Trend in the prevalence of depressive symptoms in Brazil: results from the Brazilian National Health Survey 2013 and 2019 Tendência na prevalência de sintomas

This study aims to evaluate national variation in depression prevalence and in different sociodemographic groups, health behaviors, and macroregions of Brazil from 2013 to 2019. Data were obtained from two nationwide Brazilian surveys – Brazilian National Health Survey 2013 and 2019. Participants aged 18 years or older were included, totaling 60,202 individuals in 2013 and 88,531 in 2019. Depression was evaluated with the Patient Health Questionnaire-9 (PHQ-9). All estimations accounted for the population weights and the complex sampling. The findings showed that during the six years between the two surveys, the prevalence of depression in Brazil increased by 36.7%, going from 7.9% in 2013 to 10.8% in 2019, and this increase is higher among unemployed young adults, aged 18 to 24 years, with the increase in the prevalence of depression almost tripled (3.7 in 2013 and 10.3 in 2019), an increase of 178.4%. Those dwelling in urban areas had a higher increase in the prevalence of depression in the six-year period (39.8%) when compared to residents in rural areas (20.2%). There was an increase in the prevalence of depression from 2013 to 2019 for the worst categories of the three health behaviors included in the study for both men and women: heavy drinking, smoking, and not exercising the recommended level of physical activity. Our results show a significant increase in the prevalence of depression over the six years between the two surveys, mainly among the younger and unemployed men. The country’s economic recession during this period may explain these findings. Health Surveys; Depression; Mental Health; Cross-Sectional Studies Correspondence C. S. Lopes Instituto de Medicina Social, Universidade do Estado do Rio de Janeiro. Rua São Francisco Xavier 524, 7o andar, Bloco D, Rio de Janeiro, RJ 20559-900, Brasil. cslopesims@gmail.com 1 Instituto de Medicina Social, Universidade do Estado do Rio de Janeiro, Rio de Janeiro, Brasil. 2 Faculdade de Medicina, Universidade de São Paulo, São Paulo, Brasil. This article is published in Open Access under the Creative Commons Attribution license, which allows use, distribution, and reproduction in any medium, without restrictions, as long as the original work is correctly cited.


Introduction
Cad. Saúde Pública 2022; 38 Sup 1:e00123421 metropolitan areas. The surveys were household-based with stratified sampling and a three-stage design: In the first stage, the primary sampling units were randomly selected from the master sample, from which the major surveys conducted by IBGE are sampled. In the second stage, households were randomly selected within each primary sampling units. In the third stage, an adult resident (aged 18 years or older in the 2013 edition and 15 years or older in the 2019 edition) was selected with equal probability among all adult residents in the household. Weighting factors were calculated for each of the three sampling units, considering the probabilities of selection and the non-response rate. For the selected resident, the weight was calculated considering the weight for the corresponding household, the probability of selection of the resident, the adjustment of non-response for sex, and calibration for the total population by sex and age groups estimated with the weight of all residents 18,19 .
Although the PNS 2019 included the population aged 15 years or over, to provide valid data for monitoring the indicators established by the Sustainable Development Goals (SDGs) 20 , the IBGE and the Brazilian Ministry of Health publications comprise only the population aged 18 years or older 17 . Thus, to allow comparisons between the two editions of the PNS, this study follows the same approach and uses data referring only to the population aged 18 or over.
In the PNS 2013, 69,954 occupied households were visited, and 60,202 individuals were interviewed, resulting in a response rate of 86.1%. In the PNS 2019, among the 108,525 households visited, interviews were conducted in 94,114 households, with a loss rate of 13

Assessment of depressive symptoms
In both PNS editions, depressive symptoms were assessed with the PHQ-9, which evaluates the frequency of depressive symptoms over the two weeks before data collection 21 . The instrument was validated in Brazil 22 , with good validity in diagnosing major depression at the cutoffs of > 9 and > 10. The presence of depressive symptoms was determined using the PHQ-9 score as recommended by Kroenke et al. 21 , which classifies depression severity according to the following thresholds: none (1-4 points), mild (5)(6)(7)(8)(9), moderate (10)(11)(12)(13)(14), moderately severe (15)(16)(17)(18)(19), and severe (20-27 points). In this study, the presence of depressive symptoms was defined by a PHQ-9 score of 10 or higher, which is considered the best cutoff point to detect the presence of clinically relevant symptoms 23,24 .

