Validation of ischemic cerebrovascular diagnoses in the health improvement network (THIN)

Pharmacoepidemiol Drug Saf. 2010 Jun;19(6):579-85. doi: 10.1002/pds.1919.


Purpose: To evaluate the validity of recorded diagnoses of ischemic cerebrovascular events requiring hospitalization within The Health Improvement Network (THIN) UK primary care database.

Methods: We identified 15 397 individuals aged 40-84 years with a first recorded ischemic event in 2000-2004. Of these, 4239 had a code suggestive of a hospitalization within 2 weeks of the event. A three-step strategy was used to validate the records of these patients: manual review of computerized medical records excluding free-text comments; manual review including free-text comments (which include information gained from specialists, hospital discharge letters and results of diagnostic tests) of a random sample of possible cases (n = 300) and non-cases (n = 100); and review of full medical records of this random sample and a questionnaire completed by their primary care physician. The positive predictive value (PPV) of each step was calculated. The confirmation rate was used to estimate incidence in the general population.

Results: After step 1, 3447 individuals were classified as possible cases and 792 were excluded as non-cases. After step 2, 82% of possible cases were still classified as such. Step 3 showed that inclusion of free-text comments increased the PPV of a diagnosis from 76 to 86%. The weighted incidence of hospitalized ischemic cerebrovascular events was 1.73 per 1000 person-years (95% CI:1.68-1.77).

Conclusions: THIN demonstrates a high validity for the study of ischemic cerebrovascular events when reviewing computer records with additional free-text comments. Accuracy of hospitalization status was not as well recorded.

Publication types

  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Adult
  • Aged
  • Aged, 80 and over
  • Brain Ischemia / diagnosis*
  • Brain Ischemia / epidemiology
  • Databases, Factual
  • Epidemiologic Methods*
  • Hospitalization / statistics & numerical data
  • Humans
  • Incidence
  • Medical Records Systems, Computerized / statistics & numerical data*
  • Middle Aged
  • Predictive Value of Tests
  • Primary Health Care / statistics & numerical data
  • United Kingdom / epidemiology