[Ethnic and racial iniquities in infant mortality: implications of changes in recording color/race in national health information systems in Brazil].

This descriptive study aimed to discuss the repercussions of the change in the methodology for recording the color/race variable in the Brazilian Information System on Live Births (SINASC) on infant mortality rates (IMR) according to color/race in Brazil. Annual variations were analyzed in the rates of live births and infant deaths according to color/race from 2009 to 2017. The IMR according to color/race were estimated using three strategies: (1) direct method; (2) for every year, setting the same proportions of live births by color/race as observed in 2009; and (3) for every year, setting the same proportions of deaths by color/race as observed in 2009. The strategies aimed to explore the single effect of the variations in the proportions of live births or of deaths according to color/race on the estimated IMR before and after the change in the color/race variable in the SINASC database. Between 2011 and 2012 (the year of the change in the color/race variable in SINASC), there was a sudden increase in birthdates with black, brown, and indigenous color/race, along with a reduction in birthdates with white color/race, without no corresponding variations in deaths. The increase of more socially vulnerable color/race categories in the IMR denominator resulted in the attenuation of IMR for black and indigenous infants and in an increase in the IMR for white infants and consequently an artificial reduction in iniquities in infant mortality according to color/race. The change in the color/race variable in SINASC interrupted the historical series of live births by color/race, affecting indicators that potentially depend on these data for their calculation, in this case the IMR. The resulting argument is that infant mortality rates by color/race before versus after the change in the SINASC database are distinct and noncomparable indicators.