Agreement and completeness of data on live births and infant deaths

Objective: To assess the quality of data (agreement and completeness) on infant deaths in the Mortality Information System (SIM) and in the Information System on Live Births (Sinasc), Recife, Pernambuco, Brazil. Methods: Cross-sectional study with data on infant deaths captured in Sinasc and SIM. For the deaths, the period 2013-2016 and the live births of 2012-2016 were used. The deterministic linkage was used. The percentage of incompleteness of 10 variables common to both bases pre- and post-linkage was calculated. The agreement was assessed by the Kappa index for qualitative variables and by the intraclass correlation coefficient (ICC) for the quantitative variables. Results: It was possible to relate 96.64% of the deaths to their respective declaration of live birth. All analyzed variables were classifi ed as excellent (less than 5% incompleteness), pre- and post-linkage. In Sinasc, the greatest incompleteness was in the variable length of pregnancy (1.55%) and in the SIM, the number of stillbirths (2.89%). The agreement was classified as almost perfect for all qualitative variables (Kappa between 0.8 and 1). All quantitative variables were excellent (ICC greater than 0.75). Conclusion: Despite advances in the quality of SIM and Sinasc, there were still problems of completeness of variables, especially in SIM. The linkage contributed to the improvement of information for the analysis of infant deaths by health services and for research. It is a technique that is easy to access and low operational cost, which can be included in the routine of infant mortality surveillance for the continuous improvement of information.