Bringing the individual back to small-area variation studies: a multilevel analysis of all-cause mortality in Andalusia, Spain

Starting then stopping: a nationwide register-based study on the magnitude, predictors, and urban-rural patterns of under-vaccination variation across health centers in The Gambia

OBJECTIVES

Six million children were under-vaccinated in 2022. Our study aimed to 1) quantify the magnitude of under-vaccination variation between health facilities, 2) assess to which extent individual and health center level factors contributed to the variation, 3) identify individual and health facility factors associated with under-vaccination, and 4), explore rural vs. urban health facility variations.

METHODS

We used data from 61,839 children from The Gambia national routine vaccination register. We cross tabulated under-vaccination status across study variables and fitted two-level random intercept multilevel logistic regression models to measure variance, contribution to the variance, and factors associated with the variance and under-vaccination.

RESULTS

We found that 7% of the prevalence of under-vaccination was due to variation between health facilities. Thirty-seven percent of the variation was explained by individual and health center variables. The variables explained 36% of the variance in urban and 19% in rural areas. Children who were not vaccinated at 4 months or with delayed history, due for vaccination in the rainy season, and health facilities with very small or large population to health worker ratios had higher under-vaccination odds.

CONCLUSION

Our study indicates that one of the pathways to improving vaccination coverage is addressing factors driving under-vaccination inequities between health facilities through urban-rural differentiated strategies.

Neighborhood built environment, obesity, and diabetes: A Utah siblings study

Background

This study utilizes innovative computer vision methods alongside Google Street View images to characterize neighborhood built environments across Utah.

Methods

Convolutional Neural Networks were used to create indicators of street greenness, crosswalks, and building type on 1.4 million Google Street View images. The demographic and medical profiles of Utah residents came from the Utah Population Database (UPDB). We implemented hierarchical linear models with individuals nested within zip codes to estimate associations between neighborhood built environment features and individual-level obesity and diabetes, controlling for individual- and zip code-level characteristics (n = 1,899,175 adults living in Utah in 2015). Sibling random effects models were implemented to account for shared family attributes among siblings (n = 972,150) and twins (n = 14,122).

Results

Consistent with prior neighborhood research, the variance partition coefficients (VPC) of our unadjusted models nesting individuals within zip codes were relatively small (0.5%-5.3%), except for HbA1c (VPC = 23%), suggesting a small percentage of the outcome variance is at the zip code-level. However, proportional change in variance (PCV) attributable to zip codes after the inclusion of neighborhood built environment variables and covariates ranged between 11% and 67%, suggesting that these characteristics account for a substantial portion of the zip code-level effects. Non-single-family homes (indicator of mixed land use), sidewalks (indicator of walkability), and green streets (indicator of neighborhood aesthetics) were associated with reduced diabetes and obesity. Zip codes in the third tertile for non-single-family homes were associated with a 15% reduction (PR: 0.85; 95% CI: 0.79, 0.91) in obesity and a 20% reduction (PR: 0.80; 95% CI: 0.70, 0.91) in diabetes. This tertile was also associated with a BMI reduction of -0.68 kg/m2 (95% CI: -0.95, -0.40).

Conclusion

We observe associations between neighborhood characteristics and chronic diseases, accounting for biological, social, and cultural factors shared among siblings in this large population-based study.

Individual and community level factors associated with modern contraceptive utilization among women in Ethiopia: Multilevel modeling analysis

BACKGROUND

Modern contraceptive utilization is the most effective intervention to tackle unintended pregnancy and thereby reduce abortion and improve maternal, child, and newborn health. However, multilevel factors related to low modern contraceptive utilization and the robust analysis required for decision-making were scarce in Ethiopia.

OBJECTIVE

To investigate the individual and community-level predictors of modern contraceptive utilization among reproductive-age women in Ethiopia.

METHODS

We utilized data from a cross-sectional 2019 Performance Monitoring for Action Ethiopia survey. The survey employed a stratified two-stage cluster sampling method to select households for inclusion. In Stata version 16.0, the data underwent cleaning, aggregation, and survey weighting, following which a descriptive analysis was performed utilizing the "svy" command. Subsequently, the primary analysis was executed using R software version 4.1.3. We fitted a two-level mixed effects logistic regression model on 6,117 reproductive-age women nested within 265 enumeration areas (clusters). The fixed effect models were fitted. The measures of variation were explained by intra-cluster correlation, median odds ratio, and proportional change in variance. The shrinkage factor was calculated to estimate the effects of cluster variables using the Interval odds ratio and proportion opposed odds ratio. Finally, the independent variables with a significance level of (P<0.05) and their corresponding Adjusted Odds ratios and 95% confidence intervals were described for the explanatory factors in the final model.

RESULTS

In Ethiopia, the prevalence of modern contraceptive utilization was only 37.% (34.3 to 39.8). Women who attained primary, secondary, and above secondary levels of education were more likely to report modern contraceptive utilization with AOR of 1.47, 1.73, and 1.58, respectively. Divorced/widowed women were less likely to report modern contraceptive utilization (AOR:0.18, 95% CI 0.13,0.23) compared to never-married women. Discussions between women and healthcare providers at the health facility about family planning were positively associated with modern contraceptive utilization (AOR:1.84, 95% CI: 1.52, 2.23). Community-level factors have a significant influence on modern contraceptive utilization, which is attributed to 21.9% of the total variance in the odds of using modern contraceptives (ICC = 0.219). Clusters with a higher proportion of agrarian (AOR: 2.27, 95% CI 1.5, 3.44), clusters with higher literacy (AOR: 1.46, 95% CI 1.09, 1.94), clusters with empowered women and girls about FP (AOR: 1.47, 95% CI 1.11, 1.93) and clusters with high supportive attitudes and norms toward FP (AOR: 1.37, 95% CI 1.04, 1.81) had better modern contraceptive utilization than their counterparts.

CONCLUSION

In Ethiopia, understanding the factors related to modern contraceptive use among women of reproductive age requires consideration of both individual and community characteristics. Hence, to enhance family planning intervention programs, it is essential to focus on the empowerment of women and girls, foster supportive attitudes towards family planning within communities, collaborate with education authorities to enhance overall community literacy, pay special attention to pastoralist communities, and ensure that reproductive-age women as a whole are targeted rather than solely focusing on married women.

Age-period-cohort analysis of lung cancer mortality inequalities in Southern Spain: missed opportunities for implementing equitable tobacco control policies

BACKGROUND

Lung cancer mortality in European countries shows different epidemiological patterns according to sex and socioeconomic variables. Some countries show decreasing rates in both sexes, while others show a delayed profile, with increasing mortality in women, inconsistently influenced by socioeconomic status. Our aim was to evaluate the effect of age, period and birth cohort on lung cancer mortality inequalities in men and women in Andalusia, the southernmost region in Spain.

METHODS

We used the Longitudinal Database of the Andalusian Population, which collects demographic and mortality data from the 2001 census cohort of more than 7.35 million Andalusians, followed up between 2002 and 2016. Mortality rates were calculated for men and women by educational level, and small-area deprivation. Poisson models were used to assess trends in socioeconomic inequalities in men and women. Finally, age-period-cohort (APC) models were used separately for each educational level and gender.

RESULTS

There were 39,408 lung cancer deaths in men and 5,511 in women, yielding crude mortality rates of 78.1 and 11.4 × 10⁵ person-years, respectively. In men higher mortality was found in less educated groups and inequalities increased during the study period: i.e. the rate ratio for primary studies compared to university studies increased from 1.30 (CI95:1.18-1.44) to 1.57 (CI95:1.43-1.73). For women, educational inequalities in favour of the less educated tended to decrease moderately. In APC analysis, a decreasing period effect in men and an increasing one in women were observed. Cohort effect differed significantly by educational level. In men, the lower the educational level, the earlier the peak effect was reached, with a 25-year difference between the least-educated and college-educated. Conversely, college-educated women reached the peak effect with a 12-year earlier cohort than the least-educated women. The decline of mortality followed the same pattern both in men and women, with the best-educated groups experiencing declining rates with earlier birth cohorts.

CONCLUSIONS

Our study reveals that APC analysis by education helps to uncover changes in trends occurring in different socioeconomic and gender groups, which, combined with data on smoking prevalence, provide important clues for action. Despite its limitations, this approach to the study of lung cancer inequalities allows for the assessment of gaps in historical and current tobacco policies and the identification of population groups that need to be prioritised for public health interventions.

The impact of dementia, frailty and care home characteristics on SARS-CoV-2 incidence in a national cohort of Welsh care home residents during a period of high community prevalence

BACKGROUND

dementia may increase care home residents' risk of COVID-19, but there is a lack of evidence on this effect and on interactions with individual and care home-level factors.

