Between-state Variation Research Articles

Multilevel analysis of determinants in postnatal care utilisation among mother-newborn pairs in India, 2019-21.

Postnatal care (PNC) utilisation within 24 hours of delivery is a critical component of health care services for mothers and newborns. While substantial geographic variations in various health outcomes have been documented in India, there remains a lack of understanding regarding PNC utilisation and underlying factors accounting for these geographic variations. In this study, we aimed to partition and explain the variation in PNC utilisation across multiple geographic levels in India. Using India's 5th National Family Health Survey (2019-21), we conducted four-level logistic regression analyses to partition the total geographic variation in PNC utilisation by state, district, and cluster levels, and to quantify how much of theses variations are explained by a set of 12 demographic, socioeconomic, and pregnancy-related factors. We also conducted analyses stratified by selected states/union territories. Among 149 622 mother-newborn pairs, 82.29% of mothers and 84.92% of newborns were reported to have received PNC within 24 hours of delivery. In the null model, more than half (56.64%) of the total geographic variation in mother's PNC utilisation was attributed to clusters, followed by 26.06% to states/union territories, and 17.30% to districts. Almost 30% of the between-state variation in mother's PNC utilisation was explained by the demographic, socioeconomic, and pregnancy-related factors (i.e. state level variance reduced from 0.486 (95% confidence interval (CI) = 0.238, 0.735) to 0.320 (95% CI = 0.152, 0.488)). We observed consistent results for newborn's PNC utilisation. State-specific analyses showed substantial geographic variation attributed to clusters across all selected states/union territories. Our findings highlight the consistently large cluster variation in PNC utilisation that remains unexplained by compositional effects. Future studies should explore contextual drivers of cluster variation in PNC utilisation to inform and design interventions aimed to improve maternal and child health.

Abstract 340: Automated External Defibrillator Utilization After Out-of-Hospital Cardiac Arrest at Recreational Facilities in the United States

Background: Given the importance of public access defibrillation, some U.S. states have passed laws requiring an automatic external defibrillator (AED) to be available on-site at athletic facilities. However, rates of bystander AED application during out-of-hospital cardiac arrest (OHCA) in states with and without such laws remain unknown. Research Question: What are the rates of bystander AED application during OHCA at recreational facilities in U.S. states that mandate AEDs availability at such locations? Aim: To examine the rates of bystander AED application for OHCA at recreational facilities in states with and without laws mandating the presence of AEDs at athletic facilities. Methods: We included adults (≥18 years) in the Cardiac Arrest Registry to Enhance Survival (CARES) registry with a non-traumatic OHCA during 2013-2021 at a recreational facility in the U.S. The primary outcome was bystander AED application during OHCA. Secondary outcomes included survival to hospital admission and survival to hospital discharge. We examined the rates of bystander AED application and survival outcomes in states with laws that mandate AEDs at athletic facilities and in states without such laws. Results: Overall, there were 4,145 OHCAs cases at recreational facilities in 13 states that mandated AEDs at athletic facilities and 5,145 cases in 27 states that did not. In states with laws, the median rate of bystander AED application was 19.0%, with large between-state variation (IQR: 15.1%-22.0%). The median rate of survival to hospital admission was 44.5% (IQR 39.4%-56.9%) and survival to hospital discharge was 31.0% (IQR 25.2%-32.8%). In non-law states, the median rate of bystander AED application was 18.2% (IQR: 13.9%-25.0%). The median rate of survival to hospital admission was 45.0% (IQR 38.4%-52.1%) and survival to hospital discharge was 28.4% (IQR 25.9%-37.5%). Conclusions: Despite state laws that mandate on-site availability of an AED, rates of bystander AED application for OHCA at recreational facilities in such states remain low, with notable state-level variability. Our findings highlight that improving bystander AED application during OHCA will need to go beyond making AEDs available on-site.

Interpreting Between-State Variation in 988 Suicide and Crisis Lifeline Call Volume Rates.

Interpreting Between-State Variation in 988 Suicide and Crisis Lifeline Call Volume Rates.

