The Tayside Stroke Cohort: exploiting advanced regional medical informatics to create a region-wide database for studying the pharmacoepidemiology of stroke


  • The authors have no conflict of interest to declare.



Observational studies can provide valuable information where data from randomised controlled trials is lacking. We aimed to generate a region-wide longitudinal register of stroke patients using record-linkage of existing clinical and research datasets.


The population were residents of Tayside, Scotland from 1994–2005. Stroke cases were identified from hospital inpatient admission records, death certificates and prescribing data. These were augmented with data derived from free-text CT-brain scan reports. Strokes were classified as intracerebral haemorrhage (ICH), subarachnoid haemorrhage (SAH) or ischaemic stroke (IS). The methodology was validated by audit of patient case-records. The incidence was calculated using direct standardisation to the standard European population for ages 45–84. Twenty-eight day case-fatality rates were calculated as percentages.


There were 12 620 all-cause incident strokes (ICH 1057; SAH 511; IS 6257; undetermined 4795). Standardised incidences per 100 000 by subtypes of stroke were: ICH 35 (95%CI 33–38), SAH 19 (17–22) and IS 210 (205–217). The 28-day case-fatality rates were: ICH 49% (95%CI 46–52), SAH 38% (34–43) and IS 19% (18–20). Comparisons with previous studies were favourable for ICH and SAH. For IS the incidence was lower and fatality rate higher than elsewhere. Three hundred and three sets of patient case records were audited. The positive predictive value (PPV) for identifying cases of stroke was 94.7% (95%CI 91.6–96.7).


The case ascertainment in the TSC compares favourably to established stroke cohorts. This cost effective resource can now be linked with multiple other clinical and research datasets in Tayside to further understanding of stroke and its treatment. Copyright © 2010 John Wiley & Sons, Ltd.