Search

This paper introduces a bibliometric, citation network-based method for assessing the social validation of novel research, and applies this method to the development of high-throughput toxicology research at the US Environmental Protection Agency. Social validation refers to the acceptance of novel research methods by a relevant scientific community; it is formally independent of the technical validation of methods, and is frequently studied in history, philosophy, and social studies of science using qualitative methods. The quantitative methods introduced here find that high-throughput toxicology methods are spread throughout a large and well-connected research community, which suggests high social validation. Further assessment of social validation involving mixed qualitative and quantitative methods are discussed in the conclusion. [ABSTRACT FROM AUTHOR]

Prophylactic interventions such as vaccine allocation are some of the most effective public health policy planning tools. The supply of vaccines, however, is limited and an important challenge is to optimally allocate the vaccines to minimize epidemic impact. This resource allocation question (which we refer to as VID) has multiple dimensions: when, where, to whom, etc. Most of the existing literature in this topic deals with the latter (to whom), proposing policies that prioritize individuals by age and disease risk. However, since seasonal influenza spread has a typical spatial trend, and due to the temporal constraints enforced by the availability schedule, the when and where problems become equally, if not more, relevant. In this paper, we study the VID problem in the context of seasonal influenza spread in the United States. We develop a national scale metapopulation model for influenza that integrates both short and long distance human mobility, along with realistic data on vaccine uptake. We also design GA, a greedy algorithm for allocating the vaccine supply at the state level under temporal constraints and show that such a strategy improves over the current baseline of pro-rata allocation, and the improvement is more pronounced for higher vaccine efficacy and moderate flu season intensity. Further, the resulting strategy resembles a ring vaccination applied spatiallyacross the US. [ABSTRACT FROM AUTHOR]

Background: Policies designed to restrict marketing, access to, and public use of electronic cigarettes (e-cigarettes) are increasingly under debate in various jurisdictions in the US. Little is known about public perceptions of these policies and factors that predict their support or opposition. Methods: Using a sample of US adults from Amazon Mechanical Turk in May 2015, this paper identifies beliefs about the benefits and costs of regulating e-cigarettes and identifies which of these beliefs predict support for e-cigarette restricting policies. Results: A higher proportion of respondents agreed with 8 different reasons to regulate e-cigarettes (48.5% to 83.3% agreement) versus 7 reasons not to regulate e-cigarettes (11.5% to 18.9%). The majority of participants agreed with 7 out of 8 reasons for regulation. When all reasons to regulate or not were included in a final multivariable model, beliefs about protecting people from secondhand vapor and protecting youth from trying e-cigarettes significantly predicted stronger support for e-cigarette restricting policies, whereas concern about government intrusion into individual choices was associated with reduced support. Discussion: This research identifies key beliefs that may underlie public support or opposition to policies designed to regulate the marketing and use of e-cigarettes. Advocates on both sides of the issue may find this research valuable in developing strategic campaigns related to the issue. Implications: Specific beliefs of potential benefits and costs of e-cigarette regulation (protecting youth, preventing exposure to secondhand vapor, and government intrusion into individual choices) may be effectively deployed by policy makers or health advocates in communicating with the public. [ABSTRACT FROM AUTHOR]

The goal of this paper is to estimate peer influence in video gaming time among adolescents. Using a nationally representative sample of the U.S. school-aged adolescents in 2009–2010, we estimate a structural model that accounts for the potential biases in the estimate of the peer effect. Our peer group is exogenously assigned and includes one year older adolescents in the same school grade as the respondent. The peer measure is based on peers’ own reports of video gaming time. We find that an additional one hour of playing video games per week by older grade-mates results in .47 hours increase in video gaming time by male responders. We do not find significant peer effect among female responders. Effective policies aimed at influencing the time that adolescents spend video gaming should take these findings into account. [ABSTRACT FROM AUTHOR]

