Prevalence of diabetes mellitus according to associated factors in rural traditional populations in Goiás, Brazil: a cross-sectional study

: Objective: To identify the prevalence of and factors associated with diabetes mellitus in rural traditional communities. Methods: Cross-sectional study carried out in 115 rural communities distributed in 45 municipalities in the state of Goiás, including: 13 river communities, 51 quilombolas and 63 agrarian reform settlements. Probabilistic sampling was performed, and participants were selected at random. The outcome variable was self-reported diabetes mellitus, while exposure variables were sociodemographic, lifestyle, health conditions and access to health services. Multiple regression was used to determine the association between study variables. Results: Among the 2,537 participants, the overall prevalence of diabetes was 9.8%, with 13.5% in river dwellers, 10.0% in quilombolas and 9.3% in settlers. Factors associated with diabetes were negative self-perception of health, being a former smoker, high blood pressure, hypercholesterolemia, and living in a river community. Conclusion: The results reinforce the need to strengthen strategies for the prevention and control of diabetes and its complications in rural populations, especially among river dwellers.


INTRODUCTION
Diabetes mellitus (DM) is a chronic non-communicable disease (NCD) with a global impact, and its prevalence tends to increase especially in low-income countries 1 . The International Diabetes Federation (IDF) estimates that 463 million people worldwide have DM (9.3%), and that in the next 25 years, this number will reach 700 million, representing an approximate increase of 51% 1 . Brazil ranks fifth among the ten countries with the highest prevalence of DM in the world 1 , which is linked to risk factors prevalent among Brazilians, such as overweight, obesity and sedentary lifestyle 2 .
Most studies on DM in Brazil are concentrated in urban areas 3 , and little is known about its occurrence in rural areas. The National Health Survey (PNS), one of the largest Brazilian epidemiological surveys, estimated a prevalence of 7.9% in urban areas and 6.3% in rural areas 4 . In other countries, the prevalence varies from 4.7 to 19.6% [5][6][7] in different rural populations. Brazilian studies with river dwellers in the North region 8 and rural population in the South 9 show a prevalence of 10 and 16.4%, respectively.
Rural traditional populations are historically marked by situations of inequality, inequities and exploitation of rural activity, which are reflected in significant social and health vulnerability 10 . Conceptually, they are understood as peoples and communities that have a socioeconomic and sociocultural lifestyle closely related to the land, including family farmers, peasants, settlers and/or campers, as well as river and quilombola people 10 .
In this group, the geographic location, lack of basic sanitation, less education, less access to transport, and moreover economic and social determinants and the difficulty of accessibility to basic health units 11 are among the main barriers to recognition, monitoring and the control of health problems 12 . It is noteworthy here that access to health services can be widely understood and is directly related to the ability of a group to seek and obtain care, taking into account the availability of health resources, the ability to produce services, and resistance of the service (obstacles to seeking and getting attention) 13 .
In the case of NCDs, such as DM, the lifestyle of rural populations has led to an increase in the frequency of unhealthy behaviors, such as smoking, consumption of alcoholic beverages, physical inactivity and inadequate consumption of vegetables and fruits [14][15][16][17] . In addition, rural and traditional populations have a high prevalence of hypercholesterolemia and hypertension, which are risk factors related to DM 18,19 . These factors, associated with low education and lower socioeconomic status, contribute to the occurrence of NCDs, conditions that increase the demand for health services 12 . Specifically in the case of DM, studies show that people with this disease in rural areas are at greater risk for developing complications such as diabetic foot because of working conditions, mycoses and lower health education level for diabetes care 20,21 .
It is therefore essential that surveys on the health conditions of rural populations are conducted to give visibility to the health needs of this group and thereby propose health care strategies, considering their specificities and collaborating for the execution of the objectives of the National Policy of Comprehensive Health for People of the Countryside, Forests and Waters 10 .
The aim of this study was to analyze the prevalence of self-reported DM and its associated factors in a rural traditional population in central Brazil.