METHODS

we created a national cross-sectional retrospective cohort of care home residents in Wales for 1 September to 31 December 2020. Risk factors were analysed using multi-level logistic regression to model the likelihood of SARS-CoV-2 infection and mortality.

RESULTS

the cohort included 9,571 individuals in 673 homes. Dementia was diagnosed in 5,647 individuals (59%); 1,488 (15.5%) individuals tested positive for SARS-CoV-2. We estimated the effects of age, dementia, frailty, care home size, proportion of residents with dementia, nursing and dementia services, communal space and region. The final model included the proportion of residents with dementia (OR for positive test 4.54 (95% CIs 1.55-13.27) where 75% of residents had dementia compared to no residents with dementia) and frailty (OR 1.29 (95% CIs 1.05-1.59) for severe frailty compared with no frailty). Analysis suggested 76% of the variation was due to setting rather than individual factors. Additional analysis suggested severe frailty and proportion of residents with dementia was associated with all-cause mortality, as was dementia diagnosis. Mortality analyses were challenging to interpret.

DISCUSSION

whilst individual frailty increased the risk of COVID-19 infection, dementia was a risk factor at care home but not individual level. These findings suggest whole-setting interventions, particularly in homes with high proportions of residents with dementia and including those with low/no individual risk factors may reduce the impact of COVID-19.

Small area variation in severe, moderate, and mild anemia among women and children: A multilevel analysis of 707 districts in India

India is home to the highest global number of women and children suffering from anemia, with one in every two women impacted. India's current strategy for targeting areas with a high anemia burden is based on district-level averages, yet this fails to capture the substantial small area variation in micro-geographical (small area) units such as villages. We conducted statistical and econometric analyses to quantify the extent of small area variation in the three grades of anemia (severe, moderate, and mild) among women and children across 36 states/union territories and 707 districts of India. We utilized data from the fifth round of the National Family Health Survey conducted in 2019-21. The final analytic sample for analyses was 183,883 children aged 6-59 months and 690,153 women aged 15-49 years. The primary outcome variable for the analysis was the three anemia grades among women and children. We adopted a three-level and four-level logistic regression model to compute variance partitioning of anemia among women and children. We also computed precision-weighted prevalence estimates of women and childhood anemia across 707 districts and within-district, between-cluster variation using standard deviation (SD). For severe anemia among women, small area (villages or urban blocks) account for highest share (46.1%; Var: 0.494; SE: 0.150) in total variation followed by states (39.4%; Var: 0.422; SE: 0.134) and districts (12.8%; Var: 0.156; SE: 0.012). Similarly, clusters account for the highest share in the variation in severe (61.3%; Var: 0.899; SE: 0.069) and moderate (46.4%: Var: 0.398; SE: 0.011) anemia among children. For mild and moderate anemia among women, however, states were the highest source of variation. Additionally, we found a high and positive correlation between mean prevalence and inter-cluster SD of moderate and severe anemia among women and children. In contrast, the correlation was weaker for mild anemia among women (r = 0.61) and children (0.66). In this analysis, we are positing the critical importance of small area variation within districts when designing strategies for targeting high burden areas for anemia interventions.

Intraurban socioeconomic inequalities in life expectancy: a population-based cross-sectional analysis in the city of Córdoba, Argentina (2015-2018)

OBJECTIVES

To evaluate variability in life expectancy at birth in small areas, describe the spatial pattern of life expectancy, and examine associations between small-area socioeconomic characteristics and life expectancy in a mid-sized city of a middle-income country.

DESIGN

Cross-sectional, using data from death registries (2015-2018) and socioeconomic characteristics data from the 2010 national population census.

PARTICIPANTS/SETTING

40 898 death records in 99 small areas of the city of Córdoba, Argentina. We summarised variability in life expectancy at birth by using the difference between the 90th and 10th percentile of the distribution of life expectancy across small areas (P90-P10 gap) and evaluated associations with small-area socioeconomic characteristics by calculating a Slope Index of Inequality in linear regression.

PRIMARY OUTCOME

Life expectancy at birth.

RESULTS

The median life expectancy at birth was 80.3 years in women (P90-P10 gap=3.2 years) and 75.1 years in men (P90-P10 gap=4.6 years). We found higher life expectancies in the core and northwest parts of the city, especially among women. We found positive associations between life expectancy and better small-area socioeconomic characteristics, especially among men. Mean differences in life expectancy between the highest versus the lowest decile of area characteristics in men (women) were 3.03 (2.58), 3.52 (2.56) and 2.97 (2.31) years for % adults with high school education or above, % persons aged 15-17 attending school, and % households with water inside the dwelling, respectively. Lower values of % overcrowded households and unemployment rate were associated with longer life expectancy: mean differences comparing the lowest versus the highest decile were 3.03 and 2.73 in men and 2.57 and 2.34 years in women, respectively.

CONCLUSION

Life expectancy is substantially heterogeneous and patterned by socioeconomic characteristics in a mid-sized city of a middle-income country, suggesting that small-area inequities in life expectancy are not limited to large cities or high-income countries.

Effects of Multifaceted Determinants on Individual Stress: The Mediating Role of Social Capital

Stress substantially results in various negative health outcomes. While there is a nexus between social capital and individual stress, previous studies have primarily explored the direct relationship between them. Social capital may potentially have an indirect effect on perceived stress via social networking pathways that provide accessible resources. This study addresses this research gap by exploring the mediating effect of social capital for associations between personal-level features, personal-level behaviors, physical environments, and perceived stress. A household drop-off survey of 600 respondents was collected from two neighborhoods in Korea and analyzed by structural equation models. Results showed that social capital acted as a mediator on perceived stress level. The frequency of community center use had both direct and indirect impacts on stress level through social capital. Those who were satisfied with the cleanliness of the neighborhood had a higher level of social capital and a lower level of stress indirectly through social capital. Households with more children had a lower level of social capital, while persons who had chronic disease and were more extroverted, agreeable, and open to others enjoyed a higher level of social capital. The results provide policy implications on how community revitalization affects social capital and perceived stress.

Physical and social environmental factors related to co-occurrence of unhealthy lifestyle behaviors

Previous work identifying determinants of co-occurrence of behavioral risk factors have focused on their association with individuals' characteristics with scant attention paid to their relationship to contextual factors. Data came from 21,007 individuals ≥15 years of age who participated in the cross-sectional 2011-2012 Spanish National Health Survey. Two indicators were defined by tobacco consumption, alcohol intake, diet, physical activity, and body mass index. The first indicator, based on dichotomized measures, ranges from 0 to 5. The second one (unhealthy lifestyle index), ranges from 0 to 15, with 0 denoting the healthiest score. Among the determinants, we examined social support, five perceived characteristics of the neighborhood, and the socioeconomic deprivation index of the census tract of residence. Data were analyzed using multilevel linear and logistic regression models adjusted for the main sociodemographic characteristics. Using the dichotomized indicator, the probability of having 3-5 risk factors versus <3 factors was associated with low social support (Odds Ratio [OR] 1.50; 95% Confidence Interval [CI]: 1.25-1.80). Issues surrounding neighborhood cleanliness (OR = 1.18; 95%CI: 1.04-1.33), air pollution (OR = 1.38; 95%CI: 1.16-1.64), and street crime (OR = 1.21; 95%CI: 1.03-1.42) were associated with determinants of co-occurrence. Risk factors co-occurrence increased as deprivation level increased: the OR for the highest deprivation quintile versus the lowest was 1.30 (95%CI: 1.14-1.48). Similar results were observed when using the unhealthy lifestyle index. Poorer physical and social environments are related to greater co-occurrence of risk factors for chronic diseases. Health promotion interventions targeting the prevention of risk factors should consider the contextual characteristics of the neighborhood environment.

Spatial distribution of neighborhood-level housing prices and its association with all-cause mortality in Seoul, Korea (2013-2018): A spatial panel data analysis

Housing prices are known to be a relevant indicator of the socioeconomic position of the neighborhood. In a society where the market system mainly drives housing prices, residents' spatial patterning is formulated according to their socioeconomic position. Dividing the 2013-2018 entire study period into three periods, we explored the spatial distribution of housing prices and all-cause mortality and their association in Seoul, the country's capital city. The government authorities' data and 2015 census data were used for the study. We mapped the spatial distribution of housing prices and all-cause mortality and investigated the changes in distribution. We conducted a pooled ordinary least square (OLS) and spatial panel regression analysis to estimate housing prices elasticity of all-cause mortality. We also explored the possible mediating role of housing prices on the educational composition's effect on all-cause mortality. We found the common trends of increasing spatial patterning of housing prices and all-cause mortality. The magnitude of spatial patterning was far greater in housing prices than all-cause mortality. A pooled OLS regression analysis found that a 1% increase in housing price was associated with a 0.11% reduction in all-cause mortality after controlling the explanatory variables. Attenuation in the regression coefficient's magnitude was found after adding the neighborhood's educational composition to the model. As a result of spatial panel analysis, we found a direction and scale similar to the housing price elasticity of all-cause mortality in the final pooled OLS model. The results suggested that spatial health inequality in Korea's urban space mainly stems from socioeconomic inequality.

