Building an Ontology of Adverse Event and Near Miss Data: Information Gleaned from The Patient Safet

1
Building an Ontology of Adverse Event and Near
Miss Data Information Gleaned from The Patient
Safety Event Taxonomy (PSET) Project

• Mariana Casella dos Santos, MD
• Language and Computing. Inc

2
PRESENTATION OVERVIEW

• Incident Reporting Today
• The intent of PSET
• HIRO
• What is it and why?
• Reconciling the data
• Enabling PSET classification
• Towards a HIRO powered reporting system

3
Incident Reporting Today
Passed into law on July 29, 2005, the Patient
Safety and Quality Improvement Act stipulates
that the Department of Health and Human Services
(DHHS) will develop a process for the voluntary
and confidential reporting of health care related
errors by health care organizations to Patient
Safety Organizations (PSOs) which will develop
ways to improve patient safety and reduce
medical errors.A network of patient safety
databases is to be developed that can accept,
aggregate and analyze non-identifiable patient
safety data. Voluntary use of national standards
to promote electronic exchange of healthcare
information will be developed or adopted. Use of
data standards would provide a common reporting
format for the network of patient safety
databases that will be developed, and will allow
for the comparison of data across multiple health
information technology systems (incident
reporting systems).
4
Incident Reporting Today

• Divergent Terminology

• Non-Standard Data Sets

• Laborious Processes leading to non-compliance

Low Quality of Data Captured
5
Project Overview

• This project is premised on the notion that a
  centralized, standardized national reporting
  system for adverse events will be highly valuable
  in promoting quality improvement, and in reducing
  medical errors.

6
Project Overview

• This project will provide answers to many
  questions surrounding the feasibility and
  clinical utility of combining, and analyzing data
  from across the many reporting systems that
  collect adverse events, near misses, and other
  patient safety-related information.

7
CF Causal or Contributing Factor
8
Making Systems Interoperable
Incident Reporting System C
Incident Reporting System B
Incident Reporting System A
Patient Safety Event Taxonomy
Standardized Data
9
(No Transcript)
10
Analysis of Taxonomic Relationships

• Hypothetical relationships that exist between
  nodes of the taxonomy
• Event type to Cause
• Cause to Impact

11
The Hospital Incident Reporting Ontology (HIRO)

• What is it? And what for?

12

• Research Questions
• Does deconstructing data using the PSET/HIRO
  provide a hospital with more useful information
  than it could obtain using its own pre-existing
  system?

13
Hospital Data, PSET HIRO
Making Systems Interoperable
Hospital C Incident Data
Hospital A Incident Data
Hospital B Incident Data
Standardization
Perspective
Patient Safety Event Taxonomy
14
Hypothesis

• The aggregate data provided to hospitals after
  deconstruction with PSET and analysis with the
  ontology is more useful than data aggregated
  using existing resources

15
3 Components of the HIRO project

• HIRO ontology
• Formal ontology
• Re-Built from scratch
• Populated with incidence report observables
• View independent representation of reality
• Hospital incidence report terminologies
• An aid to integrating patient data into the HIRO
  ontology
• PSET Definitions
• Patient Safety Event Taxonomy delivered by JCAHO
• View application perspective reality in terms
  of incidence reporting

16
The Basic Terminology
Hierarchical relationship Denoted also by is
subsumed by or inherits from or
is-subclass-of (OWL)
TELEPHONE
Horizontal relationship Also called property
(OWL) or predicate (FO)
IS-A
HAS-PROPERTY
BLACK TELEPHONE
BLACK
Concepts also called entity or type (FO) or Class
(OWL)
Zwarte telefoon
Black phone
Terms also called description or name Not
supported by OWL
17
Scope scale of HIRO
INCIDENT
Horizontal relationship
IS-A
HIRO 37 relationship types, 1879
Non-hierarchical relations
HAS-NON-INSTANTIATION
INCIDENT WITHOUT HARM
HARM
Concepts HIRO Originally 400 concepts, now
4053. PSET 162 Concepts
Incident, no harm
Terms HIRO 4561 English terms PSET Hospital
terminologies 5910
18
Alternate Views granular partitions
Ontology Ontology is the science of what is, of
the kinds and structures of objects, properties,
events, processes and relations in every area of
reality.
Taxonomy A specific partition of reality
(incidence reporting) A drawing of a boundary
over a certain domain An artifact of our
cognition
Think of it as Albertis grid...
19
HIRO PSET
HIRO Formal Ontology
REALITY
Bridging Horizontal relationships
PSET Taxonomy (original structure)
INCIDENCE REPORTING PERSPECTIVE
20
Differences in structuring
Taxonomy
Purpose Classification a given defined
portion of reality according to a specific
perspective
Drawbacks Rigid structure No possibility for
expansion Example Loading/merging of other
classifications is impossible Financial
Impact on Hospital Taxonomy
21
Differences in structuring example
Expandibility - Neutrality
22
Differences in structuring example
23
HIRO Framework
DOMAIN-ENTITY
Upper Level Concepts
ENDURANT
PERDURANT
MOTION PROCESS
Mid Level Concepts
FALL
Domain specific Concepts
FALL FROM BEDSIDE COMMODE
FALL CAUSED BY IMPROPER FOOTWEAR
24
HIRO concept creation
FALL IN BATHROOM WHILE ASSISTED
HOSPITAL 4
HEALTH CENTER 1
25
Interpretation examples
HIRO
Hospital specific terminology
26
Hospital
Incident report
Incident
HIRO Ontology
??
JCAHO Incident Data
JCAHO
Analysis
27
HIRO terminologies
ENTITY
META-ENTITY
DOMAIN-ENTITY
28
HIRO
PSET, CUSTOM Hierarchies..
Hospital Terminology
Instances one Incident Grouped by month /
hospital
29
HIRO
Hospital Terminology
Manual Mapping
30
HIRO
PSET, CUSTOM Hierarchies..
Hospital Terminology
Manual Mapping
Instances one Incident Grouped by month /
hospital
31

• Query the HIRO Ontology
• Straightforward queries
• All dose omissions
• All adverse Drug reactions
• All falls
• All assisted falls

HIRO

• NOT straightforward
• All medication errors

Instances one Incident Grouped by month /
hospital
32
HIRO
OR Use the ontology

• Query Medication Errors?
• Identify all HIRO concepts denoting a medication
  error.
• Use the Ontology
• Use a custom made framework (- medication error
  portal on HIRO) to ease querying

33
HIRO
PSET, CUSTOM Hierarchies..
Instances one Incident Grouped by month /
hospital
PSET Fully developed perspective Analyses HIRO
items in terms of cause, type, domain, etc GI -
GENERAL INCIDENT REPORTING TERMINOLOGY Small
Example framework Analyses HIRO items in terms
of type of incident (one of 6) persons
involved
34
Internal report (s)
Incident

• 2. Loss of information
• Attempt to categorize into type / cause / outcome
• Hospitals view on what needs and
• what not needs to be reported

• 1. Loss of information
• Incomplete Reporting
• Very concise terminology
• Data that does not fit goes into narratives
• Primitive terminology
• Might not take into account all necessary
  variables

Data sent to JCAHO
3. General Loss of information administration
HIRO
PSET
35
A Solution (utopia?) Standardized reporting
terminology system
with
without
HIRO
HIRO
Standardized Terminology software support
Loss of information Level of detail very
variable Effort to map all reports
Loss of information can be prevented Level of
detail customizable Map only standardized
terminology
36
Questions?

• Thank you!