Probabilistic record linkage has been used increasingly to identify outcomes in cohort studies. This study aimed to assess the method's accuracy for identifying deaths in a cohort of 450 patients admitted to a university hospital for decompensated heart failure over a six-year period. Vital status of cohort members was determined from electronic patient file data (gold standard). OpenRecLink software was used to link cohort records with those from the Mortality Information System, aimed at identifying deaths. Only 53.6% of patients had vital status known at the end of follow-up, and 59.3% of these had died. The method showed 97.9% sensitivity, 100% specificity, 100% positive predictive value, 97% negative predictive value, and 98.8% accuracy. The results suggest probabilistic record linkage as a valuable tool for identifying deaths in cohort studies.