The spatial distribution of income shapes the structure and organisation of cities and its understanding has broad societal implications. Despite an abundant literature, many issues remain unclear. In particular, all definitions of segregation are implicitely tied to a single indicator, usually rely on an ambiguous definition of income classes, without any consensus on how to define neighbourhoods and to deal with the polycentric organization of large cities. In this paper, we address all these questions within a unique conceptual framework. We avoid the challenge of providing a direct definition of segregation and instead start from a definition of what segregation is not. This naturally leads to the measure of representation that is able to identify locations where categories are over- or underrepresented. From there, we provide a new measure of exposure that discriminates between situations where categories co-locate or repel one another. We then use this feature to provide an unambiguous, parameter-free method to find meaningful breaks in the income distribution, thus defining classes. Applied to the 2014 American Community Survey, we find 3 emerging classes—low, middle and higher income—out of the original 16 income categories. The higher-income households are proportionally more present in larger cities, while lower-income households are not, invalidating the idea of an increased social polarisation. Finally, using the density—and not the distance to a center which is meaningless in polycentric cities—we find that the richer class is overrepresented in high density zones, especially for larger cities. This suggests that density is a relevant factor for understanding the income structure of cities and might explain some of the differences observed between US and European cities. [ABSTRACT FROM AUTHOR]

We aim to determine whether a game-theoretic model between an insurer and a healthcare practice yields a predictive equilibrium that incentivizes either player to deviate from a fee-for-service to capitation payment system. Using United States data from various primary care surveys, we find that non-extreme equilibria (i.e., shares of patients, or shares of patient visits, seen under a fee-for-service payment system) can be derived from a Stackelberg game if insurers award a non-linear bonus to practices based on performance. Overall, both insurers and practices can be incentivized to embrace capitation payments somewhat, but potentially at the expense of practice performance. [ABSTRACT FROM AUTHOR]

Objective: The Centers for Disease Control and Prevention (CDC) coordinates a labor-intensive process to measure the prevalence of autism spectrum disorder (ASD) among children in the United States. Random forests methods have shown promise in speeding up this process, but they lag behind human classification accuracy by about 5%. We explore whether more recently available document classification algorithms can close this gap. Materials and methods: Using data gathered from a single surveillance site, we applied 8 supervised learning algorithms to predict whether children meet the case definition for ASD based solely on the words in their evaluations. We compared the algorithms’ performance across 10 random train-test splits of the data, using classification accuracy, F1 score, and number of positive calls to evaluate their potential use for surveillance. Results: Across the 10 train-test cycles, the random forest and support vector machine with Naive Bayes features (NB-SVM) each achieved slightly more than 87% mean accuracy. The NB-SVM produced significantly more false negatives than false positives (P = 0.027), but the random forest did not, making its prevalence estimates very close to the true prevalence in the data. The best-performing neural network performed similarly to the random forest on both measures. Discussion: The random forest performed as well as more recently available models like the NB-SVM and the neural network, and it also produced good prevalence estimates. NB-SVM may not be a good candidate for use in a fully-automated surveillance workflow due to increased false negatives. More sophisticated algorithms, like hierarchical convolutional neural networks, may not be feasible to train due to characteristics of the data. Current algorithms might perform better if the data are abstracted and processed differently and if they take into account information about the children in addition to their evaluations. Conclusion: Deep learning models performed similarly to traditional machine learning methods at predicting the clinician-assigned case status for CDC’s autism surveillance system. While deep learning methods had limited benefit in this task, they may have applications in other surveillance systems. [ABSTRACT FROM AUTHOR]

Not much is known about the fleet level total seabird bycatch from pelagic longlines of United States vessels in the western North Atlantic or other fleets of the Atlantic or other oceans. Onboard observers generally only record seabird bycatch during line hauling. Seabirds are predominantly caught during the line setting stage, and, due to predation or mechanical action, those caught prior to the haul can drop off the hook and be lost to the onboard observer. We developed a model to gauge the size of this bycatch loss problem and provide a first approximation of its impact on estimates of total fleet bycatch. We started with a traditional loss-free bycatch model, which assumes that birds recorded were the only birds captured, and integrated into it two crucial components of the bycatch process: capture origin (set or haul) and bycatch loss of set-captures. We extracted count data on seabird bycatch loss and bycatch mortality from the literature on other longline fisheries and used these data to simulate potential total seabird bycatch in the western North Atlantic. Simulations revealed the shortcomings of both the traditional bycatch model and the current haul-only observer protocol, each of which contributed to biologically significant underestimation of total bycatch and estimation uncertainty. Based on our results, we recommend a loss-corrected modeling approach to provide a more accurate estimate of seabird mortalities in pelagic longline fisheries. Where possible, fishery-specific seabird bycatch loss rates need to be ascertained via specific set and haul observing protocols. But, even where fishery-specific estimates for a region are not available, the methodology developed here is applicable to other pelagic longline fisheries to approximate fleet-level loss-corrected bycatch. [ABSTRACT FROM AUTHOR]