STUDY DESIGN AND PARTICIPANTS
We conducted an observational study with a cross-sectional design, nested within the matrix project titled "Sanitation and Environmental Health in Rural Traditional Communities in Goiás (Sanrural Project)", which covered 45 municipalities in the state of Goiás, central region of Brazil. A total of 115 rural communities were included, distributed as follows: nine river communities, 44 quilombo remnants certified by the Palmares Quilombolas Foundation and 62 agrarian reform settlements. The choice criterion was based on the selection of municipalities that had one or more accredited quilombola and river communities, followed by the presence of agrarian reform settlements under the management of the National Institute of Colonization and Agrarian Reform, Regional Superintendence 04 (INCRA SR-04).
The parameters considered for the sample size calculation of the larger study were: 1. Estimated total number of existing families in the 115 communities; 2. 90% confidence level;

3.
Maximum error of 10% in the interval estimates of proportions of the characteristics investigated in the families of the communities; 4. Bonferroni correction for multinomial responses, with 14 categories 22 .
The parameter estimates were approximated by the simple random sampling method. Considering these parameters, the number of 3,779 families to be interviewed in the 115 communities was estimated. In the present study, families whose members were residents of permanent homes in a lot/area of the community who, during the data collection period, were present or temporarily absent, were eligible. For the interview, a family member aged ≥18 years was considered responsible for the household in a consensual way with the other family members, about whom he provided information. Thus, to estimate the prevalence of DM and associated factors, the analyses were restricted to the individual characteristics of the resident responsible for the families. Therefore, the missing answers about the presence of diseases used to compose the outcome of this study were considered as exclusion criteria. Considering the data collected in the larger project, a new sample calculation for simple proportion was performed to determine if the sample of individuals from this cut had sufficient power to estimate the prevalence of self-reported DM. Thus, a minimum sample of 1,573 individuals was estimated as sufficient to detect the minimum prevalence of DM of 9.8% found in this study, with a 90% confidence level and a sampling power of 0.95. The power of the a posteriori test for the binomial distribution was also estimated to determine the adequacy of the sample for the estimates, considering the number of samples actually collected (n=2,537). For these estimates, the results showed that the sample collected, at the 95% confidence level, had a power of 99.6% for the analysis of the prevalence of DM.

DATA COLLECTION
Data collection took place between August 2018 and August 2019 during one of the stages of the project called Workshop 2, which consisted of on-site activity in 115 communities. Three teams were formed with three field researchers, who had higher education and previous training to apply the questionnaires and approach people.
For data collection in the communities, mobilization first took place, carried out between the researchers and a community mobilizer (person of reference or identified as a leader). Upon arriving at the community, the researchers made a reconnaissance of the area, the local mobilization and the definition of the itinerary of visits. After scheduling the visit, the resident was visited by the project researchers at their home, and they interviewed them using the HP iPAQ Pocket PC device to fill in the information obtained. All participants were informed about the research objectives.
Standardized questionnaires were used that addressed demographic and health conditions, demand for and use of health services, prepared for this study. The questions had already been used in epidemiological surveys, but some questions were included for the REV BRAS EPIDEMIOL 2022; 25: E220016 purposes of the present study, after standardization in a pilot study. The health data questionnaire consisted of 36 questions.

VARIABLES
The outcome variable was self-reported DM estimated by the question: "Has a doctor ever told you that you have diabetes?" 23 . Prevalence was calculated using the number of adults interviewed who responded positively to this question as the numerator and the total number of people interviewed as the denominator.
Exposure variables were grouped as follows: 1. Sociodemographic: sex, age group, skin color and education; 2. Health conditions: self-assessment of health status, hypertension, hypercholesterolemia, use of alcohol and tobacco, physical activity and hospitalization in the last year. Regarding this item, self-rated health was measured through the question: "What do you think your state of health is?". The answers were dichotomized into very good/good/fair and bad/ very bad 24 . Hospitalization in the last year was identified through the question: "Have you been hospitalized in the last year?". Hypertension and hypercholesterolemia were verified by the question: "What diseases has the doctor said you have?" Alcohol consumption, tobacco use and physical activity were determined using the following questions, respectively: "How often do you drink alcohol?" 25 ; "How often do you smoke?" 25 and "How often do you engage in physical activity?". For the three questions, the response options were "daily", "weekly", "monthly", "on occasion" and "not at all". Thus, the daily and weekly frequencies were considered as regular practice of physical activity (yes) 26 ; regular use of alcoholic beverages (yes) was with reference to daily and weekly consumption; and daily, weekly, monthly and occasional smoking was considered as tobacco use (yes). Having health insurance was evaluated by: "Do you have health or medical insurance?"; 3. Indicators related to the availability of health services: "Have you received a visit from a member of Family Health Strategy, as a community health agent, in the last year?"; and 4. Search for services, with the questions: "Have you sought medical attention in the last year? and "Have you sought a specialist in the last year?" 25 .

