|
BACKGROUND/RATIONALE:
Background/Rationale: Troops deployed to OEF and OIF missions are at risk of exposure to biological and chemical agents. Electronic surveillance of discharge medical examinations (BDD) and VA C&P examinations could provide early detection of signs and symptoms resulting from biological and chemical agents for which troops are at risk of exposure in the Middle East deployment areas. If successful, the approach may be adopted for in-theater biosurveillance.
OBJECTIVE(S):
Objectives: The potential signs and symptoms attributable to biochemical agents can inform a knowledge algorithm that can be used for surveillance of medical records to detect the syndromes of previously unsuspected etiologies of these agents.
Aim 1. Demonstration of biochemical surveillance relevant to exposures common to OEF/OIF troops.
Aim 2. Demonstration of state-of-the-art methodological and informatics tools for searching narrative text and extracting encoded concepts.
METHODS:
Methods: General medical examinations of EOF/OIF Troops will be aggregated into a database, included will be the discharge general medical examinations referred to as BDD and the C&P disability examination. The electronic C&P examination records conducted by VHA since 2003 will be aggregated into a database. The database will reside on the TREP Center for Health Service Research server behind the VHA firewall with IT security and protections.
Using a novel and powerful medical informatics search technique of concept-based indexing, the investigators will construct clinical statements that represent the specific signs and symptoms of interest. To complete this objective, we will assemble a team of specialists, internists, statisticians, and informaticians to construct the logic base and search algorithms for the clinical statements that define the syndromes associated with biological or chemical exposure agents, which would represent the clinical "threshold of evidence" required for a reviewer to feel confident that a key defining exposure is manifest.
The Multi-threaded Clinical Vocabulary Server (MCVS) has the capability to parse and index clinical notes. The MCVS uses the following terminology and ontology systems: the SNOMED-CT for diagnoses, clinical manifestations, and microbiology data, LOINC for lab results, NDF-RT and RxNorm for drugs. The MCVS system is based on 20 years of research. We will use the MCVS technology, through a service contract with MAYO clinic, to design an electronic biosurveillance system. This knowledge base will be the first deliverable associated with this proposal. Using concept-based indexing and an expert rules search engine, we will use these clinical statements to identify the outcomes of interests from within free text general medical examination records of OEF/OIF veterans.
To estimate the sensitivity and specificity of SNOMED-CT for detection of a particular cluster of signs and symptoms, a stratified random sample of records will be drawn: a) a sample of 1000 records from those that are complete "hits", i.e. that match every one of the signs and symptoms (Group 1) b) a sample of 500 of those records that contain all but one of the signs and symptoms of interest (Group 2) and c) a sample of 500 of those records that are missing 2 or more of the relevant signs and symptoms (Group 3), and a sample of 500 records that are missing all of the relevant signs and symptoms (Group 4). The "reference standard" categorization will then be made for this sample of records based on clinician review of the patient records. Two clinicians, blinded to group, will classify each case, disagreements will go to a third review for resolution. By using the stratified sampling strategy above, we will be able to estimate with greater precision the total number of "true positives" in the entire database, which is necessary to estimate sensitivity, specificity and likelihood ratios for each stratum. Standard errors and confidence intervals will be based on use of the binomial/multinomial distributions and on asymptotic calculations using the "delta-method." The outcome will comprise an ordinal likelihood classification for agent exposure that accounts for the signs and symptoms of the veteran.
Examples of agents (anthrax, smallpox, ricin, depleted uranium, amebiasis, tuberculosis) and their associated signs and symptoms that we will consider for OEF/OIF veterans are shown in the appendix. The maked displays demonstrate the output of the MVCS system that we will use in this project.
FINDINGS/RESULTS:
No results at this time.
IMPACT:
Impact: All troops deployed to the OEF and OIF missions have potential exposure to biochemical and toxic agents. If developed, biochemical surveillance tools could provide early warning of such exposures and enable preventive measures. This project addresses the high priority of deployment health research.
PUBLICATIONS:
None at this time.
DRA:
Military and Environmental Exposures
DRE:
Diagnosis and Prognosis, Technology Development and Assessment, Epidemiology
Keywords:
Decision support, Operation Iraqi Freedom, Bioterrorism
MeSH Terms:
none
|
