|
BACKGROUND/RATIONALE:
We propose an innovative approach to use data from the electronic medical record (EMR) that draws from the field of knowledge discovery in databases (KDD). KDD techniques search large amounts of text-based data to extract structured information. We will compare several KDD techniques on their ability to identify and characterize circumstances (place and mechanism of fall) of falls that are treated in the VHA ambulatory care setting. A fall related ambulatory event is defined as "a situation in which a veteran seeks treatment for a fall-related injury (FRI) in the ambulatory care setting". Falls are an important health care issue especially among aging veterans. A history of a previous fall (a consistent predictor of future falls) is the single most important clinical indicator that identifies an elderly patient as high risk for additional falls and targets them for fall prevention programs. Our pilot data suggested significant under-coding of information about FRAE in administrative databases thereby limiting information about a history of falls to clinicians. This gap significantly reduces the value of the largest sources of ambulatory data in the work to researchers and administrators. Our pilot work further suggested that KDD has the potential of automating data extraction from the EMR providing important information to clinicians and researchers.
OBJECTIVE(S):
The goal of this study is to determine the usefulness of KDD strategies to explore the VHA databases to better identify veterans who sustain injurious falls and in predicting injurious falls. The immediate objectives of this two-year retrospective cross-sectional study are to: 1) Create a benchmark dataset on which KDD analyses will be conducted; 2) Compare the ability of KDD techniques (regular expression-based pattern matching and text mining) to identify fall-related ambulatory events with three types of data (text-based notes alone, text-based notes plus information from administrative data, and text-based notes plus information from chart review) using area under the receiver operating characteristic (ROC) curve analyses; 3) Test the generalizeabilty of the results from VISN8 in data from other VAMCs; and 4) Apply the KDD method found to be most highly predictive in O2 to identify mechanism and place of injury associated with FRAE.
METHODS:
This two-year retrospective cross-sectional study will be accomplished in two phases. In Phase 1 we will compare the results of two types of KDD analysis (regular expression-based pattern matching and text mining) under three conditions; 1) text-based information alone, 2) text-based information plus information from the administrative record, 3) text-based information plus information from chart review. The results can be used to determine the effectiveness and efficiency of KDD for extracting clinical information given increasing level of effort needed to create benchmark data. In Phase 2 the generalizeabilty of the resultant models to identify FRAE will be examined by replicating the analysis in data from other VAMCs. In addition we will extend our analyses to determine if the KDD can identify more specific information concerning the mechanism of a fall and place in which the fall occurred.
FINDINGS/RESULTS:
Currently there are no findings to report.
IMPACT:
The long-term goal of this study is to provide an improved measure of the impact of FRAE in the VHA. Capitalizing on the continued development of the centralized data resources of the Health Data Repository (HDR) we will work with the National Center of Patient Safety to develop VISN/facility level reports that accurately describe FRAE, resources expended to treat FRAE and identify high risk groups fall prevention programs. Additionally applications can be written to allow clinicians and researchers to search the EMR to identify patient at high risk. In a more general sense, the proposed research addresses data quality. The methods described here provide a means for automating data extraction traditional only possible through medical chart review. If successful these techniques could be applied to research and quality improvement programs across the VHA. As the VHA pursues projects such as the Health Data Repository and other large-scale data warehouses, issues of data quality become increasingly important. Clearly, as a leader in electronic medical record technology, the VHA is in a position to pursue such state-of-the-art initiatives.
PUBLICATIONS:
None at this time.
DRA:
Aging and Age-Related Changes
DRE:
Treatment, Diagnosis and Prognosis
Keywords:
Behavior (provider), Patient outcomes, Informatics
MeSH Terms:
none
|
