Sample Survey Design Research Articles

Method comparison and estimation of causal effects of insomnia on health outcomes in a survey sampled population

Applying causal inference methods, such as weighting and matching methods, to a survey sampled population requires properly incorporating the survey weights and design to obtain effect estimates that are representative of the target population and correct standard errors (SEs). With a simulation study, we compared various approaches for incorporating the survey weights and design into weighting and matching-based causal inference methods. When the models were correctly specified, most approaches performed well. However, when a variable was treated as an unmeasured confounder and the survey weights were constructed to depend on this variable, only the matching methods that used the survey weights in causal estimation and as a covariate in matching continued to perform well. If unmeasured confounders are potentially associated with the survey sample design, we recommend that investigators include the survey weights as a covariate in matching, in addition to incorporating them in causal effect estimation. Finally, we applied the various approaches to the Hispanic Community Health Study/Study of Latinos (HCHS/SOL) and found that insomnia has a causal association with both mild cognitive impairment (MCI) and incident hypertension 6–7 years later in the US Hispanic/Latino population.

Adapting the sampling design of research surveys to improve the biomass estimation of secondary species—the case study of Raja clavata

Abstract Research surveys are important to evaluate fishery resources’ spatial distribution and abundance. Although the underlying sampling is usually conceived with a focus on specific species, efficient designs can also collect data on secondary species. We present a framework to explore and evaluate the adequacy of alternative sampling designs for fishery research surveys aiming to maximize accuracy estimates of the secondary species abundance while maintaining the quality of the abundance estimates of primary species. A geostatistical model-based approach was developed considering the semi-continuous nature of the data and the excess of zero values commonly observed for secondary species. New sampling designs were defined according to optimization weights and evaluated based on the resulting prediction exactness. The framework was applied to the bottom trawl survey conducted along the Portuguese continental coast with European hake, Merluccius merluccius, as the primary species and thornback ray, Raja clavata, as the secondary species. The sampling design URSI provided the best balance between the accuracy for both primary and secondary species. The methodology can be replicated for other bottom trawl research surveys and an extended set of species. We recommend that a decision on which sampling design to adopt in future surveys should consider a cost-efficiency analysis.

Combining surveys in small area estimation using area‐level models

AbstractFor many sample surveys, researchers, policymakers, and other stakeholders are interested in obtaining estimates for various domains, such as for geographic levels, for demographic groups, or a cross‐classification of both. Often, the demand for estimates at a disaggregated level exceeds what the sample size and survey design can support when estimation is done by traditional design‐based estimation methods. Small area estimation involves exploiting relationships among domains and borrowing strength from multiple sources of information to improve inference relative to direct survey methods. This typically involves the use of models whose success depends heavily on the quality and predictive ability of the sources of information used. Possible sources of auxiliary information include administrative records, Censuses, big data such as traffic or cell phone data, or previous vintages of the same survey. One rich source of information is that of other surveys, especially in countries like the United States, where multiple surveys exist that cover related topics. We will provide an introduction to the topic of combining information from multiple surveys in small area estimation using area‐level models, including practical advice and a technical introduction, and illustrating with applications. We will discuss reasons to combine surveys and give an overview of some of the most common types of models.This article is categorized under: Statistical Models > Multivariate Models

Preferred Reporting Items for Complex Sample Survey Analysis (PRICSSA)

Abstract Methodological issues pertaining to transparency and analytic error have been widely documented for publications featuring analysis of complex sample survey data. The availability of numerous public use datasets to researchers without adequate training in using these data likely contributes to these problems. In an effort to introduce standards for reporting analyses of survey data and promote replication, we propose the Preferred Reporting Items for Complex Sample Survey Analysis (PRICSSA), an itemized checklist to guide researchers publishing analyses using complex sample survey data. PRICSSA is modeled after other checklists (e.g., PRISMA, CONSORT) that have been widely adopted for other research designs. The PRICSSA items include a variety of survey characteristics, such as data collection dates, mode(s), response rate, and sample selection process. In addition, essential analytic information—such as sample sizes for all estimates, missing data rates and imputation methods (if applicable), disclosing if any data were deleted, specifying what survey weight and sample design variables were used along with method of variance estimation, and reporting design-adjusted standard errors/confidence intervals for all estimates—are also included. PRICSSA also recommends that authors make all corresponding software code available. Widespread adoption of PRICSSA will help improve the quality of secondary analyses of complex sample survey data through transparency and promote scientific rigor and reproducibility.

