An Efficient Approach To Map LOINC Concepts To Notifiable Conditions

1
An Efficient Approach To Map LOINC Concepts To
Notifiable Conditions

• Wei Li, MD1, Jerome I. Tokars, MD, MPH1,
• Nikolay Lipskiy, DrPH1, Sundak Ganesan, MD2
• 1. Division of emergency Response and
  Preparedness
• National Center for Public Health Informatics
  (NCPHI)
• Centers for Disease Control and Prevention (CDC)
• 2. SAIC Consultant to NCPHI, CDC

Disclaimer The findings and conclusions in this
presentation are those of the author(s) and do
not necessarily represent the views of the
Centers for Disease Control and Prevention/the
Agency for Toxic Substances and Disease Registry.
2
Notifiable Diseases/Conditions

• Surveillance for notifiable conditions
  traditionally includes manual reporting using
  written case definitions
• The availability of laboratory data in electronic
  format makes automated reporting feasible
• Mapping of notifiable conditions to standard
  vocabularies are required
• A table mapping notifiable diseases to LOINC
  (laboratory) and SNOMED (organism) codes was
  first made by Dianne Dwyer in later 1990s and
  last updated in 2004

3
Purposes of Mapping

• A tool to filter the messages from public health
  and clinical labs to identify laboratory test
  results of public health importance
• BioSense receives laboratory test results from 30
  hospitals and from a large national
  laboratoryneed to automatically identify results
  that MAY indicate a notifiable condition
• 43 pathogens that cause notifable conditions are
  also Category A, B, or C Bioterrorism Agents
  (73)
• Electronic laboratory report (ELR) and
  interoperability
• A framework for the development and maintenance
  of a controlled vocabulary for reportable events
  of public health importance
• Developing reusable component for intelligent
  surveillance information system architectures
• Towards development of notifiable disease
  controlled vocabulary for Unified Medical
  Language System (UMLS)

4
Objective

• To develop a method to map Logical Observation
  Identifier Names and Codes (LOINC) to notifiable
  conditions
• Efficient
• Easy to update

5
Notifiable Diseases and Other Conditions of
Public Health Importance

• Notifiable Diseases and Other Conditions of
  Public Health Importance (n152)
• Reportable either nationally (n86) or to states
• Active and inactive
• Infectious diseases, injuries, toxins

6
Event Code List Notifiable Diseases and Other
Conditions of Public Health Importance
Bolded indicating NND and should be reported to
CDC on a regular basis
7
Event Code List

• Event Code changes year to year
• Retire and replace
• Enterohemorrhagic Escherichia Coli (EHEC) O157H7
  (10560,11562, 11564) were added in 1994 but
  retired in 2006
• Shiga toxin-producing Escherichia Coli (STEC) was
  added in 2006
• New code
• Polyovirus infection, nonparalytic (10405) was
  added in 2007.

8
Logical Observation Identifiers Names and Codes
(LOINC)

• A standard vocabulary identifies laboratory tests
  and clinical observations
• A common language for building electronic medical
  records
• Founded in 1994 and initial release in 1995 with
  6,000 lab tests
• The latest release (v2.21) covered 48,043 terms
• Allow users to merge clinical results from many
  sources into one database for patient care,
  clinical research, or management.

Regenstrief Institute
9
LOINC to Notifiable Conditions Mapping

• LOINC to Condition Mapping Tables
• Component of Public Health Information Network
  (PHIN)
• Original version by Dr. Dwyer, updated by Dr.
  Sable
• Last updated by PHIN in 2004
• LOINC
• There are 46,812 terms in the LOINC version 2.19,
  which revealed 11,972 new terms added (34
  increase) since 2004
• 12,000 terms have been deprecated since then
• LOINC has changed the properties of its database
  and expanded its fields to 61

10
Mapping Methods

• List of 152 Notifiable Diseases and Other
  Conditions of Public Health Importance
• Excluded events without lab test or unknown
  organism
• Silicosis, Spinal cord injury, and Head Injury,
  etc
• Kawasaki Disease, Reye Syndrome, etc
• List to be mapped includes 146 conditions
• All nationally notifiable infectious diseases
• 46 Other infectious diseases
• 7 Toxins
• Determine organisms related to each disease
• Find Systematized Nomenclature of Medicine
  (SNOMED) codes for organisms
• Map tests for organisms/toxins to LOINC codes by
  3 methods

http//www.cdc.gov/epo/dphsi/phs/infdis2007.htm
11
Mapping Methods (continued)

• The Regenstrief LOINC Mapping Assistant (RELMA
  v3.19) was used to identify lab tests and LOINC
  codes associated with each organism/toxin
• A SAS (v9.1.3) program was developed to perform a
  text search of the LOINC database to find words
  that matched notifiable conditions or
  microorganisms
• Manual process

12
Mapping Methods (continued)

• Manual process. Check the following LOINC
  database fields against the list of notifiable
  conditions
• Component
• Method Type
• Class
• Date Last Changed
• Map To
• Short Name
• Related Names 2

13
RELMA v3.19
14
Results Comparison of Mapping Tables
15
Comparison of Mapping Methods
16
Text Search

• The text search parsing program found only 4901
  LOINC
• 451 LOINC entries not found by text search most
  had a non-specific test name, e.g.
• Code 33700-6 (a spore identification test) is
  used to identify Anthrax
• Search field does not contain the word that used
  for searching, e.g. anthrax

17
LOINC to Condition Table Example Anthrax mappings

• LOINC has several Scale types
• Quantitative scale (QN) numerical results
• Ordinal scale (ORD) Positive, Negative, 1,
  2, Not detected, etc.
• Nominal scale (NOM) The name of an organism

18
Product LOINC to Condition Table Data Dictionary
Copied from LOINC database
19
Product LOINC to Condition Table Data Dictionary
Copied from LOINC database
20
Conclusions

• Compared with the 2004 LOINC to Condition Table,
  our 2007 table added 2000 entries and deleted
  200.
• The methods we used were labor intensive but
  currently necessary
• Efforts to improve the text search method by
  adding additional search terms such as spore
  may enable more frequent automated updates.
• In the interim, the use of RELMA to identify lab
  tests provides an accurate and efficient
  semi-automated process to update the LOINC to
  Condition Mapping Table

21
Future Plans

• Update LOINC to Notifiable Conditions Mapping
  Table to latest LOINC version 2.21
• Re-create/modify the text search program to
  increase accuracy
• Explore programs other than SAS to perform text
  searches
• Incorporate terms in the latest release of SNOMED
• Ultimately, use updated mapping table to improve
  laboratory-based surveillance

22
Wei Li for5_at_cdc.gov