The purpose of the study was to compare the linkage parameter estimates between hospitalization and mortality databases, calculated separately for the subsets of deaths from ill-defined causes and deaths from known causes. The databases for deaths from known causes and ill-defined causes were linked to a hospital admissions database. Parameters were estimated using two strategies: (1) first name, last name, and day, month, and year of birth, (2) full name and date of birth. In the first strategy, the estimates for the first and last name were at least 97% in both sets. However, the items day, month, and year of birth produced low values in both sets. In the second strategy there was an important difference between the two groups, with much lower values for full name and especially for date of birth in the group of deaths from ill-defined causes. Our results emphasize the need for pilot studies to evaluate possible internal heterogeneity of databases during the planning stage of linkage projects.