Creating measurement-based oil and gas sector methane inventories using source-resolved aerial surveys

Critical mitigation of methane emissions from the oil and gas (OG) sector is hampered by inaccurate official inventories and limited understanding of contributing sources. Here we present a framework for incorporating aerial measurements into comprehensive OG sector methane inventories that achieves robust, independent quantification of measurement and sample size uncertainties, while providing timely source-level insights. This hybrid inventory combines top-down, source-resolved, multi-pass aerial measurements with bottom-up estimates of unmeasured sources leveraging continuous probability of detection and quantification models for a chosen aerial technology. Notably, the technique explicitly considers skewed source distributions and finite facility populations that have not been previously addressed. The protocol is demonstrated to produce a comprehensive upstream OG sector methane inventory for British Columbia, Canada, which while approximately 1.7 times higher than the most recent official bottom-up inventory, reveals a lower methane intensity of produced natural gas (<0.5%) than comparable estimates for several other regions. Finally, the method and data are used to upper bound the potential influence of source variability/intermittency, directly addressing an open question in the literature. Results demonstrate that even for an extreme case, variability/intermittency effects can be addressed by sample size and survey design and have a minor impact on overall inventory uncertainty.

Inverse Association of the Adequacy and Balance Scores in the Modified Healthy Eating Index with Type 2 Diabetes in Women.

Type 2 diabetes (T2DM) has markedly increased among Asians as their diets and lifestyles become more westernized. We, therefore, investigated the hypothesis that the Korean healthy eating index (KHEI) scores are associated with gender-specific T2DM risk in adults. The hypothesis was tested using the data from the Korea National Health and Nutrition Examination Survey-VI (2013-2017) with a complex sample survey design. Along with the KHEI scores, the modified KHEI (MKHEI) scores for the Korean- (KSD) and Western-style diets (WSD) were used as independent parameters, calculated using a validated semi-quantitative food-frequency questionnaire (SQFFQ). We estimated the association between the KHEI or MKHEI and the T2DM risk using logistic regression after adjusting for T2DM-related covariates. The adults with T2DM were more frequently older men who were less educated, married, on a lower income, and living in rural areas compared to those without T2DM. Not only the fasting serum glucose concentrations but also the waist circumferences and serum triglyceride concentrations were much higher in adults with T2DM than in those without T2DM in both genders. Serum HDL concentrations in the non-T2DM subjects exhibited a greater inverse relationship to serum glucose than in the T2DM group in both genders. Twenty-four-hour recall data revealed that women, but not men, had higher calcium, vitamin C, saturated and monounsaturated fatty acids, retinol, and vitamin B2 intakes than the T2DM group. Furthermore, overall, the KHEI score and the adequacy and balance scores among its components were significantly higher in the non-T2DM group than in the T2DM group, but only in women. The KHEI scores were inversely associated with T2DM only in women. The mixed grain intake score was higher in the non-T2DM than the T2DM group only in men. However, there were no differences between the groups in the MKHEI scores for KSD and WSD. In conclusion, high KHEI scores in the adequacy and balance components might prevent and/or delay T2DM risk, but only in women.