Impact of COVID-19 on the Health of the General and More Vulnerable Population and Its Determinants: Health Care and Social Survey-ESSOC, Study Protocol

This manuscript describes the rationale and protocol of a real-world data (RWD) study entitled Health Care and Social Survey (ESSOC, Encuesta Sanitaria y Social). The study's objective is to determine the magnitude, characteristics, and evolution of the COVID-19 impact on overall health as well as the socioeconomic, psychosocial, behavioural, occupational, environmental, and clinical determinants of both the general and more vulnerable population. The study integrates observational data collected through a survey using a probabilistic, overlapping panel design, and data from clinical, epidemiological, demographic, and environmental registries. The data will be analysed using advanced statistical, sampling, and machine learning techniques. The study is based on several measurements obtained from three random samples of the Andalusian (Spain) population: general population aged 16 years and over, residents in disadvantaged areas, and people over the age of 55. Given the current characteristics of this pandemic and its future repercussions, this project will generate relevant information on a regular basis, commencing from the beginning of the State of Alarm. It will also establish institutional alliances of great social value, explore and apply powerful and novel methodologies, and produce large, integrated, high-quality and open-access databases. The information described here will be vital for health systems in order to design tailor-made interventions aimed at improving the health care, health, and quality of life of the populations most affected by the COVID-19 pandemic.

Small area influences on the individual unhealthy lifestyle behaviors: A multilevel analysis of discriminatory accuracy

We estimated the discriminatory power of area of residence (census tract) on the prevalence of main risk factors for chronic diseases. Results, based on a sample of 21,007 participants from the 2011-2012 National Health Survey of Spain, show a differential influence of the geosocial environment on the four health risk factors. Accounting for census tracts substantially increases the discriminatory power regarding at-risk alcohol consumption, unbalanced diet, and leisure-time sedentarism but not tobacco consumption. However, the socioeconomic characteristics of the tracts played a minor role. Further research on the specific geosocial contextual variables explaining variability in these risk factors is necessary.

Multimorbidity in large Canadian urban centres: A multilevel analysis of pooled 2015-2018 cross-sectional cycles of the Canadian Community Health Survey

BACKGROUND

There is limited knowledge on how the prevalence of multimorbidity varies within and across major Canadian urban centres. The objective of this study was to investigate the between-neighbourhood variation in the prevalence of multimorbidity in Canada's large urban centres, controlling for compositional effects associated with individual-level demographic and socioeconomic factors.

METHODS

Cross-sectional data from the 2015-2018 cycles of the Canadian Community Health Survey (CCHS) were pooled at the microdata level. Respondents (20 years and older) residing in one of the 35 census metropolitan areas (CMAs) were included (N = 100,803). Census tracts (CTs) were used as a measure of neighbourhood. To assess the between-neighbourhood differences in multimorbidity prevalence, we fitted three sequential random intercept logistic regression models.

RESULTS

During the 2015-2018 period, 8.1% of residents of large urban centres had multimorbidity. The results from the unadjusted model indicate that 13.4% of the total individual variance in multimorbidity could be attributed to the between-neighbourhood differences. After adjustment for overall characteristics of the CMAs in which these neighbourhoods are located, as well as for individual-level demographic and socioeconomic factors related to compositional effects, 11.0% of the individual variance in multimorbidity could still be attributed to the between-neighbourhood differences.

CONCLUSION

There is significant and substantial geographic variation in multimorbidity prevalence across neighbourhoods in Canada's large urban centres. Residing in some neighbourhoods could be associated with increased odds of having multimorbidity, even after accounting for overall characteristics of the CMAs in which these neighbourhoods are located, as well as individual-level factors.

Cross-classified Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) to evaluate hospital performance: the case of hospital differences in patient survival after acute myocardial infarction

OBJECTIVE

To describe a novel strategy, Multilevel Analysis of Individual Heterogeneity and Discriminatory Accuracy (MAIHDA) to evaluate hospital performance, by analysing differences in 30-day mortality after a first-ever acute myocardial infarction (AMI) in Sweden.

DESIGN

Cross-classified study.

SETTING

68 Swedish hospitals.

PARTICIPANTS

43 247 patients admitted between 2007 and 2009, with a first-ever AMI.

PRIMARY AND SECONDARY OUTCOME MEASURES

We evaluate hospital performance by analysing differences in 30-day mortality after a first-ever AMI using a cross-classified multilevel analysis. We classified the patients into 10 categories according to a risk score (RS) for 30-day mortality and created 680 strata defined by combining hospital and RS categories.

RESULTS

In the cross-classified multilevel analysis the overall RS adjusted hospital 30-day mortality in Sweden was 4.78% and the between-hospital variation was very small (variance partition coefficient (VPC)=0.70%, area under the curve (AUC)=0.54). The benchmark value was therefore achieved by all hospitals. However, as expected, there were large differences between the RS categories (VPC=34.13%, AUC=0.77) CONCLUSIONS: MAIHDA is a useful tool to evaluate hospital performance. The benefit of this novel approach to adjusting for patient RS is that it allowed one to estimate separate VPCs and AUC statistics to simultaneously evaluate the influence of RS categories and hospital differences on mortality. At the time of our analysis, all hospitals in Sweden were performing homogeneously well. That is, the benchmark target for 30-day mortality was fully achieved and there were not relevant hospital differences. Therefore, possible quality interventions should be universal and oriented to maintain the high hospital quality of care.

Geographic variation in cardiometabolic risk factor prevalence explained by area-level disadvantage in the Illawarra-Shoalhaven region of the NSW, Australia

Cardiometabolic risk factors (CMRFs) demonstrate significant geographic variation in their distribution. The study aims to quantify the general contextual effect of the areas on CMRFs; and the geographic variation explained by area-level socioeconomic disadvantage. A cross sectional design and multilevel logistic regression methods were adopted. Data included objectively measured routine pathology test data between years 2012 and 2017 on: fasting blood sugar level; glycated haemoglobin; total cholesterol; high density lipoprotein; urinary albumin creatinine ratio; estimated glomerular filtration rate; and body mass index. The 2011 Australian census based Index of Relative Socioeconomic Disadvantage (IRSD) were the area-level study variables, analysed at its smallest geographic unit of reporting. A total of 1,132,029 CMRF test results from 256,525 individuals were analysed. After adjusting for individual-level covariates, all CMRFs significantly associated with IRSD and the probability of higher risk CMRFs increases with greater area-level disadvantage. Though the specific contribution of IRSD in the geographic variation of CMRF ranged between 57.8 and 14.71%, the general contextual effect of areas were found minimal (ICCs 0.6-3.4%). The results support universal interventions proportional to the need and disadvantage level of populations for the prevention and control of CMRFs, rather than any area specific interventions as the contextual effects were found minimal in the study region.

Geographic variation in tobacco use in India: a population-based multilevel cross-sectional study

OBJECTIVE

This study aims to quantify the extent to which people's use of tobacco products varies by local areas (city ward and village) across India and the variation in this clustering by tobacco products.

DESIGN

Cross-sectional study.

SETTING AND PARTICIPANTS

Data on 73 954 adults across 2547 city wards and villages were available for analysis from 30 states and 2 union territories in India.

PRIMARY AND SECONDARY OUTCOME MEASURES

We included as primary outcomes self-reported any tobacco use, current cigarette smoking, current bidi smoking, current smokeless tobacco use and a derived variable for dual use describing respondents who engaged in both smoking and smokeless tobacco use.

RESULTS

The median risk of an individual using tobacco was 1.64 times greater if a person hypothetically moved from an area of low to high risk of tobacco use (95% CI: 1.60 to 1.69). Area-level partitioning of variation differed by tobacco product used. Median ORs ranged from 1.77 for smokeless tobacco use to 2.53 for dual use.

CONCLUSIONS

Tobacco use is highly clustered geographically in India. To be effective in India, policy interventions should be directed to influence specific local contextual factors on adult tobacco use. Where people live in India influences their use of tobacco, and this association may be greater than has been observed in other settings. Tailoring tobacco control policies for local areas in India may, therefore, provide substantial public health benefits.

