Sociodemographic determinants of multimorbidity in Brazilian adults and older adults: a cross-sectional study

ABSTRACT BACKGROUND: Multimorbidity due to non-communicable chronic diseases (NCDs) constitutes a significant challenge for healthcare systems. To attenuate its impacts, it is essential to identify the sociodemographic determinants of this condition, which can discriminate against population segments that are more exposed. OBJECTIVE: To identify associations between multimorbidity conditions and sociodemographic indicators among Brazilian adults and older adults. DESIGN AND SETTING: Cross-sectional telephone-based survey in 26 Brazilian state capitals and the federal district. METHODS: The Vigitel 2013 survey was used, with data collected via a questionnaire. The outcome was multimorbidity (2, 3 or 4 NCDs), and the exposures were sociodemographic indicators (age, sex, skin color, marital status and education). The analysis consisted of multinomial logistic regression (odds ratio), stratified by age. RESULTS: Among adults, multimorbidity comprising two, three or four diseases was associated with advancing age (P < 0.001); two and three diseases, with having a partner (P = 0.004 and P < 0.001, respectively); and two, three or four diseases, with lower education (P < 0.001). Among older adults, two, three or four diseases were associated with female sex (P < 0.001); three diseases, with living with a partner (P = 0.018); two diseases, with black skin color (P = 0.016); and two or three diseases, with lower education (P < 0.001). CONCLUSIONS: To control and prevent multimorbidity, strategies for individuals with existing chronic diseases, with partners and with lower education levels are needed. Particularly for adults, advancing age should be considered; and for older adults, being a woman and having black skin color.


OBJECTIVE
The aim of the present study was to identify the sociodemographic factors associated with multimorbidity due to non-communicable chronic diseases among adults and older adults in Brazil.

Design and sample
The data for this cross-sectional study came from the annual national survey "Surveillance of Risk Factors and Protection Against Chronic Diseases by Telephone Inquiry" (Vigitel in the Portuguese-language acronym), conducted between February and December 2013. The sampling process was performed in three steps. 12 Weighting factors were used to compensate for nonuniversal fixed-line coverage bias, adjusted for the adult Brazilian population. This was based on each individual's weight in the sample, calculated via the ranking method. Details about the methodology, including the sampling process, weighting factors and ethics procedures are provided in the official report. 12

Sample
The participants were a representative sample of adults (≥ 18 years old) living in all 27 Brazilian state capitals. To be eligible, the requirement was that the participant needed to have a residential landline telephone.

Data collection
Data were collected by means of telephone interviews simultaneously using a computer, and all the calls were recorded in case any queries arose. The instrument used was a validated questionnaire, and it was applied via telephone calls by trained staff.
The questionnaire asked about sociodemographic, behavioral, nutritional and health factors.

Measurements
The outcome variable was multimorbidity due to non-communicable diseases (NCDs). The concept investigated was co-occurrence of multiple chronic or acute diseases and medical conditions within a single person, without an index condition, giving equal attention to all diagnoses. 13 The four categories proposed were: no occurrence (zero or one disease), and multimorbidity as occurrences of two diseases, three diseases or four diseases.
The diseases considered were the four most prevalent NCDs in Brazil (diabetes, dyslipidemia, arterial hypertension and obesity). Occurrences were considered to consist of affirmative selfreports of diabetes, dyslipidemia or arterial hypertension; and for obesity, body mass index (BMI) of 30 kg/m² or higher (calculated based on self-reported weight and height). For BMI, hot-deck imputation of the data was used. The exposure variables used were sex (male or female), age groups (adult categories: 18 to 29,   30 to 39, 40 to 49 or 50 to 59 years old; older adult categories: 60 to 69, 70 to 79 or 80 years old and over), marital status (living without a partner or living with a partner), ethnicity (white or black) and education level (less than eight, nine to eleven or twelve years and over) and demographic macroregion (centerwest, northeast, north, southeast and south).

Analyses
The analysis, conducted in 2016, was stratified according to age group (adults: 18 to 59 years; and older adults: ≥ 60 years).
The descriptive analyses comprised absolute and relative frequencies, considering prevalence estimates and 95% confidence intervals (95% CI). Multinomial logistic regression (odds ratio [OR]) formed the inferential analysis that was used to investigate associations of sociodemographic indicators with each multimorbidity category (taking zero or one disease as the reference category). The first-level variables were sex, age, marital status, ethnicity and demographic macroregion; and the second level was education level. Backward selection was adopted for statistical modeling, with a critical level of P ≤ 0.20 for each variable to remain in the hierarchical regression model, so as to minimize the confounding control. The significance level was 5% for all tests. All analyses considered sample weighting obtained through the inverse of the number of telephone lines existing in the household that was interviewed and the number of adults living in the interviewee's home.