Quality of Water from Hand-Dug Wells in Nembe Town, Bayelsa State, Nigeria

This study examined the quality of water supply from hand-dug wells in Nembe town, Bayelsa State using sample survey design. Data was obtained from primary sources. To assess if the quality of groundwater supplied in Nembe town is safe for domestic and drinking purposes, the physico-chemical and microbial status of the groundwater were studied. Nembe town was stratified into ten neighbourhoods and a total of ten (10) hand dug well water samples was randomly collected from each of the sampled ten neighbourhoods and analyzed for twenty-one (21) paramaters in the month of November 2019. These samples were sent to the laboratory for physico-chemical and microbial analysis. The results of the study revealed that physico-chemical paramaters such as Total Dissolved Solids (TDS), Total Suspended Solids(TSS), Sulphate (So4-2), Bicarbonate (HCo3-), Calcium (Ca), Sodium (Na), Manganese (Mn), Electrical Conductivity (EC), Iron (Fe), Chloride (CL), Nitrate (NO), Salinity, Total Alkalinity (TA) and Magnesium (Mg) are all within the permissible level of WHO and NESREA drinking water standards. Groundwater in the area have turbidity and potassium levels higher than the WHO and NESREA maximum permissible limits for drinking water standard. All the sampled hand dug wells have pH value within the permissible limit of drinking water except one which had pH value of 6.48 and considered too acidic for human consumption. Based on hardness, 70% of the groundwater samples are slightly hard water to hard water in nature. The result of the microbial analysis showed the presence of total coliforms, Total bacteria counts and fungi in all the hand dug wells. The ANOVA statistical tool was employed in testing the stated hypothesis and the results show that there is no significant spatial variation in the physico-chemical properties of water in the study area and that there is a significant spatial variation in the microbial properties of water in the study area. The result of the study was conclusive evidence that all the hand dug wells studied are not safe for domestic purpose especially for drinking as they were affected by some physico-chemical and biological parameters. Hence the affected wells call for periodic monitoring and treatment to improve the quality of the water.

Veterans Health Administration Healthcare Coverage and Medical Financial Hardship in Low-Income Veterans

The Veterans Health Administration (VA) provides low- to no-cost care to enrolled veterans with low incomes. This study assessed the associations between VA coverage and medical financial hardship among U.S. veterans with low incomes. Using 2015-2018 National Health Interview Survey data, veterans aged ≥18 years with incomes <200% of the Federal Poverty Level were identified (crude n=2,468, weighted n=3,872,252). Four types of medical financial hardship were assessed: objective, and subjective material, psychologic, and behavioral medical financial hardship. Survey-weighted proportions of veterans with medical financial hardship were calculated, and adjusted probabilities of medical financial hardship that accounted for Veteran characteristics, year-fixed effects, and survey sampling design were estimated. Analyses were conducted from August through December 2022. Overall, 34.5% of veterans with low incomes had VA coverage. Among veterans without VA coverage, 38.7% had Medicare insurance, 18.2% had Medicaid insurance, 16.5% had private insurance, 13.5% had other public insurance, and 13.1% were uninsured. In adjusted analyses, veterans with VA coverage had lower probabilities of objective (-8.13 percentage point, p=0.008), subjective material (-6.55 percentage point, p=0.034), subjective psychologic (-10.33 percentage point, p=0.003), and subjective behavioral (-6.72 percentage point, p=0.031) medical financial hardship than veterans with Medicare and no VA coverage. VA coverage was associated with protection against four types of medical financial hardship among veterans with low incomes, yet many are not enrolled. Research is needed to understand reasons these veterans lack VA coverage and to identify strategies to address medical financial hardship.

Investigation and Research on the Cultural Construction of University Students' Associations

To investigate the construction of association culture, this study used a multi-stage sampling survey design, which is a cost- and time-effective probability sampling method. The research sample consisted of 858 Chinese college students who took part in a questionnaire study. The authors innovatively constructed an index system that can objectively evaluate the satisfaction of association construction and used the principal component method to descend to construct the comprehensive indicator of satisfaction. The authors excavate the key influencing factors of association culture construction and finally summarize them into four factors, which are internal construction factor of the association, resource allocation factor of the association, school management factor of the association and instructor factor of association. Finally, regression analysis indicated that the internal and material building of associations be prioritized, and the school authorities should give more financial support and create a relaxed environment for the culture development of associations.

Barriers to Research: A National Survey of Physician Assistant Faculty.

