ABC | Volume 111, Nº4, Octuber 2018

Original Article de Souza e Silva et al Percutaneous coronary intervention in the State of Rio de Janeiro Arq Bras Cardiol. 2018; 111(4):553-561 Methods Study population and data collection Data on PCI obtained at administrative databases of the state of RJ were analyzed retrospectively. The  DATASUS administrative database of Authorization for Hospital Admission ( Autorização de Internação Hospitalar - AIH ) was consulted to gather data on PCI performed in public or private hospitals paid by the SUS between 1999 and 2010. SUS is the Brazilian public healthcare system. It is funded from general government revenues, it is single, universal, hierarquical and integrated. 14 DATASUS contains data of the Department of Healthcare Information of the Brazilian Ministry of Health, and it manages SUS ’ healthcare and financial information. 15 AIH is a registry system 16 for any admissions that occurs in any public or private hospital that maintain a convenant with the SUS . Patient inclusion criteria : people who lived in the state of RJ, ≥ 20 years old, submitted to one single PCI between 1999 and 2010. Patient exclusion criteria: individuals submitted to coronary artery bypass grafting during the study period. From the AIH database were obtained patients’ name, date of birth, hospital admission and discharge, sex, address, mother´s name and type of PCI. PCI procedures were classified according to the AIH database codes as described in a previous study 9 as follows: a) PCI without stent placement (PCI-WS); b) PCI with stent placement (PCI-S); and c) primary PCI (PCI-P). During the study period the SUS would not pay for drug-eluting stents; therefore, PCI-S refers to the use of bare-metal stents. The post-procedure outcome was death from any cause, and information on patients’ death was obtained at the death database of the state of RJ from 1999 to 2014. In order to match information from both databases, AIH and deaths, Stata®14 probabilistic record linkage (Reclink) was used, once there is no common identification field between these two databases, and this essentially consists of a fuzzy merge. This method allows matching weights for each pre-defined variable, thus creating a new variable to hold the matching score in a zero-to-one scale, which indicates the probability that the pairs formed refer to the same patient. The pre-defined variables were patient´s name, date of birth and sex. Pairs that scored = 1.00 (perfect matches) were considered the same patient. Pairs that scored ≥ 0.99 and < 1.00 were considered possible matches and were manually reviewed using mother´s name and address to define whether or not they were going to be considered the same patient. Pairs with lower scores were considered a “non-match”. In order to test the sensitivity and specificity of the probabilistic linkage method used, in-hospital deaths found at the AIH database were compared to the matching information from the death database. Out of a total of 357 in-hospital deaths found at the AIH database, 307 were found with the linkage process with the death database, and no false positives were detected. Therefore, the estimated sensitivity and specificity were 86% and 100%, respectively. After the linkage process, patients were classified according to sex and the age groups 20-49, 50-69 and ≥ 70 years old. Underlying causes of death were obtained at the death database and classified according to the 10 th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) 17 as IHD (codes I20 to I25) or non‑IHD (any other code). As the AIH database contains no information about the exact date of the PCI procedure, only the date of the patients` hospital admission and discharge, and as the average stay of these patients was 2 days, 9 to analyze the survival rate the discharge date was considered day one. Short- and medium‑term survival rates were defined as the probability of survival until day 30 and one year after discharge, respectively. As there are two possible discharge types at the AIH database – hospital discharge or death - short-term outcomes included in-hospital mortality rates. Long-term survival was defined as the probability of survival up to 10 or 15 years after hospital discharge for comparisons among types of PCI or between age groups and sex, respectively. The study was approved by the ethics committee of Hospital Universitário Clementino Fraga Filho ( Faculdade de Medicina – UFRJ ) on 10/18/2012 (1148/12). Statistical analysis Statistical analysis was performed based on data distribution. As the Shapiro-Wilk and Kolmogorov-Smirnov tests showed that age was not normally distributed, age distributions were described as median and interquartile ranges (P25-P75). Distribution of categorical variables was described as relative frequencies. The differences among groups were analyzed with the Kruskal-Wallis test for continuous variables or chi-square test for categorical variables. Probabilities of short-, medium- and long-term survival rates were estimated with the Kaplan-Meier survival method. Survival models were estimated with Cox proportional hazards regression to compare risks among age groups, sex and type of PCI; 95% confidence intervals (CI) were calculated to express the degree of uncertainty associated with the statistics for all analyses of subgroups. Stata 14 ® was used for all analyses. Test results with a p-value < 0.05 were considered statistically significant. Results Out of 22,735 patients, 3,472 were excluded and 19,263 were selected (63.6% men). Median (P25-P75) ages for men and women were 60 (52-68) and 62 (54-70) years, respectively (p < 0.05). The frequency distribution of the age groups 20-49, 50-69 and ≥70 years old for men and women was 16.2% and 13.1%, 63.9% and 60.1%, and 19.9% and 26.8%, respectively (p < 0.05). Minimum andmaximum follow-up were 4.0 and 15.0 years, respectively, and 5,433 patients (65.1% men) died during follow-up. Probabilities of survival and 95% CI for men and women were, respectively, short-term: 97.3% (97.0-97.6%) and 97.1% (96.6-97.4%), medium‑term: 93.6% (93.2-94.1%) and 93.4% (92.8-94.0%), and long-term: 55.7% (54.0-57.4%) and 58.1% (55.8-60.3%). Men aged 20-49 years tended to have higher probability of survival in a 9-year follow-up, after which this tendency would reverse (Table 1). Men and women aged 50-69 years had the same probability of survival in a 180‑day follow‑up, after which women tended to have a higher 554