Procedures for complete case capture follow international recommendations for capture–recapture and multiple source ascertainment methods.20 (link) Cases are primarily identified through active pursuit of emergency department and directly admitted stroke patients using validated screening terms.21 (link) The abstractors also routinely canvass intensive care units and hospital floors searching for in-house strokes or those not ascertained through the screening logs. The active surveillance is supplemented by review of hospital passive listings of International Classification of Disease, 9th revision discharge codes for stroke (430–438; excluding 433.x0, 434.x0 [x = 1–9]; 437.0, 437.2, 437.3, 437.4, 437.5, 437.7, 437.8, and 438). County coroner records are screened for causes of sudden stroke death not presenting to the hospital. Several minor changes to case ascertainment procedures were made over the course of the project to maintain efficiency. In 2001, the following diagnostic terms were removed from the active surveillance list: dizziness, falling, imbalance, syncope, and trouble walking. These terms were found to be highly inefficient at identifying strokes, with a positive predict value of ≤1%. Starting January 1, 2001, a sample of Nueces County primary care and cardiology offices, as well as 95% of neurologist offices, were contacted frequently and encouraged to report stroke cases to our project. Abstractors reviewed and abstracted cases not previously screened. The sampling of primary care physicians and cardiologists was subsequently discontinued on January 31, 2004, because during the 3 years of the sampling only 13 ischemic stroke patients were identified exclusively from this method of 1,866 ischemic stroke cases identified in BASIC. Because 74 cases from the sample identified came from neurology offices, we did continue to identify these few stroke cases from neurology offices. From January 31, 2004 to July 31, 2008, only 71 strokes were identified via the neurology office of 1,971 ischemic stroke cases identified. In 2008, we therefore stopped screening neurology offices. From January 1, 2000 through December 1, 2007, BASIC identified cases through active surveillance of both the admissions log and emergency department (ED) log. A review of this methodology in 2007 using complete data from calendar year 2004 suggested that frequent passive ED surveillance in combination with active surveillance of admission logs successfully identifies ≥98% of all ischemic strokes. This new methodology was implemented on December 2, 2007. Finally, we were unable to obtain passive listings of stroke from 1 of the hospital systems for 6 months of 2008. In other 6-month periods, this never amounted to >5 cases. A sensitivity analysis was performed to determine the effects of the changes on incidence rate estimates and ethnic comparisons over time.
Cases are validated by neurologists or a stroke fellowship-trained emergency medicine physician, blinded to subjects’ ethnicity and age, using source documentation. Ischemic stroke diagnosis is based on published international clinical criteria20 (link) that require onset of a focal neurologic deficit following a defined vascular distribution without documented resolution within 24 hours (unless treated with recombinant tissue plasminogen activator) and not explainable by a nonvascular etiology. Imaging is used to discriminate ischemic stroke and hemorrhagic stroke. Because the use of brain MRI has increased greatly in the past 10 years, validators are required to use the original clinical criteria for case validation, so that trend data can be assessed without bias. Therefore, subjects having acute infarction on brain MRI without the clinical deficit described above are validated as no stroke.