Success in research is a well-known challenge for physician assistant (PA) educators. Multiple barriers to research productivity have been proposed. This study examines potential barriers to PA research productivity. Two hundred forty-seven PA programs were stratified by region and Carnegie classification, creating 28 strata. One program was randomly sampled from each stratum, and faculty from sampled programs were invited to participate (N = 214). An electronic survey was deployed via Qualtrics. Data were analyzed, accounting for the stratified survey sample design. Reported are proportions with associated 95% confidence intervals. Survey response rate was 53% (N = 114). The mean age was 46.7 years [44.5, 48.8 years] and 61% [52.4, 70.4] of respondents were female. Most respondents (53.5%; [44.2, 62.8]) were assistant professors and reported being employed as PA faculty for a mean of 12.4 years [11.2, 13.7 years]. Most respondents (68.1%; [59.7, 76.6]) reported 0 peer-reviewed research articles regardless of number of years as faculty. Few reported an interest in research (22.3%; [14.3, 29.9]) or felt internally driven to perform research (35.4%; [26.6, 44.2]). Most pursued a faculty position to teach (78.1%; [70.5, 85.7]) and spent little time pursuing research. Only 15.8% ([9.7, 21.9]) received research mentorship. This study identified lack of interest in research by faculty and lack of mentorship engagement as important factors underlying the profession's poor scholarly output. Initiatives to increase scholarship must first focus on fostering individual interest and increasing motivation.

Sampling Survey Design Presidential Election Quick Count Sumatera Island

The number of TPS on the island of Sumatra is very large, in order to save time and money in conducting surveys, a sampling survey design was created. The purpose of this study is to predict the results of the presidential election on the island of Sumatra. The TPS sample frame was obtained in four stages where each stage used a sampling technique, namely the first and second stages used stratified random sampling, the third stage used systematic random sampling, and the last used clusters. The results obtained are with different TPS sample sizes showing the same results. The victory in the presidential election on the island of Sumatra was won by candidate pair number two. Then compared with the overall TPS population in Sumatra. Based on the population, the second candidate pair is also superior. So it can be concluded that the use of a survey sampling design in this study is appropriate in predicting the results of the elected president election.

Physical Activity Levels (PAL) in US Adults-2019.

Physical activity levels (PAL) are associated with mortality risk and were instrumental in estimating national energy requirements, but we are unaware of population-based estimates of PAL in US adults. Thus, we conducted a nationwide survey using a validated previous-day recall method to estimate PAL and the behavioral determinants of low and higher PAL. Participants from the AmeriSpeak panel 20-75 yr of age ( N = 2640) completed Activities Completed over Time in 24-hours previous-day recalls. PAL values were estimated as the average metabolic equivalent value over 24 h. Recalls were conducted on randomly selected days in October and November 2019. Survey sample design weights were applied to reflect the US population. Mean age was 45.3 yr, 51% were female, 67% were non-Hispanic White, and 37% had a body mass index of ≥30 kg·m -2 . US adults reported a mean PAL of 1.63 (95% confidence interval, 1.60-1.65), and 39% (37%-42%) of adults reported PAL ≥1.6 on a given day. Men reported higher PAL than women (1.67 vs 1.59), and older adults reported lower PAL. Adults with PAL <1.4 spent 81% (12.1 h·d -1 ) of their waking day sedentary and 19% (2.7 h·d -1 ) in total physical activity. Adults with PAL considered to be "active"(1.6-1.89) spent only 49% (8.0 h·d -1 ) of their waking day sedentary, and 51% (8.3 h·d -1 ) physically active. Our study provides novel estimates of PAL in a nationwide sample of US adults and a description of the type and intensity of sedentary and physically active behaviors contributing to low and higher PAL. These findings may inform public health messages aimed at increasing physical activity in adults and potentially contribute to obesity prevention efforts.