Downscaled climate projections need to be linked to downscaled projections of population and economic growth to fully develop implications for land, natural resources, and ecosystems for future scenarios. We develop an empirical spatiotemporal approach for jointly projecting population and income at the county scale in the United States that is consistent with neoclassical economic growth theory and overlapping labor markets and that accounts for labor migration and spatial spillovers. Downscaled projections generated for the five Shared Socioeconomic Pathways used to support global scenario analysis generally show growth focused around relatively few centers especially in the southeast and western regions, with some areas in the Midwest and northeast experiencing population declines. Results are consistent with economic growth theory and with historical trends in population change and convergence of per capita personal income across US counties. [ABSTRACT FROM AUTHOR]

Objectives: Reviews of violence against children in disaster settings focus on armed conflict. Little is understood about natural disasters which has implications in planning humanitarian response. We examined the magnitude and direction of the association between exposure to natural disasters and physical, emotional, and sexual violence against children, and assessed the quality of the evidence. Methods: We searched 15 health and social science databases from first record until May 16, 2018. Publications describing all types of quantitative study design were eligible for inclusion. We presented study characteristics and quality in a narrative form and generated pooled estimates using a three-level random effects model. We evaluated Cochrane’s Q with p-values below 0.10 and radial plots to assess heterogeneity. Planned subgroup analyses explored differential results by violence form, study design, and analysis method. Results: 11 publications met inclusion criteria. The majority were cross-sectional studies examining physical or sexual violence in the United States. We found no evidence of a consistent association or directional influence between natural disasters and violence against children. Combined categorical violence outcomes had substantial heterogeneity [Q (df = 66) = 252.83, p < 0.001]. Subgroups without evidence of heterogeneity had confidence intervals that included a possible null effect. Our findings were mainly limited by inconsistencies in operational definitions of violence, a lack of representative sampling, and unclear establishment of temporal order between natural disaster exposure and violence outcomes. Conclusions: Based on the available evidence, we cannot confidently conclude that natural disasters increase the level or severity of violence against children above non-disaster settings, however heterogeneity and study quality hamper our ability to draw firm conclusions. More nuanced and rigorous research is needed to inform practice and policy as natural disasters increasingly affect human populations. [ABSTRACT FROM AUTHOR]

Most of future population growth will take place in the world’s cities and towns. Yet, there is no well-established, consistent way to measure either urban land or people. Even census-based urban concepts and measures undergo frequent revision, impeding rigorous comparisons over time and place. This study presents a new spatial approach to derive consistent urban proxies for the US. It compares census-designated urban blocks with proxies for land-based classifications of built-up areas derived from time-series of the Global Human Settlement Layer (GHSL) for 1990–2010. This comparison provides a new way to understand urban structure and its changes: Most land that is more than 50% built-up, and people living on such land, are officially classified as urban. However, 30% of the census-designated urban population and land is located in less built-up areas that can be characterized as mainly suburban and peri-urban in nature. Such insights are important starting points for a new urban research program: creating globally and temporally consistent proxies to guide modelling of urban change. [ABSTRACT FROM AUTHOR]

Breweries across the country are investing in energy efficient and low-carbon brewing practices. Drawing insights from the sustainable consumption and ecological economics literature, this analysis evaluates whether consumers are willing to pay more for sustainable beer and what predicts the value of the premium. Based on a survey of beer consumers from across the U.S. that contained one of two willingness-to-pay exercises, we evaluate what respondent attributes are associated with a higher willingness-to-pay for sustainably brewed beer. We find that the majority of beer consumers are willing to pay more for sustainable beer. Consumers who are prepared to pay a premium tend to already pay more per unit of beer, are more aware of their purchasing behavior and the manner in which their consumption patterns may affect the environment, and pursue lifestyles based on professional advancement, helping the environment, and helping other causes. [ABSTRACT FROM AUTHOR]

Accurate demand forecasting for new technology products is a key factor in the success of a business. We propose a way to forecasting a new product’s diffusion through technology diffusion and interest diffusion. Technology diffusion and interest diffusion are measured by the volume of patent citations and web search traffic, respectively. We apply the proposed method to forecast the sales of hybrid cars and industrial robots in the US market. The results show that that technology diffusion, as represented by patent citations, can explain long-term sales for hybrid cars and industrial robots. On the other hand, interest diffusion, as represented by web search traffic, can help to improve the predictability of market sales of hybrid cars in the short-term. However, interest diffusion is difficult to explain the sales of industrial robots due to the different market characteristics. Finding indicates our proposed model can relatively well explain the diffusion of consumer goods. [ABSTRACT FROM AUTHOR]

