 4
Panel 3
Panel 2
Panel 5
C1
C2
C3
C4
C5
C6
C7
C8
C9
C10
Always display these alerts
Always display this standard view
Sometimes display this standard view

Always included

Conditionally included
21
Authoring Views
Author configures a display object to specify
data content and appearance
Author includes display objects in a subview to
create a reusable reporting object
22
Authoring Views (2)
Author includes subviews in the view
subviews
display objects
IBUPROFEN 600 MG TAB Qty 90 for 30
days MAGNESIUM OXIDE 400 MG Tqy 90 for 90
days ACETAMINOPHEN 325 MG TAB Qty 100 for 30
days
23
IPDRA Phase II Project Goal
Develop and evaluate an operational prototype
that convincingly demonstrates IPDRAs
feasibility and utility.
Deliverables Testbed
IPDRA view generation web server Authoring
tool Display objects library Medical knowledge
base and tools Initial set of view
definitions Integrated Clinical Database (ICDB)
.
24
IPDRA Phase II Architecture
25
IPDRA Phase II Access ICDB Data
Web browser
ICDB web app server
IPDRA web app server
ICDB database (Oracle)
HL7
CHCS database (MUMPS)
26
IPDRA Phase II Project Plan
Month
Create View Mock-ups Basic View
Generation (display applets, database access
patient data object, html generation) Advanced
View Generation (KB-driven, ad hoc) Knowledge
Base (select data sources, editor /
loader) Authoring Tool Evaluate System
Deliver
Spec/Design
Develop/Test
User Test