Racial disparities and prevalence of cardiovascular disease risk factors, cardiometabolic risk factors, and cardiovascular health metrics among US adults: NHANES 2011–2018

This study estimated the prevalence of cardiovascular disease (CVD) risk factors, cardiometabolic (CM) risk factors, and cardiovascular health metrics (CVHMs) among US adults and across race/ethnicity groups. The study comprised 8370 US adults aged ≥ 20 years from the National Health and Nutrition Examination Survey (NHANES) 2011–2018, free of coronary heart disease/heart failure, angina/angina pectoris, heart attack, and stroke, who provided complete data for the outcome variables of interest. Age-adjusted prevalence of CVD and CM risk factors, and CVHMs were computed for all adults and across race/ethnicity groups. All analyses accounted for the complex, multi-stage survey sampling design of the NHANES. Hypertension (45.0%), obesity (40.0%), fasting plasma glucose ≥ 100 mg/dL or hypoglycemic medication (51.0%), ideal physical activity (59.2%) and ideal smoking status (56.9%) were most prevalent for the whole sample. Mexican Americans and non-Hispanic Blacks had elevated risk for some, but not all, CVD and CM risk factors compared to non-Hispanic Whites and non-Hispanic Asians. Reducing further health disparities and persisting differences among racial and ethnic groups is vital to achieving the American Heart Association vision of all people having ideal cardiovascular health, living healthier and longer.

Personal factors and the role of memory in faculty refinding of stored information

PurposeThe study examined the personal information management (PIM) challenges encountered by faculty in six universities in Ghana, their information refinding experiences and the perceived role of memory. The study tested the hypothesis that faculty PIM performance will significantly differ when the differences in the influence of personal factors (age, gender and rank) on their memory are considered.Design/methodology/approachThe study was guided by a sample survey design. A questionnaire designed based on themes extracted from earlier interviews was used to collect quantitative data from 235 faculty members from six universities in Ghana. Data analysis was undertaken with a discrete multivariate Generalized Linear Model to investigate how memory intermediates in the relationship between age, gender and rank, and, refinding of stored information.FindingsThe paper identified two subfunctions of refinding (Refinding 1 and Refinding 2) associated with self-confidence in information re-finding, and, memory (Memory 1 and Memory 2), associated with the use of complimentary frames to locate previously found and stored information. There were no significant multivariate effects for gender as a stand-alone variable. Males who were aged less than 39 could refind stored information irrespective of the memory class. Older faculty aged 40–49 who possess Memory 1 and senior lecturers who possess Memory 2 performed well in refinding information. There was a statistically significant effect of age and memory; and rank and memory.Research limitations/implicationsThis study was limited to faculty in Ghana, whereas the study itself has implications for demographic differences in PIM.Practical implicationsIdentifying how memory mediates the role of personal factors in faculty refinding of stored information will be necessary for the efforts to understand and design systems and technologies for enhancing faculty capacity to find/refind stored information.Social implicationsUnderstanding how human memory can be augmented by technology is a great PIM strategy, but understanding how human memory and personal factors interplay to affect PIM is more important.Originality/valuePIM of faculty has been extensively examined in the literature, and limitations of memory has always been identified as a constraint. Human memory has been augmented with technology, although the outcome has been very minimal. This study shows that in addition to technology augmentation, personal factors interplay with human memory to affect PIM. Discrete multivariate Generalized Linear Model applied in this study is an innovative way of addressing the challenges of assimilating statistical methodologies in psychosocial disciplines.

Examining racial and ethnic disparities in adult emergency department patient visits for concussion in the United States.