Other measures
In both surveys, sociodemographic variables were assessed, including sex (male and female); age group (18-24, 25-29, 30-39, 40-49, 50-59, 60-69, 70 years or over), race/skin color (white, black and brown, and others, which includes Asians and indigenous); level of education (uneducated or incomplete primary school; complete primary school or incomplete high school, complete high school or incomplete undergraduation, complete undergraduation); per capita household income, classified into minimum wages (0 to 1, more than 1 to 3, more than 3 to 5 and more than 5); marital status (married or living with a partner and single, divorced or widowed not living with a partner) and work status (employed, unemployed). Geographical areas were defined as macroregion of residence (North, Northeast, Central-West, Southeast, and South) and living in urban vs. rural areas.
Health behaviors selected for analysis were: current smoker of any tobacco (yes, no); excessive alcohol consumption (heavy drinking) being defined as the weekly consumption of 15 or more alcoholic drinks for men and 8 or more for women 25 , a dose being considered equivalent to a can of beer, a glass of wine, or a dose of distilled drink; and leisure-time physical activity considering the recommended level of 150 minutes per week of mild or moderate leisure-time physical activity or 75 minutes of vigorous leisure-time physical activity 26 .

Data analysis
The prevalence of depressive symptoms was described according to sociodemographic characteristics and geographical area. Estimates were computed for PNS 2013 and PNS 2019, with their respective 95% confidence intervals (95%CI) based on the t distribution, considering a large number of primary sample units. The prevalence change between the 2013 and 2019 surveys was expressed in absolute difference and percent prevalence ratio to quantify its magnitude and relative variation. Both were computed with generalized linear models, the former using a Gaussian and the latter a Poisson model. The datasets from the surveys were stacked to compose a single dataset, and an indicator variable was created to flag their respective survey iteration, 2013 or 2019. Post-stratification calibration was performed to adjust the weights accordingly. The prevalence in each grouping category in 2013 was held as the baseline by making the intercept equal to zero. To compute the prevalence change from 2013 to 2019, an interaction term between the grouping variable and the indicator variable estimated either the absolute difference or the log of the prevalence ratio. The reported percent prevalence ratio was computed as the exponential of the interaction coefficient minus one and multiplied by 100.
All statistical analyses were performed using survey-specific weighting factors adjusting the study samples to the demographic-geographic distribution of the population in Brazil. The analyses were conducted in R version 4.0.5 (http://www.r-project.org) and the survey package version 4.0.

Results
Demographic features of PNS 2013 and PNS 2019 samples are shown in Table 1. Overall, the sample characteristics of PNS 2013 and PNS 2019 showed little change in the underlying population regarding the included variables, except for age distribution, which shifted towards the older ages, and the increase in higher levels of education. Table 2 shows the prevalence of depressive symptoms according to sociodemographic characteristics and health behaviors in the general population in 2013 and 2019. Overall, there was a 37.6% increase in the prevalence of depressive symptoms between 2013 (7.9%) and 2019 (10.8%). This increase was higher among the youngest (aged 18-24 years old), women, those living in urban areas, and those living in the Southeast Region. There was a significant sex difference in the prevalence of depressive symptoms; women had a higher prevalence in both surveys, with an increase in prevalence between 2013 (10.7%) and 2019 (15%) greater than that observed among men (4.7% vs. 6.1%). Prevalence among women was significantly higher than among men in all age groups and at both time points (Table 3).
Although increase in the prevalence of depressive symptoms for all age groups was observed, this increase was more noticeable among the younger age groups, especially among those aged 18 to 24 years, where the prevalence of depressive symptoms almost doubled, being 5.6% in 2013 and 11.1% in 2019. This pattern was observed for women (8.3% vs. 15.6%) and men (2.9% vs. 6.6%) for the same age group in the stratified analysis. In contrast, for those aged 70 years or more, the prevalence remained almost constant in the whole population, changing from 10.2% to 11.1%, increasing among women but not among men.
Those living in urban areas showed a higher increase (39.8%) in the prevalence of depressive symptoms (from 8.1% in 2013 to 11.3% in 2019) when compared to those living in rural areas (from 6.4% to 7.7% in the same period), a relative increase of 20.2%. When considering the country's macroregions, the absolute differences between the prevalence of depressive symptoms in 2013 and 2019 followed the 2-3% found for the whole country, except the Southeast (Table 2). When stratified by sex, the absolute differences of 4-5% in the prevalence of depressive symptoms between 2013 and 2019 among women remained for all regions, except for the South Region, where it was less than 1% (13.4% in 2013 and 14.3% in 2019). In 2013, the prevalence of depressive symptoms among women in the South Region was higher than in the rest of the country and remained stable until 2019 (       For socioeconomic and demographic characteristics, no relevant absolute differences were found in the prevalence of depression between 2013 and 2019 in the categories under study, with increases mostly around 3% between 2013 and 2019 for the overall population, 4-5% for women, and 1.5% for men. On the other hand, considering the relative variation, one can observe increases of up to 127.1% for men aged 18 to 24 and 89.1% for women in the same age group (Tables 2 and 3).
When stratified by sex, age, and working status, the absolute difference increase in the prevalence of depressive symptoms almost tripled in the group of men aged 18 to 24, who were unemployed (3.7% in 2013 and 10.3% in 2019), a relative increase of 178.4%, when compared to those who were employed (2.6% vs. 4.9%), a relative increase of 90.5% (Table 4). Cad   Regarding health behaviors in the general population, Table 2 shows that the prevalence of depressive symptoms increased in all of them around 3% in absolute value. In 2013 and 2019, people who reported heavy drinking, who did not exercise the recommended level of physical activity (inactive), and who smoked had a higher prevalence of depressive symptoms than those without such health risk behaviors. The variation in the prevalence of depressive symptoms among smokers varied from 10.8% in 2013 to 14.7% in 2019, a 35.9% prevalence ratio. A percent prevalence ratio increase of 40.4% was found among those that did not exercise the recommended level of physical activity (inactive), higher than differences observed in the other categories of health risk behaviors.
When stratifying by sex (Table 3), among women who reported heavy drinking, there was a higher increase in the prevalence of depressive symptoms from 2013 to 2019 than among those who did not report this pattern of alcohol consumption. Women who reported smoking in 2013 showed a higher prevalence of depressive symptoms (17.7% among smokers vs. 9.9% among nonsmokers) and in 2019 (23.1% among smokers and 14.1% among nonsmokers); with an increase of 30.4% in the percent prevalence ratio among smokers from 2013 to 2019. Among men, smoking showed a higher prevalence of depressive symptoms in 2013 and 2019, with a relative increase of 41.9%, considerably higher than the variation among men in the general population. Regarding physical activity, women who did not exercise the recommended time/intensity of physical activity in 2019 showed a higher prevalence of depressive symptoms than those who did exercise, as observed in 2013, with a relative increase of 45.8%. Conversely, among men, those who did not practice the recommended time/intensity of physical activity had a relative increase of 25.4% in the prevalence of depressive symptoms as compared to those who did practice physical activity, moving from 5.6% in 2013 to 7% in 2019 (Table 3).