Predicting initiation to treatment for youth on probation: A multi-level approach

BackgroundPrevious research on youth involved in the juvenile legal system (JLS), who have higher substance use prevalence than other adolescents, has focused on treatment outcomes post-initiation. Scant research has examined factors affecting treatment initiation following JLS referral. The current work addresses this gap by estimating net impacts of state, site, and client attributes on initiating treatment. MethodsMixed effects logistic regression models analyzed data from the Behavioral Health Services Cascade framework of the JJ-TRIALS project for 5325 youth records from 22 sites across six states, using weighted hot deck imputation to address missing data. The analysis permitted estimating net impacts of individual factors while controlling for state variation factors and separating out within-state site variation. ResultsRates of treatment initiation varied significantly across sites. About 63% of outcome variation arose from site and state disparities. The odds of initiating treatment were 127% higher for youth under a higher supervision level compared to youth under low supervision. Race and ethnicity were significantly associated with initiation, with Black and Hispanic youth having 44% and 42% lower odds of initiating treatment, respectively. ConclusionIndividual factors, including race, ethnicity, and level of supervision, significantly differed in treatment initiation, controlling for between-state variation and allowing between-site-within-state effects to operate separately. A deeper understanding of factors shaping initiation may help behavioral health and JLS agencies to develop strategies to increase initiation for referred youth. This is the first study delineating individual, cross-program, and between-state variation in treatment initiation among referred youth.

Screening for Social Risk at Federally Qualified Health Centers: A National Study

Federally Qualified Health Centers serve 29.8 million low-income patients across the U.S., many of whom have unaddressed social risks. In 2019, for the first time, data on social risk screening capabilities were collected from every U.S. Federally Qualified Health Center. The objectives of this study were to describe the national rates of social risk screening capabilities across Federally Qualified Health Centers, identify organizational predictors of screening, and assess between-state heterogeneity. Using a 100% sample of U.S. Federally Qualified Health Centers (N=1,384, representing 29.8 million patients) from the 2019 Uniform Data System, the primary outcome was whether a Federally Qualified Health Center collected data on patients' social risk factors (yes/no). Summary statistics on the rates of social risk screening capabilities were generated in aggregate and by state. Linear probability models were then used to estimate the relationship between the probability of social risk screening and 7 key Federally Qualified Health Center characteristics (e.g., Federally Qualified Health Center size, Medicaid MCO contract, Medicaid accountable care organization presence). Data were analyzed in 2020‒2021. Most (71%) Federally Qualified Health Centers collected social risk data, with a between-state variation. The most common screener was the Protocol for Responding to and Assessing Patients' Assets Risks and Experiences (43% of Federally Qualified Health Centers that screened), whereas 22% collected social risk data using a nonstandardized screener. After adjusting for other characteristics, Federally Qualified Health Centers with social risk screening capabilities served more total patients, were more likely to be located in a state with a Medicaid accountable care organization, and were less likely to have an MCO contract. There has been widespread adoption of social risk screening tools across U.S. Federally Qualified Health Centers, but between-state disparities exist. Targeting social risk screening resources to smaller Federally Qualified Health Centers may increase the adoption of screening tools.

Geographic Variation in the Use of Gabapentinoids and Opioids for Pain in a Commercially Insured Adult Population in the United States.

BackgroundGabapentinoids have been prescribed off-label for almost all types of pain. The geographic variation in the use of gabapentinoids as analgesics remains unknown.ObjectiveTo describe the geographic variation in gabapentinoids, opioids and concurrent use of both for pain by US state and metropolitan statistical area (MSA).MethodsWe conducted a cross-sectional study on December 1, 2018, among commercially insured adults aged 18–64 years without epilepsy or opioid use disorders using IBM® MarketScan® Research Databases. We described the geographic variation in the analgesic regimens (gabapentinoids, opioids and concurrent use of both) by state and MSA, and assessed factors associated with the geographic variation using multilevel logistic regression.ResultsWe included 9,314,197 beneficiaries; 1.4% had gabapentinoids, 1.5% had opioids and 0.3% had concurrent use of both. The majority of gabapentinoid use lacked an FDA-approved indication. Use of the analgesic regimens varied across states (gabapentinoids (median (interquartile range)): 1.4% (1.2–1.7%); opioids: 1.5% (1.2–1.9%); both: 0.3% (0.2–0.4%)) and MSAs (gabapentinoids: 1.6% (1.3–2.0%); opioids: 1.8% (1.3–2.3%); both: 0.3% (0.2–0.5%)). Demographics explained the largest proportion of the between-state and between-MSA variation. The pattern of the geographic variation in gabapentinoids was similar to that of opioids across states and MSAs.ConclusionGabapentinoids were as commonly used as opioids for pain in a commercially insured population (mostly off-label). The geographic variation in gabapentinoids was similar to that of opioids, which suggests that gabapentinoids may be widely used as alternatives or adjuvants to opioids across the US.