We investigate the causal uncertainty surrounding the flash crash in the U.S. Treasury bond market on October 15, 2014, and the unresolved concern that no clear link has been identified between the start of the flash crash at 9:33 and the opening of the U.S. equity market at 9:30. We consider the contributory effect of mini flash crashes in equity markets, and find that the number of equity mini flash crashes in the three-minute window between market open and the Treasury Flash Crash was 2.6 times larger than the number experienced in any other three-minute window in the prior ten weekdays. We argue that (a) this statistically significant finding suggests that mini flash crashes in equity markets both predicted and contributed to the October 2014 U.S. Treasury Bond Flash Crash, and (b) mini-flash crashes are important phenomena with negative externalities that deserve much greater scholarly attention. [ABSTRACT FROM AUTHOR]

Both obesity (body mass index ≥ 30) and educational attainment have increased dramatically in the United States since the 1970s. This study analyzed the influences of educational inequalities in obesity and population improvements in education on national obesity trends between 1970 and 2010. For non-Hispanic white and black males and females aged 25–74 years, educational differences in the probability of being obese were estimated from the 1971–2012 National Health and Nutrition Examination Surveys, and population distributions of age and educational groups, from the 1970 Census and 2010 American Community Survey. In the total population, obesity increased from 15.7% to 38.8%, and there were increases in the greater obese probabilities of non-college graduates relative to four-year college graduates. The increase in obesity would have been lower by 10% (2.2 percentage points) if educational inequalities in obesity had stayed at their 1970 values and lower by one third (7.9 points) if obesity inequalities had been eliminated. Obesity inequalities were larger for females than males and for whites than blacks, and obesity did not differ by education among black males. As a result, the impact of obesity inequalities on the obesity trend was largest among white females (a 47% reduction in the obesity increase if obesity inequalities had been eliminated), and virtually zero among black males. On the other hand, without educational improvements, the obesity increase would have been 9% more in the total population, 23% more among white females and not different in the other three subpopulations. Results indicate that obesity inequalities made sizable contributions to the obesity trends, and the obesity reductions associated with educational improvements were more limited. [ABSTRACT FROM AUTHOR]

Time spent fishing is the effort metric often studied in fisheries but it may under-represent the effort actually expended by fishers. Entire fishing trips, from the time vessels leave port until they return, may prove more useful for examining trends in fleet dynamics, fisher behavior, and fishing costs. However, such trip information is often difficult to resolve. We identified ~30,000 trips made by vessels that targeted walleye pollock (Gadus chalcogrammus) in the Eastern Bering Sea from 2008–2014 by using vessel monitoring system (VMS) and landings data. We compared estimated trip durations to observer data, which were available for approximately half of trips. Total days at sea were estimated with < 1.5% error and 96.4% of trip durations were either estimated with < 5% error or they were within expected measurement error. With 99% accuracy, we classified trips as fishing for pollock, for another target species, or not fishing. This accuracy lends strong support to the use of our method with unobserved trips across North Pacific fisheries. With individual trips resolved, we examined potential errors in datasets which are often viewed as “the truth.” Despite having > 5 million VMS records (timestamps and vessel locations), this study was as much about understanding and managing data errors as it was about characterizing trips. Missing VMS records were pervasive and they strongly influenced our approach. To understand implications of missing data on inference, we simulated removal of VMS records from trips. Removal of records straightened (i.e., shortened) vessel trajectories, and travel distances were underestimated, on average, by 1.5–13.4% per trip. Despite this bias, VMS proved robust for trip characterization and for improved quality control of human-recorded data. Our scrutiny of human-reported and VMS data advanced our understanding of the potential utility and challenges facing VMS users globally. [ABSTRACT FROM AUTHOR]