Physical activity and self-rated health: role of contexts

The purpose of this study is to examine how much variance in self-rated health is attributable to household and administrative-area levels. Additionally, the study investigates the association between physical activity and self-rated health while adjusting for various covariates at the individual, household, and district levels in Seoul, South Korea. A cross-sectional study of the 2009 Seoul Welfare Panel Study conducted by Seoul Welfare Foundation was utilized. The final sample included 7,761 individuals within 3,617 households in 25 administrative areas. Three-level random intercept logistic models were fitted. The results showed that a small proportion (3.52%) of variance in self-rated health was attributed to the administrative-area level, while a relatively large proportion (33.78%) was attributed to the household level. This study also found a positive association between physical activity and self-rated health, even after controlling for covariates at multiple levels. The results indicated that public health interventions promoting physical activity may have a beneficial effect on individual health. Additionally, public health interventions for improving individual health may not be efficient if directed only at the administrative-area level, as only a small portion of variance in self-rated health was attributable to this level.

Antimicrobial prescribing and infections in long-term care facilities (LTCF): a multilevel analysis of the HALT 2016 study, Ireland, 2017

Background

The 2016 point prevalence survey (PPS) of healthcare-associated infections (HAI) and antimicrobial use (AMU) in Irish long-term care facilities (LTCF) (HALT) showed a 9.8% AMU and 4.4% HAI prevalence, based on aggregated data analysis.

Aim

Our aim was to identify institutional and resident risk factors of AMU and HAI.

Methods

HALT 2016 gathered information using institutional and resident questionnaires, for residents who met the surveillance definition of active HAI and/or AMU, limiting analysis to the aggregated institutional level. In January 2017, we requested additional data on age, sex, urinary catheter use and disorientation of current residents from HALT 2016 LTCF and matched to 2016 HALT data.

Results

Of 224 HALT 2016 LTCF, 80 provided additional information on 3,816 residents; prevalence of AMU was 10.6% and HAI was 4.7%. Presence of a coordinating physician (Odds ratio (OR): 0.3; 95% confidence interval (CI): 0.2–0.6), antimicrobial stewardship committee (OR: 0.2; 95%; CI: 0.1–0.6), healthcare assistants (OR: 0.9; 95% CI: 0.9–1.0), antimicrobial consumption feedback (OR: 0.3; 95% CI: 0.1–0.6) and medical care by personal general practitioner (OR: 0.6; 95% CI: 0.7–1.0) were associated with less AMU and feedback on surveillance of infection prevention and control (IPC) practices (OR: 0.6; 95% CI: 0.3–1.0) with less HAI. AMU and HAI varied significantly between LTCF.

Conclusions

Multilevel modelling identified significant inter-facility variation, as well as institutional factors associated with AMU and HAI. An antimicrobial stewardship committee linked with feedback on IPC and prescribing was associated with reduced AMU and HAI.

Geographic variation in cardiometabolic risk distribution: A cross-sectional study of 256,525 adult residents in the Illawarra-Shoalhaven region of the NSW, Australia

INTRODUCTION

Metabolic risk factors for cardiovascular disease (CVD) warrant significant public health concern globally. This study aims to utilise the regional database of a major laboratory network to describe the geographic distribution pattern of eight different cardiometabolic risk factors (CMRFs), which in turn can potentially generate hypotheses for future research into locality specific preventive approaches.

METHOD

A cross-sectional design utilising de-identified laboratory data on eight CMRFs including fasting blood sugar level (FBSL); glycated haemoglobin (HbA1c); total cholesterol (TC); high density lipoprotein (HDL); albumin creatinine ratio (ACR); estimated glomerular filtration rate (eGFR); body mass index (BMI); and diabetes mellitus (DM) status was used to undertake descriptive and spatial analyses. CMRF test results were dichotomised into 'higher risk' and 'lower risk' values based on existing risk definitions. Australian Census Statistical Area Level 1 (SA1) were used as the geographic units of analysis, and an Empirical Bayes (EB) approach was used to smooth rates at SA1 level. Choropleth maps demonstrating the distribution of CMRFs rates at SA1 level were produced. Spatial clustering of CMRFs was assessed using Global Moran's I test and Local Indicators of Spatial Autocorrelation (LISA).

RESULTS

A total of 1,132,016 test data derived from 256,525 individuals revealed significant geographic variation in the distribution of 'higher risk' CMRF findings. The populated eastern seaboard of the study region demonstrated the highest rates of CMRFs. Global Moran's I values were significant and positive at SA1 level for all CMRFs. The highest spatial autocorrelation strength was found among obesity rates (0.328), and the lowest for albuminuria (0.028). LISA tests identified significant High-High (HH) and Low-Low (LL) spatial clusters of CMRFs, with LL predominantly in the less populated northern, central and southern regions of the study area.

CONCLUSION

The study describes a range of CMRFs with different distributions in the study region. The results allow generation of hypotheses to test in future research concerning location specific population health approaches.

Discontinuation of Chronic Benzodiazepine Use Among Adults in the United States

BACKGROUND

Guidelines recommend most benzodiazepine (BZD) treatment be short-term, though chronic BZD use is increasing.

OBJECTIVE

Determine the rate of BZD discontinuation among chronic users and identify patient- and provider-level factors associated with discontinuation.

DESIGN, SETTING, AND PARTICIPANTS

A retrospective cohort study using nationwide insurance claims data from 2014 to 2016 of US adults ≥ 18 years old with chronic BZD use (i.e., > 120 days) during the baseline year.

MAIN OUTCOMES AND MEASURES

The primary outcome was BZD discontinuation among chronic users after 1 year of follow-up. A series of multilevel logistic regression models examined the association of BZD discontinuation with patient and provider characteristics. Covariates included patient sociodemographics, medical and psychiatric comorbidity, co-prescribed opioids and other psychotropics, and characteristics of the prescribed BZD.

KEY RESULTS

Of 141,008 chronic BZD users, 13.4% discontinued use after 1 year. Females had lower odds of discontinuation (AOR 0.83, 99% CI 0.79-0.87), while African-American patients had higher odds (AOR 1.12, 99% CI 1.03-1.22). Those prescribed a high-potency BZD had lower odds of discontinuation (AOR 0.51, 99% CI 0.47-0.54), as did those prescribed an opioid (AOR 0.94, 99% CI 0.89-0.99). After adjusting for patient- and provider-level factors, differences between providers accounted for 5.8% of variation in BZD discontinuation (p < 0.001). The median odds ratio for provider was 1.54, an association with discontinuation larger than almost all patient-level clinical variables.

CONCLUSIONS

A small minority of patients prescribed chronic BZD in a given year are no longer prescribed BZDs 1 year later. There is significant variation in the likelihood of discontinuation accounted for by non-clinical factors such as race, geography, and a patient's provider, which had a stronger association with the odds of discontinuation than almost every other patient-level variable. Provider-facing elements of interventions to reduce BZD prescribing may be critical.

Social capital and perceived stress: The role of social context

BACKGROUND

The aim of this study is twofold: to investigate how much variance in individual perceived stress is attributed to household and area levels, respectively, and to examine the association between social capital at the individual, household, and area levels and perceived stress, while adjusting for various cofounders at the individual, household, and area levels.

METHODS

This study used data from the 2010 Seoul Welfare Panel Study conducted by the Seoul Welfare Foundation. A total sample of 5881 individuals in 2820 households within 25 areas was used for multilevel analysis.

RESULTS

The results showed that a relatively large proportion of variance in perceived stress was attributed to the household level (45.86%) in comparison with the area level (6.96%), which indicates that household or family context is more important in explaining variance in perceived stress than area. This study also found that some components of social capital were negatively associated with perceived stress, and the association between social capital and perceived stress varied depending on levels and types of social capital measures.

LIMITATIONS

This study is based on a cross-sectional design, and thus it is not clear about the temporal order between the relationship between social capital and perceived stress.

CONCLUSIONS

Overall, this study showed that research on social capital and mental health can be advanced by systematically investigating the role of household social capital, not just geographical social capital.

Albuminuria measurement in diabetic care: a multilevel analysis measuring the influence of accreditation on institutional performance

Background

Studies assessing institutional performance regarding quality of care are frequently performed using single-level statistical analyses investigating differences between provider averages of various quality indicators. However, such analyses are insufficient as they do not consider patients’ heterogeneity around those averages. Hence, we apply a multilevel analysis of individual-patient heterogeneity that distinguishes between ‘general’ (‘latent quality’ or measures of variance) and ‘specific’ (measures of association) contextual effects. We assess general contextual effects of the hospital departments and the specific contextual effect of a national accreditation programme on adherence to the standard benchmark for albuminuria measurement in Danish patients with diabetes.