DATA ANALYSIS
Data analysis was performed using Stata software, version 12.0 (StataCorp, College Station, TX, United States). Initially, a descriptive analysis of the sample was performed using absolute and relative frequencies of qualitative variables and mean and standard deviation (SD) of quantitative variables. The prevalence of self-reported DM was determined with a 95% confidence interval (95%CI) for binomial distributions, for the total sample and the subpopulation under study (quilombolas, river dwellers and settlers).

REV BRAS EPIDEMIOL 2022; 25: E220016
To verify the association between the dependent variable (self-reported DM) and the independent variables, bivariate and multiple analyses were performed using the Poisson regression model. In the bivariate analysis, each independent variable was associated with the dependent variable, obtaining the crude prevalence ratio (PR) and respective 95%CI. Variables with a value of p<0.20 were then included in a multiple Poisson regression model. The input method for the independent variables was stepwise, with forward as the selection method. In this modeling, the beginning occurs without variables in the regression equation, incorporating statistically significant variables into the model, one by one. The criterion for maintaining the variable in the final model was a value of p≤0.05 in the regression. The magnitude of the association was estimated using the adjusted prevalence ratio (PRa) and 95%CI. Statistical significance was established by the Wald test, considering a significance level of 5% (p<0.05). Before the study participants filled out the questionnaire, an informed consent form was read and explained to obtain the interviewee's written or digital signature.

RESULTS
A total of 2,537 people were included, 55.5% men and 44.5% women. The mean age was 51.6 years (SD: 14.7); 41.0% reported having brown skin color, and 61.5% had completed elementary school or incomplete high school ( Table 1).
In the bivariate analysis, the prevalence of DM was higher in women (10.3%), in people aged 60 or older (10.5%), with black skin color (10.3%) and with incomplete elementary education (12.8%). Schooling showed a statistically significant difference (p=0.006) (

DISCUSSION
This study showed that the prevalence of DM in the communities studied was 9.8%, higher than that identified by the global estimates of the International Diabetes Federation (IDF), which show a prevalence of 7.2% in the rural population 1     The prevalence found in this study may be due to the general underdiagnosis of DM in Brazil 28 . It is estimated that 7.7 million people who have DM are not diagnosed 28,29 -a reality that has a higher proportion in rural populations, due to organizational and structural factors of primary care 23,29,30 . Despite the advances of the Family Health Strategy (ESF) in the country, the inequalities in the rural population's access to health services and the low tracking of this population by primary care professionals can be understood by the lower adherence to the ESF by the municipalities, especially in the North, Central-West and Southeast regions, which implies the lack of a professional community health agent (CHA) and others, who play a fundamental role in the active search and monitoring of communities and families 28,29,31 . Also, when there is the presence of the multidisciplinary team by the ESF, there is an overload of the service for the different communities because of the difficult geographical access, due to the territorial peculiarities where these communities are located, especially the quilombolas 12,32 . This results in less monitoring of health by primary care, which makes it complex for users to enter the Health Care Network (RAS) 29,32 . All these factors can contribute to late diagnosis in the presence of an emergency complication of DM 27,28 .

