[Accuracy of the probabilistic record linkage methodology to ascertain deaths in survival studies].

Probabilistic record linkage methodology has been increasingly used to ascertain outcomes in cohort studies. However, only a few studies have evaluated its accuracy. The aim of this study was to evaluate the accuracy of probabilistic record linkage methodology to ascertain deaths in a cohort of 250 elderly people hospitalized for fractures caused by falls. The vital status of cohort members was determined through household visits conducted at one and six months after hospital admission. Record linkage was used to link records from the cohort to the Rio de Janeiro State mortality database. Deaths identified by the latter process were compared to the results obtained by active follow-up (gold standard). The vital status of 20 cohort members could not be determined by active follow-up. Among the remaining 230, 60 died within 180 days after hospital admission. Record linkage had a sensitivity of 85.5%, specificity of 99.4%, positive predictive value of 98.1%, and negative predictive value of 94.9%. In conclusion, probabilistic record linkage appears to be a useful tool for identifying deaths in cohort studies conducted in Brazil.