Analysis of trend in mortality due to HIV/AIDS-defining and non-HIV/AIDS defining illnesses according to sociodemographic characteristics, by Federative Unit and Brazil, 2000-2018

Abstract Objective: To analyze the temporal trend of mortality rate due to HIV/AIDS defining and non-HIV/AIDS defining illnesses in Brazil between 2000 and 2018. Methods: This was an ecological time series study, using data from the Mortality Information System, in Brazil and the Federative Units. Trend analysis was performed by means of Prais-Winsten regression model, according to overall mortality rate, sex, age group, marital status and race/skin color. Results: A total of 237,435 deaths were recorded in the period. In the country, defining illnesses showed higher rates (7.4 to 4.4 deaths/100,000 inhabitants in the period) than those observed among non-defining diseases (0.4 to 0.8 death/100,000 inhabitants in the period). It could be seen a decrease in overall mortality due to defining diseases (-6.3%; 95%CI -8.8;-3.8); while it increased due to non-defining diseases (11.0%; 95%CI 6.5;15.7). Conclusion: There was a change in HIV/AIDS mortality profile over the years, with a decrease in deaths due to HIV/AIDS-defining diseases.


INTRODUCTION
In 1996, universal, free of charge antiretroviral therapy (ART) was made available by the Brazilian National Health System (SUS). 1 ART increased survival and improved quality of life among people living with human immunodeficiency virus (HIV) and those with manifestation of acquired immunodeficiency syndrome (AIDS). 2 This treatment led to the occurrence of deaths from age-related diseases or long-term ART use, classified as non-HIV/AIDS defining illnesses, such as arterial hypertension, diabetes mellitus, heart diseases, neoplasms, kidney diseases among others. Thus, there was a decline in deaths from diseases associated with immunodeficiency that, until then, enabled the emergence of opportunistic illnesses, which are HIV/AIDS-defining illnesses, characterized by the occurrence of some specific diseases among people living with HIV/AIDS (PLHIV) who are immunosuppressed due to the evolution of HIV infection and the manifestation of AIDS, such as pneumonia and tuberculosis. 2,3 Changing patterns of deaths from HIV/AIDS has given the health condition the status of chronic disease, although its defining illnesses still remain leading causes of death. 2 S e v e r a l f a c t o r s m a y c o n t r i b u t e t o mortality due to HIV/AIDS in Brazil, such as sociodemographic characteristics, unfavorable living conditions and clinical issues, making it complex to define what primarily determines death from the disease. 4,5 Worldwide, there was a 39% decrease in the number of deaths due to HIV/AIDS recorded between 2010 (1,100,000 deaths) and 2019 (780,000 deaths), being this decrease attributed to adherence to drug treatment for PLHIV. 6 In Brazil, it could be seen a reduction in mortality due to HIV/AIDS after the distribution of ART, from approximately 10 deaths/100,000 inhabitants in 1995 to about 7 deaths/100,000 inhabitants, in 2000, four years after the implementation of the policy for universal and free of charge distribution of ART. 7 Currently, in certain Federative Units, mortality rates are higher than those found in the country, 7 a fact that may be related to a weakening of the Brazilian initiative in the response to HIV and disease control, in addition to a context of regional inequalities. 6,8 Knowledge of the mortality profile based o n s o c i o d e m o g ra p h i c ch a ra c te ri s t i c s and according to the type (def ining and non-def ining illnesses) is necessary. In the literature, most studies focus on the occurrence of diseases among PLHIV 3 rather than a comparison between def ining and non-def ining illnesses, which is precisely this research proposal, and the purpose of which is to contribute to the understanding of deaths due to HIV/AIDS based on the type and sociodemographic factors. Moreover, this research can support the direction for public policies, incorporating the population profile according to defining and non-defining illnesses of infection and condition.

Main results
This study showed that mortality due to non-HIV/ AIDS defining illnesses has increased. Despite this scenario, it can be seen that the rates for HIV/AIDSdefining illnesses are still expressive.

Implications for services
The knowledge of the profile of mortality due to HIV/AIDS according to defining and non-defining illnesses contributes to the management of service conduct.

Perspectives
It is expected that the findings of this study will contribute to the strategies for services that provide care for people living with HIV/ AIDS, in order to reduce deaths from the disease.

