PrimeAnswers:
Knowledge at the Point of Care
NLM
Information Systems Grant LM06758
Y3
Progress Report Summary
1 December 2001
Studies and Results. Phase
2 objectives to simplify and speed access to best content through a centralized
search with local indexing, one-click search of specific external sources and a
targeted list of common conditions have been achieved. The appendices provide visual detail of the
developing system.
Content
·
Concluded institutional licensing for
final products desired for PrimeAnswers (i.e., UpToDate and Clinical Evidence). Appendix II lists content sources. Applied
Evidence articles appeared as new series in several core medical journals in
2001 and were indexed as evidence documents, including the Family Practice
Inquiry Network (FPIN) Answers. These new sources provide a rich source of
evidence-based reviews to answer specific practice questions that supplements
the gold standard Cochrane Reviews and the Clinical Evidence summaries.
Duplication between these efforts will be reported in 2002. Prediction rules,
particularly those available as interactive web tools or downloaded PDA
programs, were also selected.
- Completed 16 common
conditions for version 2 release in October based on recommendations by
primary care providers. Refinement continues and review by two primary
care providers is planned by end of Y2.
- Refocused Q&A curbside
evidence into a partnership with the Dept of Family Medicine (UWFM) and
the FPIN Consortium. UWFM led prototype development of a question gathering
tool built in Microsoft Access. We are now integrating the question tool
into an SQL-Cold Fusion model in the PA system.
System Architecture, Database, Editing Tools. (See
Appendix III)
- Designed system based on a
flexible, reproducible and low-threshold architecture using Microsoft SQL,
Cold Fusion and JavaScript. Code is designed for reusability and insures
easy system-wide changes in single location. Separate development and
production systems are in place. System is designed to put page layout,
content editing, immediate review of changes,
database reindexing, and other system
maintenance in the hands of content editors/administrators without
programmer intervention.
- Implemented data model in
SQL (see table design in Appendix III). Content is focused on sources,
queries and documents. Sources are links to an indexed
title or package of information. Queries are formatted dynamic
queries created by the programmer as a template which is reusable by
content editors. Documents are topic-specific; the
majority are evidence-based. Documents are linked to a parent source
or a parent document allowing related or sibling document displays.
Documents are indexed by the document type, keywords and triple (i.e.,
question type and semantic relationship) level and contain a “bottom line”
if copyright permits. We only store rich metadata and a pointer to the
full text. The database contains 175 source records, 86 query records, and
203 documents as of November 1st.
System Architecture, Database, Editing Tools. (See
Appendix III)
- Designed system based on a
flexible, reproducible and low-threshold architecture using Microsoft SQL,
Cold Fusion and JavaScript. Code is designed for reusability and insures
easy system-wide changes in single location. Separate development and
production systems are in place. System is designed to put page layout,
content editing, immediate review of changes,
database re-indexing, and other system maintenance in the hands of content
editors/administrators without programmer intervention.
- Implemented data model in
SQL (see table design in Appendix III). Content is focused on sources,
queries and documents. Sources are links to an indexed
title or package of information. Queries are formatted dynamic
queries created by the programmer as a template which is reusable by
content editors. Documents are topic-specific;
the majority are evidence-based. Documents are linked to a parent source
or a parent document allowing related or sibling document displays.
Documents are indexed by the document type, keywords and triple (i.e.,
question type and semantic relationship) level and contain a “bottom line”
if copyright permits. We only store rich metadata and a pointer to the
full text. The database contains 175 source records, 86 query records, and
203 documents as of November 1st.
User Interface (See Appendix IV)
- Released
hard-coded version 1 interface based on the results of the Y1 contextual
inquiry (March 2001). Version 1 allowed usability testing while the
database-driven system was under development. Version 2 was released to
clinics October 1, 2001.
The design focuses on simple, uncluttered navigation to key content items
organized in categories preferred by the providers and uniform “one box,
one click” searching of PrimeAnswers content and other key sources.