Methods

From the Danish Adult Diabetes Database, we extracted data on 137 893 patient cases admitted to hospitals between 2010 and 2013. Applying multilevel logistic and probit regression models for every year, we quantified general contextual effects of hospital department by the intraclass correlation coefficient (ICC) and the area under the receiver operating characteristic curve (AUC) values. We evaluated the specific effect of hospital accreditation using the ORs and the change in the department variance.

Results

In 2010, the department context had considerable influence on adherence with albuminuria measurement (ICC=21.8%, AUC=0.770), but the general effect attenuated along with the implementation of the national accreditation programme. The ICC value was 16.5% in 2013 and the rate of compliance with albuminuria measurement increased from 91.6% in 2010 to 96% in 2013.

Conclusions

Parallel to implementation of the national accreditation programme, departments’ compliance with the standard benchmark for albuminuria measurement increased and the ICC values decreased, but remained high. While those results indicate an overall quality improvement, further intervention focusing on departments with the lowest compliance could be considered.

General contextual effects on neglected tropical disease risk in rural Kenya

The neglected tropical diseases (NTDs) are characterized by their tendency to cluster within groups of people, typically the poorest and most marginalized. Despite this, measures of clustering, such as within-group correlation or between-group heterogeneity, are rarely reported from community-based studies of NTD risk. We describe a general contextual analysis that uses multi-level models to partition and quantify variation in individual NTD risk at multiple grouping levels in rural Kenya. The importance of general contextual effects (GCE) in structuring variation in individual infection with Schistosoma mansoni, the soil-transmitted helminths, Taenia species, and Entamoeba histolytica/dispar was examined at the household-, sublocation- and constituency-levels using variance partition/intra-class correlation co-efficients and median odds ratios. These were compared with GCE for HIV, Plasmodium falciparum and Mycobacterium tuberculosis. The role of place of residence in shaping infection risk was further assessed using the spatial scan statistic. Individuals from the same household showed correlation in infection for all pathogens, and this was consistently highest for the gastrointestinal helminths. The lowest levels of household clustering were observed for E. histolytica/dispar, P. falciparum and M. tuberculosis. Substantial heterogeneity in individual infection risk was observed between sublocations for S. mansoni and Taenia solium cysticercosis and between constituencies for infection with S. mansoni, Trichuris trichiura and Ascaris lumbricoides. Large overlapping spatial clusters were detected for S. mansoni, T. trichiura, A. lumbricoides, and Taenia spp., which overlapped a large cluster of elevated HIV risk. Important place-based heterogeneities in infection risk exist in this community, and these GCEs are greater for the NTDs and HIV than for TB and malaria. Our findings suggest that broad-scale contextual drivers shape infectious disease risk in this population, but these effects operate at different grouping-levels for different pathogens. A general contextual analysis can provide a foundation for understanding the complex ecology of NTDs and contribute to the targeting of interventions.

Non-completion of secondary education and early disability in Norway: geographic patterns, individual and community risks

Background

School non-completion and early work disability is a great public health challenge in Norway, as in most western countries. This study aims to investigate how medically based disability pension (DP) among young adults varies geographically and how municipal socioeconomic conditions interact with non-completion of secondary education in determining DP risk.

Methods

The study includes a nationally representative sample of 30% of all Norwegians (N = 350,699) aged 21–40 in 2010 from Statistic Norway’s population registries. Multilevel models incorporating factors at the individual, neighbourhood and municipal levels were applied to estimate the neighbourhood and municipality general contextual effects in DP receipt, and detect possible differences in the impact of municipal socioeconomic conditions on DP risk between completers and non-completers of secondary education.

Results

A pattern of spatial clustering at the neighbourhood (ICC = 0.124) and municipality (ICC = 0.021) levels are clearly evident, indicating that the underlying causes of DP receipt have a systematic neighbourhood and municipality variation in Norway. Non-completion of secondary education is strongly correlated with DP receipt among those younger than 40. Socioeconomic characteristics of the municipality are also significantly correlated with DP risk, but these associations are conditioned by the completion of secondary education. Living in a socioeconomically advantageous municipality (i.e. high income, high education levels and low unemployment and social security payment rates) is associated with a higher risk of DP, but only among those who do not complete their secondary education. Although the proportion of DPs was equal in rural and urban areas, it is evident that young people living in urban settings are more at risk of early DP than their counterparts living in rural parts of the country when controlling for other risk factors.

Conclusion

The association between school non-completion and DP risk varies between municipalities and local socioeconomic environments. The interplay between personal characteristics and the local community is important in DP risk among young adults, implying that preventive measures should be directed not only at the individual level, but also include the educational system and the local community.

The role of the clinical departments for understanding patient heterogeneity in one-year mortality after a diagnosis of heart failure: A multilevel analysis of individual heterogeneity for profiling provider outcomes

Purpose

To evaluate the general contextual effect (GCE) of the hospital department on one-year mortality in Swedish and Danish patients with heart failure (HF) by applying a multilevel analysis of individual heterogeneity.

Methods

Using the Swedish patient register, we obtained data on 36,943 patients who were 45–80 years old and admitted for HF to the hospital between 2007 and 2009. From the Danish Heart Failure Database (DHFD), we obtained data on 12,001 patients with incident HF who were 18 years or older and treated at hospitals between June 2010 and June2013. For each year, we applied two-step single and multilevel logistic regression models. We evaluated the general effects of the department by quantifying the intra-class correlation coefficient (ICC) and the increment in the area under the receiver operating characteristic curve (AUC) obtained by adding the random effects of the department in a multilevel logistic regression analysis.

Results

One-year mortality for Danish incident HF patients was low in the three audit years (around 11.1% -13.1%) and departments performed homogeneously (ICC ≈1.5% - 3.5%). The discriminatory accuracy of a model including age and gender was rather high (AUC≈ 0.71–0.73) but the increment in AUC after adding the department random effects into these models was only about 0.011–0.022 units in the three years.

One-year mortality in Swedish patients with first hospitalization for heart failure, was relatively higher for 2007–2009 (≈21.3% - 22%) and departments performed homogeneously (ICC ≈ 1.5% - 3%). The discriminatory accuracy of a model including age, gender and patient risk score was rather high (AUC≈ 0.726–0.728) but the increment in AUC after adding the department random effects was only about 0.010–0.017 units in the three years.

Conclusion

Using the DHFD standard benchmark for one-year mortality, Danish departments had a good, homogeneous performance. In reference to literature, Swedish departments had a homogeneous performance and the mortality rates for patients with first hospitalization for heart failure were similar to those reported since 2000. Considering this, if health authorities decide to further reduce mortality rates, a comprehensive quality strategy should focus on all Swedish hospitals. Yet, a complementary assessment for the period after the study period is required to confirm whether department performance is still homogeneous or not to determine the most appropriate action.

Clinic variation in glycaemic control for children with Type 1 diabetes in England and Wales: a population-based, multilevel analysis

AIM

To understand the scope for improving children's glycaemic outcomes by reducing variation between clinics and examine the role of insulin regimen and clinic characteristics.

METHODS

Cross-sectional analysis of 2012-2013 National Paediatric Diabetes Audit data from 21 773 children aged < 19 years with Type 1 diabetes cared for at 176 clinics organized into 11 regional diabetes networks in England and Wales. Variation in HbA_1c was explored by multilevel models with a random effect for clinic. The impact of clinic context was quantified by computing the per cent of total variation in HbA_1c which occurs between clinics (intraclass correlation coefficient; ICC).

RESULTS

Overall, 69 of the 176 diabetes clinics (39%) had a glycaemic performance that differed significantly from the national average after adjusting for patient case-mix with respect to age, gender, diabetes duration, deprivation and ethnicity. However, differences between clinics accounted for 4.7% of the total variation in HbA_1c . Inclusion of within-clinic HbA_1c standard deviation led to a substantial reduction in ICC to 2.4%. Insulin regimen, clinic volume and diabetes networks had a small or moderate impact on ICC.

CONCLUSIONS

Differences between diabetes clinics accounted for only a small portion of the total variation in glycaemic control because most of the variation was within clinics. This implies that national glycaemic improvements might best be achieved not only by targeting poor centres but also by shifting the whole distribution of clinics to higher levels of quality.