Ethics
The Brazilian Ministry of Health's National Ethics Committee for Research on Human Beings approved this study under registration number 355.590/2013, on June 26, 2013.

RESULTS
Within the response rate of 71.5%, the total number of participants was 52,929 (37,947 adults and 14,982 older adults). Among the adults, the following were predominant: females (52.9%), individuals living without a partner (52.0%), blacks (55.1%), individuals with nine to eleven years of schooling (41.1%) and those living in the southeastern region (44.1%). Among the older adults, the following were predominant: females (59.5%), individuals living with a partner (56.9%), those of white non-Hispanic ethnicity (61.3%), those eight or fewer years of schooling (69.3%) and those living in the southeastern region (51.8%) ( Table 1).
The frequency of multimorbidity in adults was 13.7% (9.8% with two, 3.3% with three and 0.6% with four NCDs), and 42.9% among old adults (27.9% with two, 12.4% with three and 2.7% with four NCDs) (Figure 1). The odds for this group were also associated with the demographic macroregion, indicating that the southeastern and southern regions were more exposed than the other demographic macroregions.   Table 3).

DISCUSSION
In this study, the aim was to identify sociodemographic factors associated with multimorbidity due to non-communicable chronic diseases among adults and older adults in Brazil.
Among adults, the main sociodemographic factors related to multimorbidity due to two, three and four chronic diseases were age between 40 to 59 years, the lowest educational level and the fact that they lived in the southern and southeastern regions. Among older adults, being female and having the lowest educational level were the characteristics that presented the most consistent associations with occurrences of multimorbidity, while residing in the southeastern and central-western regions was an inverse feature.
In the present study, older women were at greater risk of NCD multimorbidity. However, among adults, this same result was not seen. In a review of the literature by Marengoni et al., 14 on the multimorbidity process in relation to aging, it was indicated that women characterized this state of health more clearly. Another study identified higher risk for women regardless of age, with a magnitude lower than in the present study (OR = 1.12). 15 The higher risk presented by women in this age group can be attributed to the postmenopausal period 16  aging can be attributed to genetically engineered mechanisms, neuronal-endocrine failures and modifications resulting from the oxidative stress proteins and deoxyribonucleic acid (DNA) of cellular lipids. These alterations increase chronic inflammatory states, since they are not controlled mainly by behavioral factors, and they therefore exposed personal effects to the appearance of NCDs. 22 Considering marital status, the adults and the older adults with partners were at greater risk of having multimorbidity consisting of two and three NCDs, respectively. Other studies have also found this association of risk among adults, 4,23 which can be attributed to changes to unhealthy habits, with increasing burden of disease appearing over the course of the routine that characterizes marital transition. 24 Evidence for this association had already been presented, 25 but assessment of multimorbidity consisting of concomitant diseases constitutes a new approach. This may be explained differently among adults, in the light of the transition of the marital situation. Another interpretation that might be suggested is that individuals with a partner are more likely to use healthcare services than are those without a partner. 26 The adults' skin color did not show any association with multimorbidity, but among the older adults, there was higher risk of multimorbidity with two NCDs among individuals with black skin color. This result can be characterized in social or biological terms.
Socially, older adults with black skin color have social, economic and cultural barriers regarding their living and health conditions, which reflect the unequal distribution of risk factors, protective factors and health problems accumulated over the course of their lives. 27 From a biological point of view, people with black skin color We recognize that there were some limitations to this study that need to be considered before interpreting the results. The main limitation was the number of NCDs included as multimorbidities.
This hampers comparisons with data from other countries or from using different instruments. Moreover, the measurements were self-reported, which required that the interviewees were aware that the disease had been diagnosed. The fact that Brazil's public healthcare programs give the population broad access to primary healthcare also needs to be considered. 31 Our study did not assess the contributions that may have been made by substantial numbers of other diseases and their severity.
Furthermore, the obesity classification was based on self-reported weight and height information and should be considered cautiously. Lastly, even though the sample was representative nationally, it is necessary to restrict the extrapolation of the results. Only individuals residing in the capitals of the Brazilian states and those who had landline telephones were considered, although the weighting used in Vigitel aimed to minimize this source of bias.