ENERGY STAR Appliance Market Shares: Do They Respond to Electricity Prices, and Does It Matter?

We address an apparent paradox that the market shares of four ENERGY STAR appliances in the United States do not respond to within-state changes in electricity prices. We resolve the paradox by showing that market shares do respond to between-state variation in electricity prices. We also suggest an economic explanation for the paradox through the timing of appliance purchases. Using the estimation results, we find that the four ENERGY STAR appliances reduce carbon emissions by 1.9 million megawatt-hours per year, equivalent to removing 0.1% of all U.S. vehicles, and that the addition of a $100/ton CO2 price would increase these figures to 2.1 million megawatt-hours and 0.11%.

Multilevel regression and poststratification for estimating population quantities from large health studies: a simulation study based on US population structure

IntroductionRecruiting a representative sample of participants is becoming increasingly difficult in large-scale health surveys. Multilevel regression and poststratification (MRP) has been shown to be effective in estimating population descriptive quantities...

Quantifying and explaining variation in life expectancy at census tract, county, and state levels in the United States

Studies on geographic inequalities in life expectancy in the United States have exclusively focused on single-level analyses of aggregated data at state or county level. This study develops a multilevel perspective to understanding variation in life expectancy by simultaneously modeling the geographic variation at the levels of census tracts (CTs), counties, and states. We analyzed data from 65,662 CTs, nested within 3,020 counties and 48 states (plus District of Columbia). The dependent variable was age-specific life expectancy observed in each of the CTs. We also considered the following CT-level socioeconomic and demographic characteristics as independent variables: population density; proportions of population who are black, who are single parents, who are below the federal poverty line, and who are aged 25 or older who have a bachelor's degree or higher; and median household income. Of the total geographic variation in life expectancy at birth, 70.4% of the variation was attributed to CTs, followed by 19.0% for states and 10.7% for counties. The relative importance of CTs was greater for life expectancy at older ages (70.4 to 96.8%). The CT-level independent variables explained 5 to 76.6% of between-state variation, 11.1 to 58.6% of between-county variation, and 0.7 to 44.9% of between-CT variation in life expectancy across different age groups. Our findings indicate that population inequalities in longevity in the United States are primarily a local phenomenon. There is a need for greater precision and targeting of local geographies in public policy discourse aimed at reducing health inequalities in the United States.

State variation in the characteristics of Medicare-Medicaid dual enrollees: Implications for risk adjustment.

To examine between-state differences in the socioeconomic and health characteristics of Medicare beneficiaries dually enrolled in Medicaid, focusing on characteristics not observable to or used by policy makers for risk adjustment. 2010-2013 Medicare Current Beneficiary Survey. Retrospective analyses of survey-reported health and socioeconomic status (SES) measures among low-income Medicare beneficiaries and low-income dual enrollees. We used hierarchical linear regression models with state random effects to estimate the between-state variation in respondent characteristics and linear models to compare the characteristics of dual enrollees by state Medicaid policies. Between-state differences in health and socioeconomic risk among low-income Medicare beneficiaries, as measured by the coefficient of variation, ranged from 17.5 percent for an index of socioeconomic risk to 20.3 percent for an index of health risk. Between-state differences were comparable among the subset of low-income beneficiaries dually enrolled in Medicare and Medicaid. Dual enrollees with incomes below the Federal Poverty Level were in better health and had higher SES in states that offered Medicaid to individuals with relatively higher incomes. Duals' average incomes were higher in states with Medically Needy programs. Characteristics of dual enrollees differ substantially across states, reflecting differences in states' low-income Medicare populations and Medicaid policies. Risk-adjustment methods using dual enrollment to proxy for poor health and low SES should account for this state-level heterogeneity.