METHODS
This was an ecological time series study of mortality rates due to HIV/AIDS defining and non-HIV/AIDS def ining illnesses. This study used microdata from the Mortality Information System (SIM, as per the Brazilian acronym), via the Brazilian National Health System Information Technology Department (DATASUS, as per the Brazilian acronym) website in January 2022, 9 in order to make it possible to classify deaths from defining and non-defining illnesses. It included all deaths registered in the International Statistical Classification of Diseases and Related Health Problems -10 th Revision (ICD-10), between the items B20 and B24, related to the group "Disease caused by the human immunodeficiency virus" in the underlying cause or in lines A, B, C, D and II of the Death Certificate (DC).
The classification of a death as HIV/AIDSdefining illness included all deaths presented in the classification of HIV/AIDS-defining illnesses of the Coding Causes of Death in HIV Protocol, published by the Centers for Disease Control and Prevention (CDC) of the United States (Box 1). 10 With regard to non-defining illnesses, all those that were not in the list of defining illnesses were taken into consideration (Box 1).
Deaths were analyzed and classif ied as "death from defining illness" or "death from non-defining illness", based on ICD-10 items registered in lines A, B, C, D and II of the DC. Deaths whose underlying cause presented ICD were identified, however, lines A, B, C, D and II did not have a register in the Classification. In these cases, the ICDs registered were those classified between items B20 and B24 of ICD-10 as HIV/AIDS -defining illnesses, because they are ICDs that characterize the occurrence of death from opportunistic infections.
Mortality rates due to HIV/AIDS defining and non-HIV/AIDS defining illnesses per 100,000 inhabitants were standardized using the direct method, and the Brazilian population was def ined as a standard population. 11

ORIGINAL ARTICLE
Mortality due to HIV/AIDS-defining and non-HIV/AIDS defining illnesses by Federative Unit and Brazil Standardization using the direct method aims at adjusting the effect of age on the general mortality rate (GMR), taking into consideration a standard population, with known age distribution, from which we can identify the weights known by the proportion of people in each age group, which will be applied to agespecific mortality rates. 1 Regarding the direct method, GMRs by age groups of a population are applied to the population size of the standard population, making it possible to identify the expected deaths in each age group and, from the division of the total number of expected deaths by the standard population, calculate the standardized GMR. 11 Mortality rates considered, as numerator, deaths from defining/non-defining illnesses, and as denominator, the population in the period. This calculation was performed according to the sociodemographic characteristics aforementioned.
The standardizations was made for general mortality, sex (female; male), marital status (married; unmarried) and race/skin color (White; Black). The analysis by race/skin color comprised the White and Black categories, the latter resulting from the aggregation of Black and Brown categories, while for the White race/ skin color no aggregation was necessary since it was a single category. The aggregation of Blacks and Browns as Blacks was performed to enable the analysis of the Afro-descendant population in a unified way.
The variable "marital status" was also categorized, between married and unmarried. This aggregation was performed by the categories that represented the same meaning, married and unmarried, namely: single; married; widower; legally separated; in consensual union. For the category "married", the categories "married" and "legally separated" were aggregated, while for "unmarried", the stratifications "single", "widower" and "legally separated" were aggregated.
Mortality rates were also calculated according to age group, in years: 0 to 14; 15 to 29; 30 to 59; 60 or over.
Population data, necessary to calculate mortality rates, were retrieved f rom the DATASUS and Instituto Brasileiro de Geografia e Estatística (IBGE) websites in November 2021. 8 Regarding the calculation of the general mortality rates and age group, the population estimates available at the DATASUS website were used. 8 With regard to the populations, according to race/skin color and marital status, it was necessary to extract data from IBGE. These data were accessed via the Application Programming Interface (API) of the Automatic Recovery System (SIDRA, as per the Brazilian acronym), using the SidraR package via the Rstudio statistical program. These populations are available for census years 2000 and 2010, so projections were made for the intercensal years up to 2018. The projections were calculated using the geometry projection method, which considers population growth to be constant.
Prais-Winsten regression model was used for trend analysis. 12 Independent variables (X) refer to the years when the deaths occurred, and dependent variables (Y) correspond to the mortality rates. This model is applied in order to correct serial autocorrelation in time series, and it is necessary to use the Durbin-Watson test, in which the value of the test is measured from a scale ranging from 0 to 4. Values close to zero indicate the existence of maximum positive autocorrelation. When the values are close to 4, the serial autocorrelation is negative. However, if the Durbin-Watson value is close to 2, there is no serial autocorrelation. 12 After the analysis of serial autocorrelation, the logarithmic transformation of (Y) values was performed in order to reduce the heterogeneity of the variance of residuals of the model. Positive APC indicates an upward trend, while the negative corresponds to a decreasing trend; the series is called stationary when there is no significant difference between its value and zero. 12 Finally, the 95% confidence intervals (95%CI) of the study measurements were calculated using the following formula: The minimum and maximum values of b were identified from the 95%CI parameters generated by the statistical analysis software and applied in the formula, with the minimum value of b corresponding to the minimum point of CI, and the maximum value of b corresponding to the maximum point of CI.
The significance level considered was 5% in the Prais-Winsten model, for trend analysis. The steps of data organization, rate calculation, trend analysis and graphs were developed using the RStudio software, version 4.0.2.
This study was performed using publicly available secondary data, and did not involve interactions with humans. However, the project was submitted to the Research Ethics Committee of the Escola Nacional de Saúde Pública Sergio Arouca/Fundação Oswaldo Cruz (CEP/ENSP/Fiocruz), and was approved, Opinion No. 16, issued on November 23, 2020.

