Abstract WP221: A Pragmatic Computer Algorithm Successfully Matches De-identified Regional Quality Improvement Database Records and Emergency Medical Services Records

    loading  Checking for direct PDF access through Ovid

Abstract

Introduction: Linking prehospital data from emergency medical services (EMS) patient care records with hospital-based data in quality improvement databases, such as the American Heart Association’s Get With the Guidelines (GWTG)-Stroke database, can help inform initiatives that aim to improve both prehospital and hospital-based acute stroke care. However, inconsistent data collection, discrete systems, and privacy concerns prevent simple data merging of prehospital and hospital-based databases.

Hypothesis: A simple, pragmatic computer algorithm using probabilistic matching can successfully link EMS data with hospital-based quality improvement data.

Methods: A retrospective pilot study was performed that matched hospital data from GWTG-Stroke, a regional quality improvement database, and prehospital data from a single municipal fire-based EMS provider agency that responds to all 9-1-1 calls in a large US city with 15 primary and comprehensive stroke centers participating in GWTG-Stroke data collection. Using data for patients with confirmed stroke arriving via EMS from July to December 2013, we implemented a rule-based probabilistic matching algorithm that incorporated patient age, sex, time of hospital arrival +/- 30 minutes, and destination hospital to match de-identified records between the GWTG-Stroke and EMS databases. A subset of records at one stroke center was audited to verify successful matches. Python (Python Software Foundation, Beaverton, Oregon) was used to facilitate analysis.

Results: Among 328 patients with confirmed stroke arriving by EMS, a probabilistic matching algorithm successfully linked the prehospital and hospital records for 300 (91%). The 28 unmatched records were due to typographical errors or missingness of entered data. Of 40 algorithm-matched records audited at one stroke center for accuracy, 40 (100%) had been matched correctly.

Conclusions: A simple algorithm using patient age, sex, time of hospital arrival, and destination hospital successfully matched greater than 90% of de-identified prehospital and hospital records. Leveraging high fidelity database matching can facilitate future work that links prehospital and hospital stroke interventions to patient outcomes.

Related Topics

    loading  Loading Related Articles