Evidence that social-economic factors play an important role in drug overdose deaths

BackgroundDrug overdose deaths in the United States increased from approximately 16,000 per year in 2001 to 41,000 per year in 2014. Although every US state witnessed an increase, the increases were much larger in some states than others. There was also variation as a function of race and ethnicity. Non-Hispanic Whites accounted for more than 80% of the deaths, and in some states their rates were about ten times greater per capita than Hispanic and Non-White rates. State and temporal differences provide an opportunity to evaluate explanations of what is driving drug overdose deaths. In this report, we evaluate the degree to which state level variation in opioid prescription rates and social-economic conditions explain state level variation in overdose death rates. Methods and dataWe used publicly available data from the Center for Disease Control (CDC), Bureau of Labor Statistics (BLS), Drug Enforcement Agency (DEA) and the Opportunity Insights project. ResultsLegally prescribed opioids, social capital and work force participation accounted for 53–69% of the between-state variation in overdose deaths in Non-Hispanic Whites. Prescriptions and the two social economic measures accounted for about the same amounts of unique variation, but shared variation among the three independent variables was the strongest predictor of overdose deaths. Panel regression results of the year-to-year changes in overdose deaths were similar. However, the pattern of correlations for Hispanics and Non-Whites was quite different. Neither opioid prescriptions nor social capital were significant predictors of overdose deaths in the between-state and between-year Hispanic and Non-White regression analyses. ConclusionsCommon variation in opioid prescriptions rates, social capital, and work force participation proved the strongest predictor of drug overdose deaths in Non-Hispanic Whites. We discuss reasons why the same did not hold for the Hispanic/Non-White population.

Spatial variation in teens’ crash rate reduction following the implementation of a graduated driver licensing program in Michigan

Motor vehicle crashes are a leading cause of injury, and teen drivers contribute disproportionately to that burden. Graduated Driver Licensing (GDL) programs are effective at reducing teen crash risk, but teen crash rates remain high. Between-state variation in the teen crash rate reduction following GDL implementation has been documented, but this is the first study to examine small-area variation in such a reduction. Fusing together crash data from the Michigan State Police, census data, and organizational data (alcohol outlet, movie theatre, and school locations), we analyzed spatial correlates of teen injury crash, and place-based features that modified the injury crash rate difference following GDL implementation. Specifically, using census-based units, we estimated changes in injury crash rates among teens using negative binomial regression controlling for spatial autocorrelation, and tested whether any measured spatial characteristics modified the crash rate change in the pre versus post GDL periods. There was a substantial reduction in teen crashes after GDL implementation (RR = 0.66, 95%CI: [0.65, 0.67]), and this effect was robust across gender and time-of-day (light/dark). We found evidence that this reduction varied across space; areas with more alcohol outlets corresponded to a larger daytime crash rate reduction post-GDL, while areas near schools corresponded to a smaller daytime crash rate reduction. Concentrations of movie theatres corresponded to larger post-GDL crash rate reductions after dark. Maximizing the substantial successes of GDL programs requires understanding why crash rate reductions were larger in some areas following GDL implementation, and harnessing that understanding to improve its effectiveness across a state, focusing on identifying priorities for improving driver training (e.g., by parents and driver educators), law enforcement, and future policy changes to current GDL laws.

Patterns of Racial/Ethnic Disparities and Prevalence in HIV and Syphilis Diagnoses Among Men Who Have Sex With Men, 2016: A Novel Data Visualization.

To describe disparities in HIV infection and syphilis among gay, bisexual, and other men who have sex with men (MSM) in US states through ratio-based measures and graphical depictions of disparities. We used state-level surveillance data of reported HIV and syphilis cases in 2015 and 2016, and estimates of MSM population sizes to estimate HIV and syphilis prevalence by race/ethnicity and rate ratios (RRs) and to visually display patterns of disparity and prevalence among US states. State-specific rates of new HIV diagnoses were higher for Black than for White MSM (RR range = 2.35 [Rhode Island] to 10.12 [Wisconsin]) and for Hispanic than for White MSM (RR range = 1.50 [Tennessee] to 5.78 [Pennsylvania]). Rates of syphilis diagnoses were higher for Black than for White MSM in 42 of 44 states (state RR range = 0.89 [Hawaii] to 17.11 [Alaska]). Scatterplots of HIV diagnosis rates by race showed heterogeneity in epidemic scenarios, even in states with similar ratio-based disparity measures. There is a widely disparate impact of HIV and syphilis among Black and Hispanic MSM compared with White MSM. Between-state variation suggests that states should tailor and focus their prevention responses to best address state data.