RESULTS
A total of 237,435 deaths from HIV/AIDS were recorded between 2000 and 2018. General mortality rates due to HIV/AIDS-defining illnesses ranged from 7.4 death/100,000 inhabitants, in 2000, to 4.4 death/100,000 inhabitants, in 2018. For non-HIV/AIDS defining illnesses, general mortality rates ranged from 0.4 deaths/100,000 inhabitants, in 2000, to 0.8 deaths/100,000 inhabitants, 2018 ( Figure 1A).

Box 1 -Correspondence of the International Statistical Classification of Diseases and Related Health Problems -10 th Revision (ICD-10) by defining diseases established by the Centers for Disease Control and Prevention (CDC) of the United States
To be continued

Box 1 -Correspondence of the International Statistical Classification of Diseases and Related Health Problems -10 th Revision (ICD-10) by defining diseases established by the Centers for Disease Control and Prevention (CDC) of the United States
To be continued

Box 1 -Correspondence of the International Statistical Classification of Diseases and Related Health Problems -10 th Revision (ICD-10) by defining diseases established by the Centers for Disease Control and Prevention (CDC) of the United States
To be continued

ORIGINAL ARTICLE
Mortality due to HIV/AIDS-defining and non-HIV/AIDS defining illnesses by Federative Unit and Brazil However, an upward trend was found in the states of the North and Northeast regions, mainly for HIV/AIDS-defining illnesses ( Figure 1B).
With regard to mortality due to defining illnesses, according to age group, the rates were significant among those aged 60 years and over in the country (from 13.1 deaths in 2000 to 26.0 deaths/100,000 inhabitants in 2018), as well as among those aged 30 to 59 years (from 16.0 deaths in 2000 to 4.8 deaths/100,000 inhabitants in 2018) (Figure 2A). Rates related to defining illnesses in the 0 to 14 and 15 to 29 age groups showed close values, the lowest being close to 0.2 death/100,000 inhabitants (Figure 2A). Trend in mortality due to defining illnesses showed that in the 0 to 14 age group, only Ceará (APC = 10.5%; 95%CI 2.9;-18.7) and Alagoas (APC = 9.8%; 95%CI 2.1;18.1) showed an upward trend in mortality rates due to def ining illnesses ( Figure 2B). It could be seen that in the 15 to 29 age group, there was no increasing trend in any units of analysis, a fact that was also observed in the 30 to 59 age group ( Figure 2B Regarding non-HIV/AIDS defining illnesses, taking into consideration the country as a whole, there was a mostly decreasing trend in the 15 to 29 (APC = -20.1%; 95%CI -20.7;-19.5) and 30 to 59 (APC = -6.6%; 95%CI -9.1;-3.9) age groups. However, for those aged 60 years or over, there was an upward trend in most states and in Brazil (APC = 29.1%; 95%CI 12.6;-48.0) ( Figure 2D).
Mortality rates due to defining illnesses and according to marital status were higher among unmarried individuals, throughout the period (from 27.0 deaths in 2000 to 12.4 deaths/100,000 inhabitants in 2018) ( Figure 3A). In the state of Amazonas, rates due to defining illnesses were higher among unmarried individuals (from 9.1 deaths in 2000 to 22.9 deaths/100,000 inhabitants in 2018), with oscillations over the period. The states of Roraima, Pará and Amapá also showed significant rates among unmarried individuals ( Figure 3A). Mortality due to nondefining illnesses showed lower rates when compared to defining illnesses ( Figure 3A).       In Brazil, there was a decreasing trend for defining illnesses among married (APC = -7.7%; 95%CI -9.3;-6.0) and unmarried individuals (APC = -9.1%; 95%CI -10.8;-7.5), while there was an upward trend in both categories when the trends in non-defining diseases were evaluated: for married individuals (APC = 9.1%; 95%CI 4.5;13.8) and for unmarried individuals (APC = 7.5%; 95%CI 2.5;-12.7) ( Figure 3B).