Reduction of “clicks”
is a core design objective and has been favorably reviewed by anecdotal
feedback. Current PrimeAnswers searching is site-wide, with one click to
evidence documents that can be resorted by alternative quality criteria,
and one click to bottom line, summary tables and native document.
·
Speed and clean visual layout continue to be
core design objectives. Minimized very selective “redhot”
clinical care news and linked to native document (e.g., Baycol
recall, Bioterrorism clinical guidelines). Deferred
personalization due to provider objections to a personal sign-on. Shared
computer use in clinics complicates a cookie per machine approach. Based on
provider recommendation developed a mouse-over popup description which created
50% overhead for page downloading noticeable on slow computers in one clinic. A
slightly more cluttered but faster icon image mouse-over pop-up or one-click
mini-window for descriptive information replaced the original design.
PDA Devices
(See Appendix V)
- Equipped the project’s three primary care providers with a personal
digital assistant (palm-based TRGpro) loaded
with ePocrates qRX and
Tarascon's ePharmacopoeia
as part of our ongoing investigation of pocket devices for answering
clinical questions. Pre- and post- questionnaires were used for a 3 month
test. While these PCPs did not prefer pocket devices anecdotal reports
from other providers in the test clinics indicate that PDAs
are used in patient care, but only for drug reference and simple calculators.
These results are confirmed by the literature and a recent UW Dept of
Anesthesiology survey. PrimeAnswers began indexing the best downloadable
pocket resources, particularly prediction rules, calculators and drug
reference as recommended by providers. System developments will not focus
specifically on PDA synchronization, but will maintain device-independence
as an objective.
Significance
Informal critiques of the value of the
“one-click” PrimeAnswers interface locally and externally have been very
encouraging. Basing our system on “off the shelf” software allows us to
replicate and sustain the system collaboratively with other institutions, for
example, the FPIN academic health libraries. We believe the model is replicable
in other health domains. An interesting body of knowledge about duplication of
evidence answers across primary care is being developed as the project
continues. Reporting redundancy may encourage collaboration or an opportunity
for overarching summarization, potentially in collaboration with the FPIN
project. Collaboration with UWFM, FPIN and the informal Question Coalition
grounds PrimeAnswers in a larger national scene. We are working with UWFM to
integrate collection of questions and key physician data to evidence-based practice-educational
objectives across the residency network. We are also considering an editorial
option for UW authored FPIN Answers. Three librarians piloted providing
searching support for FPIN answers during Y2, which revealed that support of UW
physician authors would maximize time/benefit. It is anticipated that the
PrimeAnswers and FPIN projects will continue to collaborate on synergistic
goals and objectives.
Plans
The project timeline is on target. Key
outcomes for Year 3 are: 1) develop and
implement an indexing schema to include primary care question type, semantic
relationship of concepts (i.e., condition:task:intervention)
that capitalizes on the UMLS thesaurus and semantic net; collaborate with the
UWFM and FPIN to insure an extensible taxonomy; reassess
licensure with publishers for direct “bottom line” linkage to accomplish our
three-tiered approach to evidence answers; 2) explore
optimal display techniques for “bottom line” answers;
3) analyze and report on duplication of primary care Q&A pairs in
multiple evidence products with similar goals and audiences; 4) complete
integration the UWFM question collection program into the PA system using SQL
and Cold Fusion; 5) investigate a partnership
with UWFM to serve as the editorial team to develop FPIN Answers for
publication in the American Family Physician; 6) collaborate in the
development of FPIN’s
new database-driven web architecture for FPIN Answers;
7) conduct PrimeAnswers system usability and analyze web log tracking
statistics to inform design improvements; 8) investigate
reproducibility of the FPIN speed/answer study published in Journal of
Family Practice, Oct 2001; and 9) test
usability and sustainability in the following areas: a) expand to all UW the
general internal medicine and family medicine clinics; b) add a Seattle Native
American Health Clinic affiliated with UWFM; c) open the system to non-urban UW
primary care clinics to identify potential access barriers; and d) investigate
replication into another subject domain or to another academic health sciences
library.