Improving Efficiency of Stroke Research: The Brain Attack Surveillance in Corpus Christi Study
Stroke, Epidemiology, Screening, Surveillance, Computer, Clinical
We studied whether a computer algorithm or abstractor could diagnose stroke as well as a fellowship-trained stroke neurologist. As part of an ongoing prospective, community-based stroke surveillance project, a diagnostic algorithm was developed, and patients' neurologic signs and symptoms were collected in a computerized database. The abstractors were blinded to the results of this algorithm and were asked to verify whether the patient had a stroke. The separate results of the computer and abstractor were compared with the final diagnosis given by the blinded neurologist. From 1 January through 31 July 2000, 3418 cases were screened. The abstractors yielded sensitivity 91%, specificity 97%, positive predictive value (PPV) 85%, and negative predictive value (NPV) 99%. Three computer algorithms were evaluated. The sensitivities ranged from 83% to 96%, specificity ranged from 88% to 97%, PPV ranged from 54% to 81%, and NPV ranged from 97% to 99%. The use of computer verification or abstractors may obviate the need for physician stroke verification and may greatly improve study efficiency.
Digital Object Identifier (DOI)
Citation / Publisher Attribution
Journal of Clinical Epidemiology, v. 56, issue 4, p. 351-357
Scholar Commons Citation
Al-Wabil, Areej; Smith, Melinda A.; Moyé, Lemuel A.; Burgin, W. Scott; and Morgenstern, Lewis B., "Improving Efficiency of Stroke Research: The Brain Attack Surveillance in Corpus Christi Study" (2003). Neurology Faculty Publications. 65.
Was this content written or created while at USF?