Avaliação do rastreamento do câncer de mama em Minas Gerais a partir de dados do SISMAMA

Abstract

This thesis aims to evaluate breast cancer screening in Minas Gerais according to data from SISMAMA and national guidelines. Three studies were conducted, and results are presented as articles. Probabilistic data linkage was one of the main techniques used in the thesis. According to size of linked databases, the procedure may lead to multiple pairs, hindering manual conference to identifying a real pair. Thus, the first study created an automated method to select scores and identify pairs in probabilistic data linkage. Sixteen steps linked databases of altered exams and references. Every step was manually verified to obtain a gold standard, and samples were selected to calculate accuracy of the method. The method simplified and facilitated identification of real pairs and found good quality of assessment measures from linkages performed. The second study estimated mammographic screening in the state and health microregions and associated factors were evaluated. Duplicates and repetitions were removed by deterministic and probabilistic methods to calculate screening coverage. Associations between coverage and other variables were analyzed using linear regression model. Estimated coverage for the population of mammographic screening was 36.3%. In microregions, coverage varied from 2.66% to 61.70%. Higher coverage occurred in micro-regions with better socioeconomic development levels and low health vulnerability (central and south) ; north and northeast micro-regions presented lower coverage since they are less developed and have higher health vulnerability. A positive association was found between coverage and income, number of mammograms, and capacity of using mammograms. In contrast, analysis showed a negative association with Gini Index, indicating the higher the mammography coverage, the higher the income, number of mammograms, and capacity of using this equipment, and the lower the income inequality. In the third study, over-screening was evaluated in a non-concurrent cohort of registers in SISMAMA using probabilistic data linkage between the reference database and follow-up. Percentage of over-screening was estimated by age (exams in women aging < 50 or ≥ 70 years) and excessive periodicity (repetition of the exam in a period lower than 18 months). Associations between over-screening and sociodemographic and assistance variables were analyzed using regression model in an ecological study using 77 health microregions of the state. Over-screening by age and periodicity reached 47.1% and 27.8% of women, respectively. The latter was associated with income, number of mammograms, and low-income inequality. No associations were found between over-screening by age and variables studied. Micro-regions with low coverage, number of human resources, and equipments performed over-screening, demonstrating the need for adequacies. We conclude that mammography coverage between 50-69 years in Minas Gerais was low in the period studied , and recommendations by age and periodicity were not adopted. Screening actions require improvements and adequacies focusing on resource distribution according to characteristics and needs of each region, organized and optimized screening actions, awareness of the general population, and professional training regarding evidence that support guidelines adopted in Brazil.