BackgroundRacial and ethnic differences in emergency department (ED) visits have been reported among adolescent patients but are unsubstantiated among adults. Therefore, our purpose in this study was to examine the relationship between race/ethnicity and adult ED visits for concussions, their injury mechanisms, and computed tomography (CT) scan use among a nationally representative sample.MethodsWe used the National Hospital Ambulatory Medical Care Survey database from 2010–2015 to examine 63,725 adult (20–45 years old) patient visits, representing an estimated 310.6 million visits presented to EDs. Of these visits, 884 (4.5 million national estimate) were diagnosed with a concussion. Visit records detailed patient information (age, sex, race/ethnicity, geographic region, primary payment type), ED visit diagnoses, injury mechanism (sport, motor vehicle, fall, struck by or against, “other”), and head CT scan use. The primary independent variable was race/ethnicity (non-Hispanic Asian, non-Hispanic Black or African American, Hispanic/Latinx, non-Hispanic multiracial or another, and non-Hispanic White). We used multivariable logistic and multinomial regression models with complex survey sampling design weighting to examine the relationship between concussion ED visits, injury mechanisms, and CT scan use separately by race/ethnicity while accounting for covariates.ResultsThere were no associations between race/ethnicity and concussion diagnosis among adult ED visits after accounting for covariates. Relative to sports-related injuries, non-Hispanic Black or African American patient visits were associated with a motor vehicle (OR = 2.69, 95% CI: 1.06–6.86) and “other” injury mechanism (OR = 4.58, 95% CI: 1.34–15.69) compared to non-Hispanic White patients. Relative to sports-related injuries, non-Hispanic Asian, multiracial, or patients of another race had decreased odds of falls (OR = 0.20, 95% CI: 0.04–0.91) and “other” injuries (OR = 0.09, 95% CI: 0.01–0.55) compared to non-Hispanic White patients. The odds of a CT scan being performed were significantly lower among Hispanic/Latinx patient visits relative to non-Hispanic White patients (OR = 0.52, 95% CI: 0.30–0.91), while no other race/ethnicity comparisons differed.ConclusionOur findings indicate that the overarching concussion ED visit likelihood may not differ by race/ethnicity in adults, but the underlying mechanism causing the concussion and receiving a CT scan demonstrates considerable differences. Prospective future research is warranted to comprehensively understand and intervene in the complex, multi-level race/ethnicity relationships related to concussion health care to ensure equitable patient treatment.

The estimated impact of changes to otolith field-sampling and ageing effort on stock assessment inputs, outputs, and catch advice

Generating accurate data for stock assessments is resource-demanding, necessitating periodic evaluation of survey sampling designs and potential impacts on stock assessments. We developed a framework for bootstrapped resampling of survey age data and calculation of input sample sizes as a function of among-bootstrap variance in age compositions. Data from this bootstrap estimator were then used to evaluate the influence of alternative sampling rates and methods on uncertainty in estimates of the overfishing limit (OFL) calculated using stock assessment models. For dusky rockfish ( Sebastes variabilis) and Pacific ocean perch ( Sebastes alutus), a 10% decrease in the number of tows sampled upon led to a predicted 5%–6% increase in the CV of OFL (log–log slope = −0.576 to −0.486), which was greater than the 0%–2% increase from a 10% decrease in otoliths-per-tow (log–log slope = −0.238 to −0.029). Application of this approach across all stocks monitored in the survey of interest is required to identify which stocks ( i) benefit the most from increased sampling of ageing structures or ( ii) cost the least in terms of OFL uncertainty owing to reduced sampling.

Pseudo Bayesian Mixed Models under Informative Sampling

Abstract When random effects are correlated with survey sample design variables, the usual approach of employing individual survey weights (constructed to be inversely proportional to the unit survey inclusion probabilities) to form a pseudo-likelihood no longer produces asymptotically unbiased inference. We construct a weight-exponentiated formulation for the random effects distribution that achieves approximately unbiased inference for generating hyperparameters of the random effects. We contrast our approach with frequentist methods that rely on numerical integration to reveal that the pseudo Bayesian method achieves both unbiased estimation with respect to the sampling design distribution and consistency with respect to the population generating distribution. Our simulations and real data example for a survey of business establishments demonstrate the utility of our approach across different modeling formulations and sampling designs. This work serves as a capstone for recent developmental efforts that combine traditional survey estimation approaches with the Bayesian modeling paradigm and provides a bridge across the two rich but disparate sub-fields.

