The decline in mortality due to acute complications of diabetes mellitus in Brazil, 1991–2010

We employed secondary data to characterize national trends in mortality. We downloaded public domain demographic data available from the Brazilian Institute of Geography and Statistics (IBGE) and mortality data reported through Brazil’s Mortality Information System (SIM) for the years 1991 to 2010 [17]. Death from diabetes was defined by one of the following three-character codes of the International Statistical Classification of Diseases and Related Health Problems, 9th Revision (ICD-9): “250” (Diabetes mellitus) or 10th Revision (ICD-10): “E10” (Insulin-dependent diabetes mellitus), “E11” (Non-insulin-dependent diabetes mellitus), “E12” (Malnutrition-related diabetes mellitus), “E13” (Other specified diabetes mellitus), or “E14” (Unspecified diabetes mellitus). A death due to an acute complication of diabetes was defined as any death listing the above codes and also presenting the fourth-character subdivision “.1” (with ketoacidosis), “.2” (with hyperosmolarity), or “.3” (with other coma) in ICD-9 or “.0” (Coma with or without ketoacidosis, hyperosmolar coma, hypoglycemic coma, or hyperglycemic coma NOS) or “.1” (Ketoacidosis or acidosis without mention of coma) in ICD-10. We investigated mortality over the complete age range, and also before 40 years of age. The literature suggests that acute causes are relatively more frequent among diabetes deaths prior to age 40, are more preventable, and given the greater number of years of life lost with each death, contribute importantly to the disease’s burden [14‐16].

We used IBGE census (1991, 2000, 2010), IBGE population count (1996), or linear interpolation between the above-listed years to determine population size. The study period was confined to the years 1991 to 2010, as the former year was the first for which data necessary for determining completeness of death reporting were available, and the latter was the last for which mortality data were available.

Brazilian mortality registration has made steady process in completeness of reporting (estimated completeness – 1991: 77.6 %, 2011: 94.2 %): and definition of the cause of death (ill-defined deaths – 1991: 18.2 %, 2010: 6.7 %) over the past two decades [18‐20]. In trend analyses, corrections must be made in order to take greater past lack of completeness (deaths not notified) and more frequent ill-defined causes into account. Using an approach similar to that taken in key analyses of the Ministry of Health [21, 22], we corrected reported deaths for both estimated lack of completeness of death registration and for ill-defined causes of death (chapter XVI of ICD-9, “Signs, symptoms and ill-defined conditions” (codes 780–799), and chapter XVIII of the ICD-10, “Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified” (codes R00-R99).). The approach to these corrections has evolved over the years. We based correction for lack of completeness for 1991 on the estimated number of deaths furnished by IBGE, and calculated correction factors separately for each gender/age group/region strata. We corrected lack of completeness for the years 2000 to 2010 based on a study of active investigation of undeclared deaths [23] and performed the correction separately for each federative unit. Whenever the number of estimated deaths, after correction, was higher than that reported, we used the ratio of estimated to reported deaths as a correction factor. We applied linear interpolation of the data of 1991 and 2000 to estimate lack of completeness of death registration for the years 1992–1999. We corrected ill-defined causes of death by redistributing these deaths under the assumption that their distribution followed that of reported non-external cause deaths [24, 25].

We calculated the mortality rate for acute complications of diabetes annually. Given relatively small absolute numbers of deaths due to acute complications among those under 40 years of age in the north, center-west, and south regions of Brazil, the means of the annual mortality rate within the periods 1991–1995 and 2006–2010 were used to show trends for these groups. We tested statistical significance in the comparison between the two rates using either the confidence intervals of the rate ratios [26] or the p-value of the z-score for the difference between rates [27], the significance level being set at 0.05.

To facilitate comparisons between countries and across regions within Brazil, we directly standardized mortality rates by age and sex to the world standard population [24, 28]. In these analyses, age was stratified in intervals of 5 years up to 79 (0–4 years, 5–9, 10–14, 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79), the final strata being 80 years and older. We performed calculations with Stata 10.0. We analyzed trends in mortality from acute complications during the period with the Joinpoint Regression Program, starting with a maximum of 3 joinpoints and using the permutation method to determine the best model fit [29]. Though these numbers represent the universe, and as such do not require confidence intervals, we have included these intervals at times.

The project (number 100056) was approved by the Ethics Committee of Hospital de Clínicas de Porto Alegre. As analyses were based on surveillance databases from the Ministry of Health, no patient consent was necessary.