Associations between life history speed and sexually coercive behavior

The current study seeks to further understand risk factors for sexually coercive behavior by evaluating how indicators of population level average Life History Speed (LHS; e.g., teen birth rate) compare to typical criminogenic variables (e.g., Socioeconomic status) as predictors of state variation in rape rates across the 50 United States, as well as the relationship between individuals' LHS and self-reported proclivity for, and perpetration of, sexually coercive behaviors in a community sample (n = 162). LH strategies are described as a continuum of “LH speeds,” and variation in LHS has been connected to variation in aggressive and violent behavior. The current project extends this research by testing population level variation in LHS indicators and individual variation in psychometric LHS as a predictor of variation in sexually coercive behavior. At the U.S. state level, the teen birth rate (B = 0.63, p = .016) was the strongest predictor of between-state variation in rape rates. Although significant bivariate associations were found between psychometric LHS and sexually coercive behavior, at the multivariate level, facets of LHS were linked to self-reported propensity to engage in sexual coercion, but less so with actual perpetration.

State Variation in Quality Outcomes and Disparities in Outcomes in Community Health Centers.

Prior studies have reported that community health centers perform as well as other sites of care, despite serving more vulnerable patient populations. However, there is little prior study of geographic variation in quality outcomes and disparities in outcomes in this setting. Quantifying geographic variation is important so as to target quality improvement efforts and funding and to learn from states where total quality is highest and racial/ethnic disparities are lowest. To estimate between-state variation in hypertension, diabetes, and pregnancy outcomes and racial/ethnic disparities in these outcomes for health center patients. Using data on all US health centers from 2010 to 2014 (N=1047 health centers/year, or 21.2 million patients in 2014), we used linear regression models to estimate adjusted quality outcomes and racial/ethnic disparities in quality outcomes by state for hypertension control, diabetes control, and normal birthweight. We found wide variation in both outcomes and racial/ethnic disparities in outcomes between states for patients seen at health centers. For instance, between states, the mean proportion of patients with hypertension control ranged from 58% to 70% for white patients, from 49% to 64% for black patients, and from 53% to 74% for Hispanic patients (P<0.001). Racial/ethnic disparities in outcomes ranged from negative or nonsignificant in some states to positive in others. Wide variation in health center patient outcomes and disparities in outcomes is observed between states. This variation suggests that policymakers should target funding and interventions to underperforming states, and identify determinants of high quality in higher performing states.

Will Public Pre-K Really Close Achievement Gaps? Gaps in Prekindergarten Quality Between Students and Across States

Publicly funded pre-K is often touted as a means to narrow achievement gaps, but this goal is less likely to be achieved if poor and/or minority children do not, at a minimum, attend equal quality pre-K as their non-poor, non-minority peers. In this paper, I find large “quality gaps” in public pre-K between poor, minority students and non-poor, non-minority students, ranging from 0.3 to 0.7 SD on a range of classroom observational measures. I also find that even after adjusting for several classroom characteristics, significant and sizable quality gaps remain. Finally, I find much between-state variation in gap magnitudes and that state-level quality gaps are related to state-level residential segregation. These findings are particularly troubling if a goal of public pre-K is to minimize inequality.

Ignition Interlock Laws: Effects on Fatal Motor Vehicle Crashes, 1982-2013.

Alcohol-involved motor vehicle crashes are a major cause of preventable mortality in the U.S., leading to more than 10,000 fatalities in 2013. Ignition interlocks, or alcohol-sensing devices connected to a vehicle's ignition to prevent it from starting if a driver has a predetermined blood alcohol content (BAC) level, are a promising avenue for preventing alcohol-involved driving. This study sought to assess the effects of laws requiring ignition interlocks for some or all drunk driving offenders on alcohol-involved fatal crashes. A multilevel modeling approach assessed the effects of state interlock laws on alcohol-involved fatal crashes in the U.S. from 1982 to 2013. Monthly data on alcohol-involved crashes in each of the 50 states was collected in 2014 from the National Highway Traffic Safety Administration Fatality Analysis Reporting System. Random-intercept models accounted for between-state variation in alcohol-involved fatal crash rates and autocorrelation of within-state crash rates over time. Analysis was conducted in 2015. State laws requiring interlocks for all drunk driving offenders were associated with a 7% decrease in the rate of BAC >0.08 fatal crashes and an 8% decrease in the rate of BAC ≥0.15 fatal crashes, translating into an estimated 1,250 prevented BAC >0.08 fatal crashes. Laws requiring interlocks for segments of high-risk drunk driving offenders, such as repeat offenders, may reduce alcohol-involved fatal crashes after 2 years of implementation. Ignition interlock laws reduce alcohol-involved fatal crashes. Increasing the spread of interlock laws that are mandatory for all offenders would have significant public health benefit.