. Research carried out in
The present study also shows that the prevalence of DM was higher in riverine people, compared to quilombolas and rural settlers. To our knowledge, no studies were found with these groups that could be compared with our findings. Only in river communities, was there a study in the state of Amazonas that observed a prevalence of 16.4 and 10.0% in two investigated communities 8 . The ease of access by this group to urban areas through river vessels can contribute to changing lifestyles, such as increased consumption of processed foods, resulting in a higher prevalence of DM 33 . Also, the Mais Médicos (More Doctors) project, established in 2017, the expansion of the Basic River Health Units (UBSF) and the establishment of the Riverbank Family Health Teams (eSFR) and the River Family Health Teams (eSFF) contribute to greater screening and diagnosis of DM in this population [34][35][36] .
Regarding health conditions, hypertension and hypercholesterolemia were associated with DM, a result consistent with previous studies in rural and urban areas 1,27,37 . In the Brazilian rural population, both conditions are frequent, with prevalence between 20 and 35% 18,19,37 . Specifically in rural traditional populations, some studies explain these associations by genetic components and sociodemographic factors, such as low education and difficulties in accessing health services 19,37 . Also noteworthy is the increase in the frequency of unhealthy behaviors, with the ease of consumption of inappropriate foods due to the urban-rural connection, and better income conditions, linked to such social welfare projects as Bolsa Família, Bolsa Floresta and Seguro Defeso 10,18,19 .
Negative self-perception of health is also an indicator that is often associated with DM, which may be linked to the demand for care perceived by the need for continuous metabolic and glycemic control, the complexity of care and the possibility of complications 14 . In rural populations, this negative perception may be greater, mainly due to the risk of developing foot neuropathies, bone deformities, fungal infections and ulcers, due to housing and occupation conditions related to agribusiness and agriculture 9,16 .
Being a former smoker was associated with DM, similarly to previous studies 38,39 . The rural area has a significant prevalence of smokers and ex-smokers, ranging from 10 to 16.6%, higher than in urban areas, which favors the association with DM 15,38,39 .
Alcohol use was a protective factor for DM in this study. PNS data with urban populations show that alcohol abuse is associated with DM 23 . However, a meta-analysis of 38 observational studies showed that reduction in the risk of type 2 DM was present at all levels of alcohol intake up to 63 g per day, with greater risks above this threshold, especially in women and in Western populations 40 . Specifically in rural populations in China, only the consumption of high daily doses of alcohol was related to an increase in the risk of type 2 DM 17 . In this study, information on the daily dose of alcohol and type of beverage consumed was not collected, which constitutes a limitation for comparison with the cited works. Despite this, it is important to point out alcohol as a risk factor for DM and other chronic diseases, accidents and violence, as well as the difficulties in seeking health services by rural populations in view of the possible consequences of its use.
As for indicators related to availability and demand for health services, although they did not remain associated with DM in the multiple analysis, they should be considered. In the present study, between 60 and 70% of the population with DM received a visit from a CHA or sought medical attention with a general practitioner in the last year, which shows that people with NCDs tend to seek health services more assiduously 10,12,34,38 . In Brazil, the National Policy of Comprehensive Health for People of the Countryside, Forests and Waters foresees in its first axis the guarantee of the population's access to quality services, with equity and in an adequate time to meet their needs 10 . Access to services by rural traditional communities occurs mainly through the ESF, which ensures the territorialization and coverage of a given area by a multidisciplinary team 30,31 , enabling the construction of relationships of bond and trust, as well as continuity, resolution and longitudinality of care 29 . However, access to health services by the rural population faces several difficulties, such as geographical barriers that make it difficult for professionals to travel to communities and for users themselves to get to health services when necessary in a timely manner 4,11 .
The use of health insurance and the search for a specialist among individuals with DM were also found in the PNS of 2019 4 , because of the need to have an assistance complement to public health services, which still have structural and organizational weaknesses in the care of patients 4,11,31 .
Some limitations can be highlighted in this investigation. First, there were the limitations of cross-sectional studies, which restrict inferences about the directionality of some associations in the multiple analysis model. Second, the use of a self-reported measure can lead to underestimations, given that it depends on access to diagnostic services. Finally, the evaluation of some exposure variables, such as physical activity and alcohol, may have suffered information bias due to the way in which it was performed.
Despite these limitations, this study makes several contributions. The results are unprecedented in the state of Goiás and allow us to establish a set of factors associated with DM, contributing to better planning and execution of intersectoral strategies and to the synthesis of evidence for the process of formulating and implementing policies and programs for rural populations, enabling better quality of life for rural traditional populations in the state, as provided for in the Evidence-Informed Policy 41 . Specifically, in Goiás, such actions may involve the sectors of the Department of Education, Sport and Leisure, of Transport, and also of the State Department of Agriculture, Livestock and Supply (SEAPA), responsible for rural traditional areas in the state 42 .
Finally, this study allowed us to identify the prevalence of and factors associated with DM in rural communities, pointing out possible choices, due to unhealthy lifestyle habits in the communities, which result in hypertension, hypercholesterolemia and smoking history, factors associated with DM. Accordingly, the risk of complications from DM can become even greater, which requires more intersectoral and singular strategies, which are appropriate in relation to income distribution and the use and appreciation of resources already available in communities. For future research, an objective assessment of DM is recommended, as well as the assessment of other variables related to access and use of health services and healthy habits, and also existing barriers in the view of users and professionals.