Geographical disparities in obesity prevalence: small-area analysis of the Chilean National Health Surveys

Halting the increase in obesity prevalence is an ever-pressing issue for public health policymakers. The worldwide rate of obesity has trebled over the past four decades, reaching 650 million (13%) and 1.9 billion (39%) obese and overweight adults in 2016, accounting for more deaths than undernutrition. This problem is more challenging in middle-low income countries due to the fact that the rapid upsurge of obesity coexists with undernutrition, while resources to deal with both are scarce [1, 2].

National Health Surveys (NHS) and complex surveillance systems are the usual frameworks to quantify the burden of diseases for planning interventions and resource allocation. Nevertheless, surveillance systems are seldom available in lower-middle-income countries. Moreover, the NHS is not updated as fast as needed and only represents large geographical areas (e.g., national and urban-rural levels). On the other hand, national-level or large geographical area estimates do not account for the disease variability in smaller geographical areas within a country, potentially missing information and target populations. For example, emerging evidence reported substantial geographical disparities in childhood mortality, cardiovascular and suicide rates [3‐5], infectious diseases [6‐8], and healthcare costs [9, 10] at subnational levels, highlighting the importance of geographic variation and the fast-growing demand for reliable small-area estimates. Concerning obesity, some studies based on NHS and surveillance records found a substantial difference in the small-area pattern of obesity rates among adults in high-income countries [11‐16] and in children [17‐19]. Since these studies cross-sectionally estimate obesity rates in smaller areas [20, 21], there is a gap of knowledge about the increase of the obesity rate among adults, especially in a setting where low-middle income households persist.

Chile can be a benchmark for anticipating the course and prevention policies of obesity in medium-low income settings due to the country experienced the fastest nutritional transition during the last four decades, tracking from high rates of undernutrition in the 1960s to high rates of obesity in the 2000s [22]. According to the World Bank, Chile has experienced the fastest growing economy in Latin America in recent decades; nevertheless, it has substantial socioeconomic variability. Thus, low-middle settings at subnational levels or lower geographic administrative divisions persist [23]. The last Chilean NHS reported that the obesity prevalence was 31% in 2016, affecting predominantly women and urban areas, with no examination of regional variation of the increase of obesity rates or the potential influence of smaller geographic areas.

The main objective of this study was to analyze the geographic variation of the increase of the obesity prevalence rates among adults and adolescents (≥ 15 years) at the national and regional levels, which are appropriate and valid for inference, comparing data from two representative NHS from Chile. Secondly, we assessed the geographic variation of the obesity rate in a smaller geographical unit related to health care services distribution, using small-area estimations (SAE).

A total of 4884 (90.2%) and 5220 (83.7%) complete registers of NSH-2009 and NSH-2016, respectively, were analyzed in this study. Sample size distribution by the regional level and regions aggregated by macro zones are displayed in Table 1. Chile has 15 regions with 29 HS imbalanced distributed across each region. For instance, the Valparaiso, Metropolitan, Biobío, La Araucanía, and Los Lagos regions have from 2 to 6 HS areas; in contrast, there is just one HS for the rest of the regions (Table 1).

Table 2 shows the sample size distribution by the subpopulations of the study. In NSH-2009, 59.7% of the sample were female, 24.7% older than 65 years, and 26.6% low socioeconomic status. While 63.4% were female, 24.7% older than 65 years, and 22.3% had low socioeconomic status in the NSH-2016 (Table 2).

To our knowledge, this is the first study that reported the geographic variation of obesity rate increments and obesity rates at smaller geographic areas in Chile. The main purpose of studying geographic variation in disease rates is to provide health management information and formulate hypotheses about the environmental determinants of disease. The “hot spot” can also be identified as a high concentration of cases in a particular area. Subsequently, the correlation between geographic disease variation and concomitant variation in the degree of exposure to the environment or lifestyle can be addressed [13, 24]. Identifying small areas of highest prevalence allows to address the health consequences of obesity and provide specific resources to provide an early diagnosis and timely treatment. Moreover, identifying environmental drivers will promote better urban planning [27, 28].

Chile presents a great regional variation in socio-economic development and disparities in health indicators [23]. Consequently, our study’s results show that the prevalence of obesity and its increments are very heterogeneous in the territory, even within regions. Disclosing that the “hot spots” are different between subnational levels and smaller areas, represented by regions and the HS, respectively, and suggesting geographic disparities in obesity. Scrutinizing our results, HS areas with higher obesity prevalence coincide with less developed communes in Chile, according to the Community Development Indicator (CDI) [23], even in regions with better socioeconomic development. However, this seems to be different in the relative increments.

The Metropolitan Region (the region with the most population density), as an example, has the highest CDI (0.50), and its obesity rate is near to the national average (31.7%, NHS-2016). Nevertheless, HS areas belonging to less developed communes showed higher obesity rates (M. Occidente: 40%, M. Norte 35%), contrary to those with higher CDI (M. Oriente 26%). On the other hand, other HS areas have the largest relative increases (M.Central and M. Suroriente). Even though there is a knowledge gap regarding factors that could explain the observed differences, further hypotheses to be tested are the influence of socioeconomics or educational factors, and other environmental drivers of obesity. Literature also suggests access to green areas that encourage physical activity, safe public transport planning and promote joint mobility with bicycles and walks, barriers to the consumption of junk food, etc.

In addition, our findings based on the relative obesity increases indicators make visibly hidden “hot spots” susceptible to earlier preventive interventions. Males, the youngest age group, highest socioeconomic status, and regions from the central and norths zone of the country are becoming the targeted groups of concern. High prevalence and the emerging relative obesity increases in other subpopulations also lead to reflection on the public policy measures taken in the country to solve or confront the problem. The establishment of laws (such as nutrition labeling) as the only preventable intervention has been insufficient to curb obesity. The above reinforces the importance of examining prevalence at sub-national labels and the need to adjust preventive measures to the local reality to face the determinants of this problem. These can be studied using modern approaches such as big data, artificial intelligence, geo-analysis, agent-based models, and public health dashboards [29].

Strengths of our study include the large sample size, which represents adolescents and adults 15 years of age and older, and the standardized weight and height measurements conducted by trained personnel in both surveys, which gives reliable indicators for comparison. Traditional and well-known modeling for SAE allowed us to estimate obesity prevalence in smaller areas, where sample sizes are too small or insufficient. Although the above could explain the high prevalence due to high variability, evidence supports the reliability of different SAE approaches; moreover, our interim analysis reinforces the robustness of our estimates (Supplement, Fig. 5). Still, local surveillance data for direct estimation would be preferable, and local government should consider resources to better track obesity and other obesity-related chronic health conditions at this level.

The principal limitation of our study is that we have no complete data at the province or commune level. Instead, we use HS data as a proxy variable and the education level as a proxy socioeconomic status. HS data may misrepresent communes, especially in regions with less health service access or rural areas. Nevertheless, this is the first study exploring the utility of this data and brightening to light important geographic variations in obesity prevalence among adults and adolescents. In addition, due to the potential for bias, ecological studies should only be considered the first step in investigating the potential associations between environmental agents and disease geography differences [30].

In conclusion, this study determined that the obesity rates and relative increase at subnational levels substantially differ from the national estimates, highlighting geographical disparities in obesity prevalence among adults and adolescents. These findings suggest that stakeholders should stake prevention programs and resource investment locally and evaluate determinants of these disparities and their impact on other diseases.