Discussion
This is the first study that compares the prevalence of depressive symptoms in the Brazilian population in two distinct time-periods. The study shows an increase in the prevalence of depressive symptoms from 2013 to 2019, from 7.9% to 10.8%, mainly among women and the youngest. Across all age groups, the prevalence of depressive symptoms showed different change, with a higher increase among those aged 18 to 24 years. This increase among the youngest was even higher among unemployed individuals and especially among younger men.
Overall, the current results differ from those in the initially presented meta-analysis and from the Global Burden of Disease Study findings, which found no differences in the prevalence of depression over time 1,3 . A study in Chile, which compared the prevalence of major depression in 2003 and 2010, also found no significant variation (20.5% vs. 18.4%, respectively) 27 . However, studies that assessed the impact of financial crises and economic recessions on the prevalence of depression in adult popu-Cad. Saúde Pública 2022; 38 Sup 1:e00123421 lations from different countries found results similar to ours. In Spain, a study conducted to assess the impact of the economic crisis that began in 2007 on different health outcomes showed that, compared with the pre-crisis period of 2006, the 2010 survey revealed that the highest increase in frequency was for mood disorders, major depression (an absolute increase of 19.4%) and dysthymia (10.8%) 28 . In Greece, the prevalence of major depression increased from 3.3% in 2008 to 8.2% in 2011, and this increase was attributed to the economic crisis experienced by the country in 2008 29 . To assess shortterm differences in population mental health before and after the 2008 recession in England, a study conducted with representative samples of the general population in the working-age (25-64 years) was made between 1991 and 2010. The results showed an increase of common mental disorders from 13.7% in 2008 to 16.4% in 2009 and 15.5% in 2010 30 .
Some of these studies also found that financial crises, periods of recession, and unemployment have a higher impact in specific subgroups, especially among the youngest and among men 31 . Probably, these groups are more affected by economic crisis. The loss of their jobs, or the impossibility to get one, may conduct them to disillusion and hopelessness situations. On the other hand, women, who already have a higher risk of depression, are also very affected by periods of economic crisis. However, none of these studies found differences in the prevalence of depression as large as those observed in this study. Our findings showing a 178.4% increase in the prevalence of depressive symptoms among unemployed men aged 18 to 24 years, and 89.8% among women in the same age group and work situation, is unparalleled in the literature. Brazil had a period of good economic growth from the beginning of the 21st century to 2014, which was followed by a period of deep economic crisis, with a significant increase in unemployment, which led to a dramatic drop in the population's standard of living, affecting mainly those who were at the age of entry into the labor market. As Brazil did not have effective mechanisms for social protection in such periods of crisis, it is possible that more vulnerable groups suffered the consequences of economic hardships more intensively.
This study also found that individuals dwelling in urban areas of the country had a higher prevalence of depressive symptoms in 2013 and 2019 and a higher increase in the percent prevalence ratio of depressive symptoms in the 6 years (39.8%) when compared to residents in rural areas (20.2%), and this pattern was similar for men and women. We did not find any other study that investigated living in urban vs. rural areas and the mental health trend in Brazil, but previous studies on the prevalence of common mental disorders/depression in urban and rural areas have shown inconclusive results; some studies showed association, whereas others did not 32,33,34 . However, studies conducted in other countries corroborate our findings and show that living in urban regions with high demographic density is associated with a higher risk of depression 35,36 . Among the studies that have investigated the trend of depression over time, some have observed a tendency towards an increase in the prevalence of depression in urban vs. rural regions, following accelerated urbanization processes 37,38 . Other studies, however, did not observe such a trend 27,39 . A recent study to assess trends in the prevalence of depression between 2014 and 2018, conducted in Peru, found no significant differences in the prevalence of depression in that period for urban and rural regions 40 .
Regarding health behaviors, this work found an increase in the prevalence of depressive symptoms between 2013 and 2019 for the worst categories of the three health behaviors under study, for both sexes: heavy drinking, smoking, and no physical activity during leisure-time, following the pattern observed for the general population. However, when stratified by sex, the differences in the prevalence of depressive symptoms are higher for women who reported excessive alcohol consumption than for men who reported such behavior. The relationship between health risk behaviors and depressive symptoms is well established in the literature 41, 42,43,44 . However, few studies have evaluated the role of health risk behaviors on changes in the prevalence of depression over time. Although estimating associations is not the goal of this study, the stratified analysis suggests that such relationship may be worth investigating. Overall, the current results are in line with those presented by a study based on the annually cross-sectional U.S. National Health Interview Surveys (NHIS) of 1997-2016, among individuals aged 18 years and older 45 . They found that psychological distress became more strongly associated with smoking and physical inactivity but less strongly associated with heavy alcohol consumption. Another study, also in the American population, examined changes in the prevalence of major depression in the United States between 1991-1992 and 2001-2002 and sought to determine whether these changes were associated with changes in substance abuse (includ-Cad. Saúde Pública 2022; 38 Sup 1:e00123421 ing alcohol). They found that increases in the prevalence of depression associated with substance use disorders were consistent only for black men aged 18 to 29 years 46 . Our results are also in line with longitudinal studies showing that women in the higher risk drinking group at baseline were at a higher risk of developing depression disorder at follow-up 44,47 .