Background: Disease-associated malnutrition has been identified as a prevalent condition, particularly for the elderly, which has often been overlooked in the U.S. healthcare system. The state-level burden of community-based disease-associated malnutrition is unknown and there have been limited efforts by state policy makers to identify, quantify, and address malnutrition. The objective of this study was to examine and quantify the state-level economic burden of disease-associated malnutrition. Methods: Direct medical costs of disease-associated malnutrition were calculated for 8 diseases: Stroke, Chronic Obstructive Pulmonary Disease, Coronary Heart Failure, Breast Cancer, Dementia, Musculoskeletal Disorders, Depression, and Colorectal Cancer. National disease and malnutrition prevalence rates were estimated for subgroups defined by age, race, and sex using the National Health and Nutrition Examination Survey and the National Health Interview Survey. State prevalence of disease-associated malnutrition was estimated by combining national prevalence estimates with states’ demographic data from the U.S. Census. Direct medical cost for each state was estimated as the increased expenditures incurred as a result of malnutrition. Principal Findings: Direct medical costs attributable to disease-associated malnutrition vary among states from an annual cost of $36 per capita in Utah to $65 per capita in Washington, D.C. Nationally the annual cost of disease-associated malnutrition is over $15.5 billion. The elderly bear a disproportionate share of this cost on both the state and national level. Conclusions: Additional action is needed to reduce the economic impact of disease-associated malnutrition, particularly at the state level. Nutrition may be a cost-effective way to help address high health care costs. [ABSTRACT FROM AUTHOR]

The rapid increase of wealth inequality in the past few decades is one of the most disturbing social and economic issues of our time. Studying its origin and underlying mechanisms is essential for policy aiming to control and even reverse this trend. In that context, controlling the distribution of income, using income tax or other macroeconomic policy instruments, is generally perceived as effective for regulating the wealth distribution. We provide a theoretical tool, based on the realistic modeling of wealth inequality dynamics, to describe the effects of personal savings and income distribution on wealth inequality. Our theoretical approach incorporates coupled equations, solved using iterated maps to model the dynamics of wealth and income inequality. Notably, using the appropriate historical parameter values we were able to capture the historical dynamics of wealth inequality in the United States during the course of the 20th century. It is found that the effect of personal savings on wealth inequality is substantial, and its major decrease in the past 30 years can be associated with the current wealth inequality surge. In addition, the effect of increasing income tax, though naturally contributing to lowering income inequality, might contribute to a mild increase in wealth inequality and vice versa. Plausible changes in income tax are found to have an insignificant effect on wealth inequality, in practice. In addition, controlling the income inequality, by progressive taxation, for example, is found to have a very small effect on wealth inequality in the short run. The results imply, therefore, that controlling income inequality is an impractical tool for regulating wealth inequality. [ABSTRACT FROM AUTHOR]

Objectives: Assessment and identification of spatial structures in the distribution and abundance of invasive species is important for unraveling the underlying ecological processes. The invasive agricultural insect pest Halyomorpha halys that causes severe economic losses in the United States is currently expanding both within United States and across Europe. We examined the drivers of H. halys invasion by characterizing the distribution and abundance patterns of H. halys and native stink bugs (Chinavia hilaris and Euschistus servus) across eight different spatial scales. We then quantified the interactive and individual influences of temperature, and measures of resource availability and distance from source populations, and their relevant spatial scales. We used Moran’s Eigenvector Maps based on Gabriel graph framework to quantify spatial relationships among the soybean fields in mid-Atlantic Unites States surveyed for stink bugs. Findings: Results from the multi-spatial scale, multivariate analyses showed that temperature and its interaction with resource availability and distance from source populations structures the patterns in H. halys at very broad spatial scale. H. halys abundance decreased with increasing average June temperature and distance from source population. H. halys were not recorded at fields with average June temperature higher than 23.5°C. In parts with suitable climate, high H. halys abundance was positively associated with percentage developed open area and percentage deciduous forests at 250m scale. Broad scale patterns in native stink bugs were positively associated with increasing forest cover and, in contrast to the invasive H. halys, increasing mean July temperature. Our results identify the contrasting role of temperature in structuring regional patterns in H. halys and native stink bugs, while demonstrating its interaction with resource availability and distance from source populations for structuring H. halys patterns. Conclusion: These results help predicting the pest potential of H. halys and vulnerability of agricultural systems at various regions, given the climatic conditions, and its interaction with resource availability and distance from source populations. Monitoring and control efforts within parts of the United States and Europe with more suitable climate could focus in areas of peri-urban developments with deciduous forests and other host plants, along with efforts to reduce propagule pressure. [ABSTRACT FROM AUTHOR]