Box 1 -Correspondence of the International Statistical Classification of Diseases and Related Health Problems -10 th Revision (ICD-10) by defining diseases established by the Centers for Disease Control and Prevention (CDC) of the United States
Mortality rates due to defining diseases in Brazil were higher when compared to nondefining diseases, also in the stratification by race/skin color ( Figure 4A and 4D). Black people showed the highest rates for defining diseases, with the exception of the states of Amazonas, Roraima, Amapá, Maranhão, Santa Catarina, Rio Grande do Sul and Mato Grosso ( Figure 4A). Mortality rates due to non-defining diseases were very close among White and Black races/ skin color ( Figure 4C).  Figure 4B). Mortality due to non-defining diseases showed an upward trend, either among individuals of White race/ skin color, or among those of Black race/skin color ( Figure 4C).

DISCUSSION
This study showed that the general mortality rates due to HIV/AIDS-defining illnesses are higher when compared to non-HIV/AIDS defining illnesses, regarding the analysis of the years between 2000 and 2018. According to sex, the rates for defining diseases were higher among males, compared to the female population.
In the observed period, mortality due to HIV/AIDS-def ining illnesses showed a decreasing trend in the states of the South and Southeast regions, and upward and stationary trends in the North and Northeast regions, indicating the need for different measures aimed at controlling mortality due to the disease in these regions. 5,[13][14][15] Inequality in the regional distribution of services is likely to be contributing to this scenario. 4 Nevertheless, non-HIV/AIDS def ining illnesses showed an increasing trend in most states, when the general population and by sex were taken into consideration. The same pattern was found in other studies that pointed to an increase in deaths from non-defining illnesses. 2,3,16 High mortality due to def ining illnesses among males follows the pattern of mortality due to HIV/AIDS observed in the country and worldwide. 3,[17][18][19] However, females also need attention, because they have maintained an increasing trend in some states, when the defining illnesses are taken into consideration.
The oldest age groups, 30 to 59 and 60 years or over, showed higher rates of defining illnesses, while for non-defining illnesses, the age of 60 years or over showed more significant values. This scenario indicates that currently, mortality due to the disease can include individuals reaching advanced age who have acquired illnesses that are typical of HIV/AIDS, while the remaining people may be those who,

ORIGINAL ARTICLE
Mortality due to HIV/AIDS-defining and non-HIV/AIDS defining illnesses by Federative Unit and Brazil after long-term ART use, suffered from the side effects of treatment and even from common conditions of older age. 3,[16][17][18][19][20][21] Regarding marital status, unmarried individuals showed higher rates of HIV/AIDSdef ining illnesses. In addition, although mortality rates were higher among unmarried individuals, there was an upward trend, according to defining diseases, in the states of the Northeast region, and it could be seen an increasing trend, in most states, also among non-HIV/AIDS defining illnesses. High mortality rates among unmarried individuals were found in a study conducted in Papua, Indonesia, focusing on overall mortality, and the type of mortality was not detailed, whether due to defining diseases or not. 21 Studies show that, depending on local culture, married individuals may be more vulnerable to HIV/AIDS, such as that observed in a rural area of South Africa, from 2000 to 2017, while in other locations the disease may be concentrated among unmarried individuals. 23,24 The analysis based on race/skin color showed that mortality rates were higher among Black individuals in most states, according to defining diseases. There was also an upward trend in these rates, both for defining and non-defining illnesses. The highest proportion of the upward trend in non-defining illnesses was observed among Black individuals, while upward and stationary trends were found among White individuals, a result that may be linked to an ongoing increase in mortality due to nondefining illnesses among the population. Black population is in a greater social vulnerability [25][26][27][28] and the change in their mortality patterns may occur at a slower rate than that observed in the population of White race/skin color.
A number of factors expose the population of Black race/skin color to mortality due to HIV/AIDS, such as unemployment, difficulty in accessing health care services and fear of seeking treatment due to discrimination. 26 A study that analyzed the trend in mortality