The tyranny of the averages and the indiscriminate use of risk factors in public health: The case of coronary heart disease

Modern medicine is overwhelmed by a plethora of both established risk factors and novel biomarkers for diseases. The majority of this information is expressed by probabilistic measures of association such as the odds ratio (OR) obtained by calculating differences in average “risk” between exposed and unexposed groups. However, recent research demonstrates that even ORs of considerable magnitude are insufficient for assessing the ability of risk factors or biomarkers to distinguish the individuals who will develop the disease from those who will not. In regards to coronary heart disease (CHD), we already know that novel biomarkers add very little to the discriminatory accuracy (DA) of traditional risk factors. However, the value added by traditional risk factors alongside simple demographic variables such as age and sex has been the subject of less discussion. Moreover, in public health, we use the OR to calculate the population attributable fraction (PAF), although this measure fails to consider the DA of the risk factor it represents. Therefore, focusing on CHD and applying measures of DA, we re-examine the role of individual demographic characteristics, risk factors, novel biomarkers and PAFs in public health and epidemiology. In so doing, we also raise a more general criticism of the traditional risk factors’ epidemiology. We investigated a cohort of 6103 men and women who participated in the baseline (1991–1996) of the Malmö Diet and Cancer study and were followed for 18 years. We found that neither traditional risk factors nor biomarkers substantially improved the DA obtained by models considering only age and sex. We concluded that the PAF measure provided insufficient information for the planning of preventive strategies in the population. We need a better understanding of the individual heterogeneity around the averages and, thereby, a fundamental change in the way we interpret risk factors in public health and epidemiology.

•

There is a plethora of differences in “average” risk between exposed and unexposed groups of individuals.

•

Individual heterogeneity around average values is seldom considered in Public Health.

•

Measures of discriminatory accuracy (DA) informs on the underlying individual heterogeneity.

•

Most know risk factors and other categorizations associated with diseases have low DA.

•

We need a fundamental change in the way we investigate risk factors and other categorizations in Public Health.

Intermediate and advanced topics in multilevel logistic regression analysis

Multilevel data occur frequently in health services, population and public health, and epidemiologic research. In such research, binary outcomes are common. Multilevel logistic regression models allow one to account for the clustering of subjects within clusters of higher‐level units when estimating the effect of subject and cluster characteristics on subject outcomes. A search of the PubMed database demonstrated that the use of multilevel or hierarchical regression models is increasing rapidly. However, our impression is that many analysts simply use multilevel regression models to account for the nuisance of within‐cluster homogeneity that is induced by clustering. In this article, we describe a suite of analyses that can complement the fitting of multilevel logistic regression models. These ancillary analyses permit analysts to estimate the marginal or population‐average effect of covariates measured at the subject and cluster level, in contrast to the within‐cluster or cluster‐specific effects arising from the original multilevel logistic regression model. We describe the interval odds ratio and the proportion of opposed odds ratios, which are summary measures of effect for cluster‐level covariates. We describe the variance partition coefficient and the median odds ratio which are measures of components of variance and heterogeneity in outcomes. These measures allow one to quantify the magnitude of the general contextual effect. We describe an R² measure that allows analysts to quantify the proportion of variation explained by different multilevel logistic regression models. We illustrate the application and interpretation of these measures by analyzing mortality in patients hospitalized with a diagnosis of acute myocardial infarction. © 2017 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

Do family and neighbourhood matter in secondary school completion? A multilevel study of determinants and their interactions in a life-course perspective

BACKGROUND

Completion of secondary education is important for individuals' future health and health behaviour. The fundamental purpose of this study is to investigate the variation and clustering of school completion in families and neighbourhoods. Secondly, we aim to examine the impact of individuals' family structure and neighbourhood of residence and examine to what extent parental education level moderates these associations.

METHODS

Longitudinal register data for 30% of the entire Norwegian population aged 21-27 years in 2010 (N = 107,003) was extracted from Statistic Norway´s event database. Three-level logistic regression models, which incorporated individual, family, and neighbourhood contextual factors, were applied to estimate the family and neighbourhood general contextual effects and detect possible educational differences in the impact of family structure and urban place of residence in school completion.

RESULTS

Completion rates were significantly higher within families with higher education level (79% in tertiary educated families vs. 61% and 48% in secondary and primary educated families respectively) and were strongly correlated within families (ICC = 39.6) and neighbourhoods (ICC = 5.7). Several structural factors at the family level negatively associated with school completion (e.g., family disruption, large family size, and young maternal age) were more prevalent and displayed more negative impact among primary educated individuals. Urban residence was associated with school completion, but only among the tertiary educated.

CONCLUSIONS

Investment in the resources in the individuals' immediate surroundings, including family and neighbourhood, may address a substantial portion of the social inequalities in the completion of upper secondary education. The high intra-familial correlation in school completion suggests that public health policies and future research should acknowledge family environments in order to improve secondary education completion rates among young people within lower educated families.

Intersectionality and risk for ischemic heart disease in Sweden: Categorical and anti-categorical approaches

Intersectionality theory can contribute to epidemiology and public health by furthering understanding of power dynamics driving production of health disparities, and increasing knowledge about heterogeneities within, and overlap between, social categories. Drawing on McCall, we relate the first of these potential contributions to categorical intersectionality and the second to anti-categorical intersectionality. Both approaches are used in study of risk of ischemic heart disease (IHD), based on register data on 3.6 million adults residing in Sweden by 2010, followed for three years. Categorical intersectionality is here coupled with between-group differences in average risk calculation, as we use intersectional categorizations while estimating odds ratios through logistic regressions. The anti-categorical approach is operationalized through measurement of discriminatory accuracy (DA), i.e., capacity to accurately categorize individuals with or without a certain outcome, through computation of the area under the curve (AUC). Our results show substantial differences in average risk between intersectional groupings. The DA of social categorizations is found to be low, however, due to outcome variability within and overlap between categories. We argue that measures of DA should be used for proper interpretation of differences in average risk between social (or any other) categories. Tension between average between-group risk and the DA of categorizations, which can be related to categorical and anti-categorical intersectional analyses, should be made explicit and discussed to a larger degree in epidemiology and public health.

Contextual generalized trust and immunization against the 2009 A(H1N1) pandemic in the American states: A multilevel approach

The aim of the study was to investigate the association between contextual generalized trust and individual-level 2009 A(H1N1) pandemic immunization acceptance. A second aim was to investigate whether knowledge about the A(H1N1) pandemic mediated the association between contextual generalized trust and A(H1N1) immunization acceptance. Data from the National 2009 H1N1 Flu Survey was used. To capture contextual generalized trust, data comes from an aggregation of surveys measuring generalized trust in the American states. To investigate the association between contextual generalized trust and immunization acceptance, while taking potential individual-level confounders into account, multilevel logistic regression was used. The investigation showed contextual generalized trust to be significantly associated with immunization acceptance. However, controlling for knowledge about the A(H1N1) pandemic did not substantially affect the association between contextual generalized trust and immunization acceptance. In conclusion, contextual state-level generalized trust was associated with A(H1N1) immunization, but knowledge about A(H1N1) was not mediating this association.

Appropriate assessment of ethnic differences in adolescent use of psychotropic medication: multilevel analysis of discriminatory accuracy

OBJECTIVE

In the present study, we used a multilevel approach to investigate the role of maternal country of birth (MCOB) in predicting adolescent use of psychotropic medication in Sweden.

DESIGN

Using the Swedish Medical Birth Register we identified all 428,314 adolescents born between 1987 and 1990 and who were residing in Sweden in the year they turned 18. We applied multilevel logistic regression analysis with adolescents (level 1) nested within MCOBs (level 2). Measures of association (odds ratio) and measures of variance (intra-class correlation (ICC)) were calculated, as well as the discriminatory accuracy by calculating the area under the Receiver Operator Characteristic (AU-ROC) curve.

RESULTS

In comparison with adolescents with Swedish-born mothers, adolescents with mothers born in upper-middle, lower-middle and low-income countries were less likely to use psychotropic medication. However, the variance between MCOBs was small (ICC = 2.5 in the final model) relative to the variation within MCOBs. This was confirmed by an AU-ROC value of 0.598.

CONCLUSIONS

Even though we found associations between MCOB and adolescent use of psychotropic medication, the small ICC and AU-ROC indicate that MCOB appears to be an inaccurate context for discriminating adolescent use of psychotropic medication in Sweden.