Despite increased awareness and concern about children with developmental disabilities wandering away from adult supervision, there is a paucity of research about elopement. This is the first study to examine and report the prevalence and correlates of elopement in a nationally representative sample of school-age children in the United States with an autism spectrum disorder (ASD) and/or cognitive impairment. Data were obtained from the CDC's "Pathways" Survey, a follow-up telephone survey of the parents of 4,032 children with a developmental condition. 3,518 children that had ASD, intellectual disability (ID), and/or developmental delay (DD) at the time of survey administration were included for analysis. Children were divided into three condition groups: ASD-only; ID/DD-only; ASD+ID/DD. Logistic regression analyses were used to compare the prevalence of elopement and rates of preventive measure use (barriers and/or electronic devices) across condition groups, and to examine the clinical and demographic correlates of elopement. T-tests were also performed to compare scores on the Children's Social Behavior Questionnaire (CSBQ) between wanderers and non-wanderers. Overall, 26.7% of children had reportedly eloped within the previous year, most commonly from public places. Children with ASD-only and ASD+ID/DD were more likely to have eloped than those with ID/DD-only. Across all groups, wanderers scored higher than non-wanderers on five out of six CSBQ subscales; they were more likely not to realize when there is danger, to have difficulty distinguishing between strangers and familiar people, to show sudden mood changes, to over-react to everything/everyone, to get angry quickly, to get lost easily, and to panic in new situations or if change occurs. Even after controlling for elopement history, parents of children in the ASD+ID/DD group were more likely than those in the other condition groups to report using physical or electronic measures to prevent wandering. [ABSTRACT FROM AUTHOR]

Background: Several studies have focused on stratifying patients according to their level of readmission risk, fueled in part by incentive programs in the U.S. that link readmission rates to the annual payment update by Medicare. Patient-specific predictions about readmission have not seen widespread use because of their limited accuracy and questions about the efficacy of using measures of risk to guide clinical decisions. We construct a predictive model for readmissions for congestive heart failure (CHF) and study how its predictions can be used to perform patient-specific interventions. We assess the cost-effectiveness of a methodology that combines prediction and decision making to allocate interventions. The results highlight the importance of combining predictions with decision analysis. Methods: We construct a statistical classifier from a retrospective database of 793 hospital visits for heart failure that predicts the likelihood that patients will be rehospitalized within 30 days of discharge. We introduce a decision analysis that uses the predictions to guide decisions about post-discharge interventions. We perform a cost-effectiveness analysis of 379 additional hospital visits that were not included in either the formulation of the classifiers or the decision analysis. We report the performance of the methodology and show the overall expected value of employing a real-time decision system. Findings: For the cohort studied, readmissions are associated with a mean cost of $13,679 with a standard error of $1,214. Given a post-discharge plan that costs $1,300 and that reduces 30-day rehospitalizations by 35%, use of the proposed methods would provide an 18.2% reduction in rehospitalizations and save 3.8% of costs. Conclusions: Classifiers learned automatically from patient data can be joined with decision analysis to guide the allocation of post-discharge support to CHF patients. Such analyses are especially valuable in the common situation where it is not economically feasible to provide programs to all patients. [ABSTRACT FROM AUTHOR]

Objective: Reported associations between socioeconomic status (SES) and obesity are inconsistent depending on gender and geographic location. Globally, these inconsistent observations may hide a variation in the contextual effect on individuals' risk of obesity for subgroups of the population. This study explored the regional variability in the association between SES and BMI in the USA and in Canada, and describes the geographical variance patterns by SES category. Methods: The 2009–2010 samples of the Behavioral Risk Factor Surveillance System (BRFSS) and the Canadian Community Health Survey (CCHS) were used for this comparison study. Three-level random intercept and differential variance multilevel models were built separately for women and men to assess region-specific BMI by SES category and their variance bounds. Results: Associations between individual SES and BMI differed importantly by gender and countries. At the regional-level, the mean BMI variation was significantly different between SES categories in the USA, but not in Canada. In the USA, whereas the county-specific mean BMI of higher SES individuals remained close to the mean, its variation grown as SES decreased. At the county level, variation of mean BMI around the regional mean was 5 kg/m2 in the high SES group, and reached 8.8 kg/m2 in the low SES group. Conclusions: This study underlines how BMI varies by country, region, gender and SES. Lower socioeconomic groups within some regions show a much higher variation in BMI than in other regions. Above the BMI regional mean, important variation patterns of BMI by SES and place of residence were found in the USA. No such pattern was found in Canada. This study suggests that a change in the mean does not necessarily reflect the change in the variance. Analyzing the variance by SES may be a good way to detect subtle influences of social forces underlying social inequalities. [ABSTRACT FROM AUTHOR]