Utilization of Screening Mammography in Women Before 50: Cross-Sectional Survey Results from the National Health Interview Survey

While the American College of Radiology recommends annual screening mammography starting at age 40 years, the US Preventive Services Task Force (USPSTF) recommends that screening mammography in women younger than age 50 years should involve shared- decision making (SDM) between clinicians and patients, considering benefits and potential harms in younger women. Using a nationally representative cross-sectional survey, we aimed to evaluate patient-reported reasons and predictors of screening mammography utilization in this age group. Respondents aged 40-49 years from the 2018 National Health Interview Survey (NHIS) without a history of breast cancer were included (response rate 64%). Participants reported sociodemographic variables and reasons they did not engage in mammography screening within the last two years. Multiple variable logistic regression analyses were performed to evaluate the association between sociodemographic characteristics and patient-reported screening mammography use, accounting for complex survey sampling design elements. 1,948 women between the ages of 40-49 years were included. Of this group, (758/1948) 46.6% reported receiving a screening mammogram within the last year, and 1196/1948 (61.4%) reported receiving a screening mammogram within the last two years. The most common reasons for not undergoing screening included: "No reason/never thought about it" 744/1948 (38.2%), "Put it off" 343/1948 (17.6%), "Didn't need it" 331/1948 (16.9%), "Doctor didn't order it" 162/1948 (8.3%), and "I'm too young" 63/1948 (5.3%). Multiple variable analyses demonstrated that lack of health insurance was the strongest predictor of mammography non-engagement (p< 0.001). Deficits in shared- decision-making in women younger than 50 years related to mammography utilization exist. Radiologists may be key in addressing this issue among ambulatory care providers and patients, educating about the benefits and harms of screening younger women, particularly in racial/ethnic minorities and uninsured patients, who experience additional barriers to care and SDM discussions.

Time trends in the prevalence and determinants of age-appropriate breast feeding among children aged 0–23 months in Ghana: a pooled analysis of population-based surveys, 2003–2017

ObjectiveWe assessed the sociodemographic and maternal–child characteristics associated with age-appropriate breast feeding among children aged 0–23 months in Ghana.MethodsWe pooled data on 12 743 children aged 0–23 months from three...

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

BackgroundPrevious representative health surveys conducted in Chile evidenced a high obesity prevalence rate among adults, especially in female and urban areas. Nevertheless, these have limited utility for targeted interventions and local source allocation for prevention. This study analyzes the increments in obesity prevalence rates in populations ≥15 years of age and the geographic variation at the regional level. We also assessed whether the obesity rates have different patterns on a smaller geographic level than national and regional ones.MethodsThis ecological study analyzed data from two representative national samples of adolescents and adults ≥15 years old, who participated in the last Chilean health surveys, 2009 (n = 5412) and 2016 (n = 6233). Obesity (body mass index≥30 kg/m2) rates were calculated on the national, regional, and Health service (HS) levels, being HS the smallest unit of analysis available. Obesity rates and relative increase to early identify target populations and geographic areas, with 95% confidence intervals (95% CI), were calculated using the sampling design of the national surveys, at the national and regional level, and by gender, age groups, and socioeconomic status. The Fay-Herriot (FH) models, using auxiliary data, were fitted for obesity rate estimates at the HS level.ResultsThe relative increase in obesity rate was 37.1% (95%CI 23.3–52.9) at the national level, with a heterogeneous geographic distribution at the regional one. Southern regions had the highest obesity rates in both surveys (Aysén: 35.2, 95%CI 26.9–43.5 in 2009, 44.3 95%CI 37–51.7 in 2016), but higher increases were predominantly in the northern and central areas of the country (relative increase 91.1 95%CI 39.6–110.1 in Valparaiso and 81.6 95%CI 14.4–196.2 in Tarapacá). Obesity rates were higher in females, older age, and lower socioeconomic groups; nevertheless, relative increases were higher in the opposite ones. The FH estimates showed an obesity rates variation at the HS level, where higher rates tend to converge to specific HS areas of each region.ConclusionObesity rates and relative increase are diverse across subnational levels and substantially differ from the national estimates, highlighting a pattern that converges to areas with low-middle income households. Our results emphasize geographical disparities in obesity prevalence among adults and adolescents.