The median hazard ratio: a useful measure of variance and general contextual effects in multilevel survival analysis

Multilevel data occurs frequently in many research areas like health services research and epidemiology. A suitable way to analyze such data is through the use of multilevel regression models (MLRM). MLRM incorporate cluster‐specific random effects which allow one to partition the total individual variance into between‐cluster variation and between‐individual variation. Statistically, MLRM account for the dependency of the data within clusters and provide correct estimates of uncertainty around regression coefficients. Substantively, the magnitude of the effect of clustering provides a measure of the General Contextual Effect (GCE). When outcomes are binary, the GCE can also be quantified by measures of heterogeneity like the Median Odds Ratio (MOR) calculated from a multilevel logistic regression model. Time‐to‐event outcomes within a multilevel structure occur commonly in epidemiological and medical research. However, the Median Hazard Ratio (MHR) that corresponds to the MOR in multilevel (i.e., ‘frailty’) Cox proportional hazards regression is rarely used. Analogously to the MOR, the MHR is the median relative change in the hazard of the occurrence of the outcome when comparing identical subjects from two randomly selected different clusters that are ordered by risk. We illustrate the application and interpretation of the MHR in a case study analyzing the hazard of mortality in patients hospitalized for acute myocardial infarction at hospitals in Ontario, Canada. We provide R code for computing the MHR. The MHR is a useful and intuitive measure for expressing cluster heterogeneity in the outcome and, thereby, estimating general contextual effects in multilevel survival analysis. © 2016 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

An Original Stepwise Multilevel Logistic Regression Analysis of Discriminatory Accuracy: The Case of Neighbourhoods and Health

BACKGROUND AND AIM

Many multilevel logistic regression analyses of "neighbourhood and health" focus on interpreting measures of associations (e.g., odds ratio, OR). In contrast, multilevel analysis of variance is rarely considered. We propose an original stepwise analytical approach that distinguishes between "specific" (measures of association) and "general" (measures of variance) contextual effects. Performing two empirical examples we illustrate the methodology, interpret the results and discuss the implications of this kind of analysis in public health.

METHODS

We analyse 43,291 individuals residing in 218 neighbourhoods in the city of Malmö, Sweden in 2006. We study two individual outcomes (psychotropic drug use and choice of private vs. public general practitioner, GP) for which the relative importance of neighbourhood as a source of individual variation differs substantially. In Step 1 of the analysis, we evaluate the OR and the area under the receiver operating characteristic (AUC) curve for individual-level covariates (i.e., age, sex and individual low income). In Step 2, we assess general contextual effects using the AUC. Finally, in Step 3 the OR for a specific neighbourhood characteristic (i.e., neighbourhood income) is interpreted jointly with the proportional change in variance (i.e., PCV) and the proportion of ORs in the opposite direction (POOR) statistics.

RESULTS

For both outcomes, information on individual characteristics (Step 1) provide a low discriminatory accuracy (AUC = 0.616 for psychotropic drugs; = 0.600 for choosing a private GP). Accounting for neighbourhood of residence (Step 2) only improved the AUC for choosing a private GP (+0.295 units). High neighbourhood income (Step 3) was strongly associated to choosing a private GP (OR = 3.50) but the PCV was only 11% and the POOR 33%.

CONCLUSION

Applying an innovative stepwise multilevel analysis, we observed that, in Malmö, the neighbourhood context per se had a negligible influence on individual use of psychotropic drugs, but appears to strongly condition individual choice of a private GP. However, the latter was only modestly explained by the socioeconomic circumstances of the neighbourhoods. Our analyses are based on real data and provide useful information for understanding neighbourhood level influences in general and on individual use of psychotropic drugs and choice of GP in particular. However, our primary aim is to illustrate how to perform and interpret a multilevel analysis of individual heterogeneity in social epidemiology and public health. Our study shows that neighbourhood "effects" are not properly quantified by reporting differences between neighbourhood averages but rather by measuring the share of the individual heterogeneity that exists at the neighbourhood level.

Neighbourhood effects on body constitution-A case study of Hong Kong

Traditional Chinese Medicine (TCM) has long perceived environment as an integral part of the development of body constitution, which is a personal state of health closely related to disease presence. Despite of the ever-growing studies on the clinical effectiveness of TCM and the scientific linking between body constitution and diseases, the geographical influence on body constitution has yet remained an unexplored territory. This study sought to investigate whether the neighbourhood environment is relevant to the composition of body type of a population through statistical multilevel and Geographic Information Systems modelling. The analysis comprised 3277 participants who had completed their body type assessment between 2009 and 2012 inclusive. The multilevel analysis also took simultaneous accounts of both individual-level (gender, age, BMI, type of housing) and area-level (percent greenery, percent road surface, total road intersection, sky view factor, temperature, relative humidity, rainfall and social deprivation index) characteristics to explain geographical variation by body types. Significant random or place effects (p < 0.001) were identified in the multilevel models. The spatial variation of body constitution involved the dynamic interplay between individual and environmental factors. The findings amassed the first scientific indications to back the common belief that place does play a role in the development of body constitution and is worthy of further investigation. By considering spatial and personal attributes simultaneously, the study can yield valuable insights into the patterning of area variation in body constitution and disease presence.

Short Term Survival after Admission for Heart Failure in Sweden: Applying Multilevel Analyses of Discriminatory Accuracy to Evaluate Institutional Performance

BACKGROUND

Hospital performance is frequently evaluated by analyzing differences between hospital averages in some quality indicators. The results are often expressed as quality charts of hospital variance (e.g., league tables, funnel plots). However, those analyses seldom consider patients heterogeneity around averages, which is of fundamental relevance for a correct evaluation. Therefore, we apply an innovative methodology based on measures of components of variance and discriminatory accuracy to analyze 30-day mortality after hospital discharge with a diagnosis of Heart Failure (HF) in Sweden.

METHODS

We analyzed 36,943 patients aged 45-80 treated in 565 wards at 71 hospitals during 2007-2009. We applied single and multilevel logistic regression analyses to calculate the odds ratios and the area under the receiver-operating characteristic (AUC). We evaluated general hospital and ward effects by quantifying the intra-class correlation coefficient (ICC) and the increment in the AUC obtained by adding random effects in a multilevel regression analysis (MLRA). Finally, the Odds Ratios (ORs) for specific ward and hospital characteristics were interpreted jointly with the proportional change in variance (PCV) and the proportion of ORs in the opposite direction (POOR).

FINDINGS

Overall, the average 30-day mortality was 9%. Using only patient information on age and previous hospitalizations for different diseases we obtained an AUC = 0.727. This value was almost unchanged when adding sex, country of birth as well as hospitals and wards levels. Average mortality was higher in small wards and municipal hospitals but the POOR values were 15% and 16% respectively.

CONCLUSIONS

Swedish wards and hospitals in general performed homogeneously well, resulting in a low 30-day mortality rate after HF. In our study, knowledge on a patient's previous hospitalizations was the best predictor of 30-day mortality, and this information did not improve by knowing the sex and country of birth of the patient or where the patient was treated.

Maternal education and perinatal outcomes among Spanish women residing in southern Spain (2001-2011)

Evidence suggests that educational differences in perinatal outcomes have increased in some countries (Eastern Europe) while remained stable in others (Scandinavian countries). However, less is known about the experience of Southern Europe. This study aims to evaluate the association between maternal education and perinatal outcomes derived from birthweight (low birthweight and macrosomia) and gestational age (pre-term and post-term births) among Spaniards living in the Autonomous Community of Andalusia during the period 2001-2011 (around 19 % of births in Spain); and to evaluate whether the educational differences narrowed or widened during that period, which includes both an economic boom (2001-2008) and the global economic crisis (2009-2011). This study uses the Andalusian Population Longitudinal Database and the Vital Statistics Data provided by the Spanish National Statistics Institute. We study live and singleton births of Spanish mothers who lived in Andalusia at the time of delivery (n = 404,951). ORs with 95 % confidence intervals (crude and adjusted) were estimated using multinomial regression models. A negative educational gradient is observed in all perinatal outcomes studied (i.e., the higher the educational status, the lower the risk of negative perinatal outcomes). However, when disaggregating the sample in two periods, the gradient is only statistically significant for pre-term birth during 2001-2008, while a full gradient is observed in all perinatal indicators in the period 2009-2011 with an increase in the educational inequalities in macrosomia and post-term. Further studies are needed in order to confirm whether there is a causal association between the widening of the educational differences in perinatal outcomes and the onset of the economic crisis in Spain, or the widening can be explained by other factors, such as changes in childbearing patterns and the composition of women accessing motherhood.

Mortality and work disability in a cohort of Norwegian couples--the HUNT study

BACKGROUND

Poor health is clustered in families, and partners might influence each other. We studied possible consequences of living with a spouse with poor health or unhealthy lifestyle on mortality and work disability.

METHODS

In total, 18 943 couples from the HUNT2 Study (1995-97) were linked to national registries and followed until December 2007, identifying deaths and disability pension retirements. Couple's mean exposures were included together with the individual's deviation from the couple mean in discrete time multilevel logistic regression.

