Acurácia do relacionamento probabilístico na avaliação da alta complexidade em cardiologia

Abstract

OBJECTIVE:To evaluate the viability of a probabilistic record linkage strategy to identify deaths of patients who underwent complex cardiology procedures among the total deceased population. METHODS:The processing cost was estimated based on 1,672 records of patients undergoing coronary artery bypass grafting that were compared with all death records in Brazil in 2005. The accuracy of the relationship was based on the probabilistic linkage of 99 hospital admissions records of patients, with known vital status, who underwent cardiac surgery at a single cardiology institute, with the death records of the state of Rio de Janeiro, Southeastern Brazil, in 2005. Linkage was conducted in four stages: standardizing the databases, blocking, matching, and rating peers. Blocking in fi ve steps was used, with blocking keys formed by a combination of variables such as soundex codes for the fi rst and last names, sex, and year of birth. The variables used for pairing were "full name" with the use of Levenshtein distance and "birth date". RESULTS:The second and fi fth blocking steps resulted in the largest number of formed pairs and the largest processing times for the matching. The fourth step required a lower processing cost. In the accuracy study, after fi ve blocking steps, the sensitivity of the linkage was 90.6%, and the specifi city was 100%. CONCLUSIONS:The probabilistic strategy used has high accuracy and can be used in studies of the effectiveness of high-complexity, high-cost cardiology procedures.

Accuracy of a probabilistic record-linkage methodology used to track blood donors in the Mortality Information System database

“…In Brazil, it has been used increasingly for studies involving the health information system databases managed by the Brazilian Ministry of Health, including the Mortality Information System (SIM). The process is used not only to detect and remove duplicate records relating to the same individual from within a database 1 but also to integrate records across health information system databases and in between them and other databases 2,3,4,5,6,7 . By doing so, these studies help to improve data quality and integrity, allow for the reuse of existing data that can be used for answering a broader range of study questions, reduce costs and efforts related to data acquisition and also help to improve record linkage methodology and its assessment 8 .…”

Inclusão de etapa de pós-processamento determinístico para o aumento de performance do relacionamento (linkage) probabilístico

“…O ganho de performance das medidas de desempenho em todas as estratégias quando acrescidas da EPPD já era esperado, uma vez que se trata de um acréscimo metodológico para contrapor as deficiências do OpenRecLink, que tende a apresentar um número elevado de falso-positivos com escores intermediários 11,13 e um pequeno, mas relevante, número de falso-negativos com escores baixos 27 .…”

“…É notória a proporção de registros com nomes abreviados no SINAN, 1 a cada 14, bem como a diferença em relação ao SIM, 1 a cada 302, que foi encontrada no estudo (Figura 1). Embora as abreviações de nomes fossem descritas em outros estudos como erro de preenchimento que dificulta o processo de relacionamento de dados 27,30,31 , a frequência de nomes com abreviação não foi informada.…”

Estratégias usadas no relacionamento entre Sistemas de Informações em Saúde para seguimento das mulheres com mamografias suspeitas no Sistema Único de Saúde

“…In Brazil, several studies have applied the probabilistic linkage technique using the software RecLink ®16-23 , and some report satisfactory sensitivity and specificity [17][18][19] . It is important, however, to consider that the accuracy of data linkage is related to the number of identifiers to be compared, as well as to the quality of their completion 17 .…”