Strengths and limitations of this study
One of the strengths is that this study is the first one to assess the trend in the prevalence of depressive symptoms based on two national representative surveys. Thus, it allows assessing changes in the prevalence of depressive symptoms according to sociodemographic characteristics, region of residence, and health behaviors. Moreover, both surveys used the same standardized questionnaire (PHQ-9), widely used in national and international studies to assess depressive symptoms, according to internationally accepted criteria, that allows the comparison of this study results with those of other international studies.
Some limitations need to be addressed: Firstly, the presence of depressive symptoms was assessed with the PHQ-9 with the cutoff 10 to classify depression, and pooled estimates for such cutoff are 0.77 for sensitivity and 0.85 for specificity, implying that some degree of random misclassification may have occurred. It may have biased the prevalence estimates and, therefore, the differences and ratios. Secondly, more severe cases of depression may have been unaccounted due to non-response, measurement error, and exclusion of institutionalized individuals in both surveys. Another limitation is that the primary sample units use different identification codes across surveys; since primary sample units may be overlapped in both samples, there may be some variability due to dependency not accounted for. Thus, the standard errors for the differences and ratios may be slightly underestimated, and the confidence limits close to the null hypothesis should be interpreted with caution.

Conclusion
Our findings show a significant increase in the prevalence of depressive symptoms over the six years between the two surveys. The finding that the group of younger and unemployed men showed the highest variation in the prevalence of depressive symptoms draws attention. It encourages us to seek explanations based on the literature and the country's socioeconomic context during this period. It is possible that such subgroup is today one of the most vulnerable and this condition may affect their mental health. Although economic crises tend to reduce healthcare budgets, mental health care budget must be maintained or even increased, so that economic recovery and mental health of the population can achieve faster and better results. Cad