Geographic Variation in Early MRI for Acute Work-Related Low Back Pain and Associated Factors.

Retrospective cohort study using medical claims data. To document the extent of geographic variation in utilization of magnetic resonance imaging (MRI) for working-age patients early in the course of acute, disabling low back pain (LBP); to identify potential factors associated with the most extreme variations. Although guidelines discourage MRI in acute uncomplicated LBP, this practice is highly prevalent. Geographic variation in radiologic testing is common, and may indicate problems with access or quality of care, yet this has not been studied in working-age patients with LBP (a frequent cause for acute care visits). All cases of acute, disabling LBP with onset between 1/1/2002 and 12/31/2007 were selected from a large workers' compensation data source. Detailed information from medical bills was used to identify persons who received early MRI (within 30 days of onset), classify cases by LBP severity, and exclude those with concurrent injuries or diseases, and/or prior LBP disability. Individual predictors included age, gender, job tenure, and industry. State-level predictors included economic, physician supply and practice variables, workers compensation system features, and MRI testing location. Generalized linear mixed models were constructed to evaluate within- and between-state variability, selecting the six highest and six lowest MRI utilization states. State rates of early MRI scanning varied from 6.0% to 58.4%. In the 12 selected most extreme states, non-hospital MRI sites and lower state median income were associated with higher rates of early MRIs, explaining 84% of between-state variation, and 12.5% of all observed variability. Inter-state differences in MRI rates were greatest for lower-severity cases. Higher severity diagnoses were more common in high utilization states. Between-state inappropriate early MRI variability is largely explained by rate of non-hospital MRI sites and state median income. Potential solutions include efforts to address inappropriate referral patterns based on private MRI facility ownership, and to improve quality of communication with low-income patients. 4.

CHIP-ing Away at Health Disparities: Has State-Provided Health Insurance Reduced Race- and Nativity-Based Differences in Health Care Utilization among US Children?

Although recent legal permanent residents and undocumented immigrants are generally barred from accessing public health insurance, some US states cover immigrant children through the Children’s Health Insurance Program (CHIP). In this study, we examine the contextual effect of US state health insurance eligibility policy, particularly with respect to immigrant children, on race/ethnic and nativity-based disparities in children’s routine health care. Utilizing our original data on state CHIP eligibility policies and child-level data from the Survey of Income and Program Participation, we find that a significant portion of between-state variation in children’s routine health care results from diversity in CHIP eligibility rules for poor and foreign-born children. Immigrant-specific disparities are reduced when states do not require five years of residency for CHIP participation.

Individual and Ecological Variation in Child Undernutrition in India

Despite the substantial burden of child undernutrition in South Asia, little is known on the relative importance and contribution of individual and micro/macro environments in shaping variation in child undernutrition. Using measures of anthropometry, we decompose the variation in child undernutrition in India to the levels of child, communities and states, quantifying the extent to which variation at each of these levels can be explained by known proximal and distal risk factors, measured at the individual (child/household) level. Data are from under-five singleton children participating in the 2005–2006 National Family Health Survey (NFHS-3). The outcome variables were: height-for-age z-score (HAZ), weight-for-age z-score (WAZ) and weight-for-height z-score (WHZ), as well as their associated measures of anthropometric failure: stunting, underweight and wasting, defined as more than two standard deviations below the median of the referred z-scores, respectively. We also considered the composite index of anthropometric failure (CIAF), defined by combinations of child anthropometric failure. After accounting for risk factors, of the total variation in HAZ, 93.2 per cent, 4.9 per cent and 1.9 per cent were attributable to the individual, community and state levels, respectively. The observed risk factors explained 6.3 per cent and 46.9 per cent of the variation at the individual and community level, respectively; however, between-state variation was not explained by these risk factors. Variability in other measures of anthropometry and anthropometric failure largely followed this pattern. Additionally, there were also considerable differences in the amount of variation at the individual and community levels among different states. Hence, there is a substantial variability at the community level compared to the state level, suggesting the presence of micro-geographies of undernutrition. Additionally, while a substantial majority of the variation in child undernutrition is at the individual level, our ability to explain variability in undernutrition at the individual-level risk factors is extremely limited. Further research is needed to explore community level or environmental factors affecting child undernutrition, generating evidence for policies to target these determinants.