Validity of Danish register diagnoses of myocardial infarction and stroke against experts in people with screen-detected diabetes

Publikation: Bidrag til tidsskrift/Konferencebidrag i tidsskrift /Bidrag til avisTidsskriftartikelForskningpeer review

DOI

BACKGROUND: Administrative patient registers are often used to estimate morbidity in epidemiological studies. The validity of register data is thus important. This study aims to assess the positive predictive value of myocardial infarction and stroke registered in the Danish National Patient Register, and to examine the association between cardiovascular risk factors and cardiovascular disease based on register data or validated diagnoses in a well-defined diabetes population.

METHODS: We included 1533 individuals found with screen-detected type 2 diabetes in the ADDITION-Denmark study in 2001-2006. All individuals were followed for cardiovascular outcomes until the end of 2014. Hospital discharge codes for myocardial infarction and stroke were identified in the Danish National Patient Register. Hospital medical records and other clinically relevant information were collected and an independent adjudication committee evaluated all possible events. The positive predictive value for myocardial infarction and stroke were calculated as the proportion of cases recorded in the Danish National Patient Register confirmed by the adjudication committee.

RESULTS: The positive predictive value was 75% (95% CI: 64;84) for MI and 70% (95% CI: 54;80) for stroke. The association between cardiovascular risk factors and incident cardiovascular disease did not depend on using register-based or verified diagnoses. However, a tendency was seen towards stronger associations when using verified diagnoses.

CONCLUSIONS: Our results show that studies using only register-based diagnoses are likely to misclassify cardiovascular outcomes. Moreover, the results suggest that the magnitude of associations between cardiovascular risk factors and cardiovascular outcomes may be underestimated when using register-based diagnoses.

OriginalsprogEngelsk
TidsskriftB M C Public Health
Vol/bind19
Nummer1
Sider (fra-til)228
Antal sider8
ISSN1471-2458
DOI
StatusUdgivet - 22 feb. 2019

Se relationer på Aarhus Universitet Citationsformater

ID: 148229739