Comparison of different procedures to identify probable cases of myocardial infarction and stroke in two Swedish prospective cohort studies using local and national routine registers.

In prospective cohort studies, person-time time is calculated from baseline until the first definite event occurs or until censoring. A way to correctly identify and date definite events when only routine registers are available is to retrieve all hospital discharge notes and death certificates with a diagnosis of probable event and perform a consecutive revision. It is important to detect all possible hospital stays as they may contain useful information for the revision study. Furthermore, loss to follow-up can be avoided by extending the retrieval outside the specific geographical area where the cohort was defined. The aims of this study were (i) to describe a comprehensive retrieval of probable myocardial infarctions (diagnosis with International Classification of Diseases 8th and 9th revisions codes 410-414) or stroke (codes 430-438), (ii) to quantify the relative efficiency of different local and national routine registers or their combination compared with the use of all available registers together, and (iii) to audit local and national registers by comparing their outcome at the county level. The study was performed in two prospective cohorts studies i.e., 'Men-born-1914' (n = 500) from Skåne (period 1982-1993), and Skara-1 (n = 683) from Skaraborg (period 1988-1994.). All available routine registers were linked to the cohorts. The use of all available routine registers improved retrieval of both individual and hospital stays with a discharge diagnosis of probable event and gave an enhanced basis for a future validation study. Local registers were not completely covered by the national register, but the accessible combination of national inpatient and mortality registers was an efficient alternative.