RESULTS

There was weak evidence of associations between partner's health and risk of dying. Associations between couples slightly exceeded associations within couples for smoking [odds ratio (OR) within 1.57 (95% confidence interval (CI): 1.38-1.78); OR between 1.88 (95% CI: 1.70-2.08), P value for difference 0.027] and education [OR within 1.07 (95% CI: 0.99-1.15); OR between 1.17 (1.11-1.23), P value for difference 0.065]. Indicators of partner's health, such as self-rated health [OR within 3.17 (95% CI: 2.80-3.58); OR between 3.92 (95% CI: 3.50-4.40), P value for difference 0.014], insomnia [OR within 1.39 (95% CI: 1.18-1.64); OR between 2.11 (95% CI: 1.86-2.53), P value for difference <0.001] and symptoms of depression [OR within 1.45 (95% CI: 1.22-1.71); OR between 1.98 (95% CI: 1.69-2.31) P value for difference 0.009] were, however, associated with risk of work disability. Self-rated health and symptoms displayed stronger associations with work disability among partners than reported somatic diseases.

CONCLUSIONS

This study did not indicate strong consequences of living with a spouse with poor health or unhealthy lifestyle on mortality. It did, however, indicate associations of partner's health with work disability.

Does Maternal Country of Birth Matter for Understanding Offspring's Birthweight? A Multilevel Analysis of Individual Heterogeneity in Sweden

Background

Many public health and epidemiological studies have found differences between populations (e.g. maternal countries of birth) in average values of a health indicator (e.g. mean offspring birthweight). However, the approach based solely on population-level averages compromises our understanding of variability in individuals’ health around the averages. If this variability is high, the exclusive study of averages may give misleading information. This idea is relevant when investigating country of birth differences in health.

Methods and Results

To exemplify this concept, we use information from the Swedish Medical Birth Register (2002–2010) and apply multilevel regression analysis of birthweight, with babies (n = 811,329) at the first, mothers (n = 571,876) at the second, and maternal countries of birth (n = 109) at the third level. We disentangle offspring, maternal and maternal country of birth components of the total offspring heterogeneity in birthweight for babies born within the normal timespan (37–42 weeks). We found that of such birthweight variation about 50% was at the baby level, 47% at the maternal level and only 3% at the maternal countries of birth level.

Conclusion

In spite of seemingly large differences in average birthweight among maternal countries of birth (range 3290–3677g), knowledge of the maternal country of birth does not provide accurate information for ascertaining individual offspring birthweight because of the high inter-offspring heterogeneity around country averages. Our study exemplifies the need for a better understanding of individual health diversity for which group averages may provide insufficient and even misleading information. The analytical approach we outline is therefore relevant to investigations of country of birth (and ethnic) differences in health in general.

Comparing multilevel and Bayesian spatial random effects survival models to assess geographical inequalities in colorectal cancer survival: a case study

BACKGROUND

Multilevel and spatial models are being increasingly used to obtain substantive information on area-level inequalities in cancer survival. Multilevel models assume independent geographical areas, whereas spatial models explicitly incorporate geographical correlation, often via a conditional autoregressive prior. However the relative merits of these methods for large population-based studies have not been explored. Using a case-study approach, we report on the implications of using multilevel and spatial survival models to study geographical inequalities in all-cause survival.

METHODS

Multilevel discrete-time and Bayesian spatial survival models were used to study geographical inequalities in all-cause survival for a population-based colorectal cancer cohort of 22,727 cases aged 20-84 years diagnosed during 1997-2007 from Queensland, Australia.

RESULTS

Both approaches were viable on this large dataset, and produced similar estimates of the fixed effects. After adding area-level covariates, the between-area variability in survival using multilevel discrete-time models was no longer significant. Spatial inequalities in survival were also markedly reduced after adjusting for aggregated area-level covariates. Only the multilevel approach however, provided an estimation of the contribution of geographical variation to the total variation in survival between individual patients.

CONCLUSIONS

With little difference observed between the two approaches in the estimation of fixed effects, multilevel models should be favored if there is a clear hierarchical data structure and measuring the independent impact of individual- and area-level effects on survival differences is of primary interest. Bayesian spatial analyses may be preferred if spatial correlation between areas is important and if the priority is to assess small-area variations in survival and map spatial patterns. Both approaches can be readily fitted to geographically enabled survival data from international settings.

Invited commentary: multilevel analysis of individual heterogeneity-a fundamental critique of the current probabilistic risk factor epidemiology

In this issue of the Journal, Dundas et al. (Am J Epidemiol. 2014;180(2):197-207) apply a hitherto infrequent multilevel analytical approach: multiple membership multiple classification (MMMC) models. Specifically, by adopting a life-course approach, they use a multilevel regression with individuals cross-classified in different contexts (i.e., families, early schools, and neighborhoods) to investigate self-reported health and mental health in adulthood. They provide observational evidence suggesting the relevance of the early family environment for launching public health interventions in childhood in order to improve health in adulthood. In their analyses, the authors distinguish between specific contextual measures (i.e., the association between particular contextual characteristics and individual health) and general contextual measures (i.e., the share of the total interindividual heterogeneity in health that appears at each level). By doing so, they implicitly question the traditional probabilistic risk factor epidemiology including classical "neighborhood effects" studies. In fact, those studies use simple hierarchical structures and disregard the analysis of general contextual measures. The innovative MMMC approach properly responds to the call for a multilevel eco-epidemiology against a widespread probabilistic risk factors epidemiology. The risk factors epidemiology is not only reduced to individual-level analyses, but it also embraces many current "multilevel analyses" that are exclusively focused on analyzing contextual risk factors.

Early-life school, neighborhood, and family influences on adult health: a multilevel cross-classified analysis of the Aberdeen children of the 1950s study

Lifetime exposures to adverse social environments influence adult health, as do exposures in early life. It is usual to examine the influences of school on teenage health and of adult area of residence on adult health. We examined the combined long-term association of the school attended, as well as the area of residence in childhood, with adult health. A total of 6,285 children from Aberdeen, Scotland, who were aged 5–12 years in 1962, were followed up at a mean age of 47 years in 2001. Cross-classified multilevel logistic regression was used to estimate the associations of family, school, and area of residence with self-reported adult health and mental health, adjusting for childhood family-, school-, and neighborhood-level factors, as well as current adult occupational position. Low early-life social position (as determined by the father's occupational level) was associated with poor adult self-rated health but not poor mental health. There were small contextual associations between childhood school environment (median odds ratio = 1.08) and neighborhood environment (median odds ratio = 1.05) and adult self-rated health. The share of the total variance in health at the family level was 10.1% compared with 89.6% at the individual level. Both socioeconomic context and composition in early life appear to have an influence on adult health, even after adjustment for current occupational position.

When does hardship matter for health? Neighborhood and individual disadvantages and functional somatic symptoms from adolescence to mid-life in The Northern Swedish Cohort

A large body of research has shown that health is influenced by disadvantaged living conditions, including both personal and neighborhood conditions. Little is however known to what degree the health impact of different forms of disadvantage differ along the life course. The present study aims to examine when, during the life course, neighborhood and individual disadvantages relate to functional somatic symptoms. Participants (n = 992) came from The Northern Swedish Cohort and followed from age 16, 21, 30 until 42 years. Functional somatic symptoms, socioeconomic disadvantage, and social and material adversity were measured through questionnaires and linked to register data on neighborhood disadvantage. Data was analyzed with longitudinal and cross-sectional multilevel models. Results showed that neighborhood disadvantage, social and material adversity and gender all contributed independently to overall levels of symptoms across the life course. Cross-sectional analyses also suggested that the impact of disadvantage differed between life course periods; neighborhood disadvantage was most important in young adulthood, and the relative importance of material versus social adversity increased as participants grew older. In summary, the study suggests that disadvantages from different contextual sources may affect functional somatic health across the life course, but also through life course specific patterns.

The impact of neighbourhood deprivation on adolescent violent criminality and substance misuse: a longitudinal, quasi-experimental study of the total Swedish population

BACKGROUND

A number of studies suggest associations between neighbourhood characteristics and criminality during adolescence and young adulthood. However, the causality of such neighbourhood effects remains uncertain.

METHODS

We followed all children born in Sweden from 1975-1989 who lived in its three largest cities by the age of 15 years and for whom complete information was available about individual and contextual factors (N = 303 465). All biological siblings were identified in the sample (N = 179 099). Generalized linear mixed-effects models were used to assess the effect of neighbourhood deprivation on violent criminality and substance misuse between the ages of 15 and 20 years, while taking into account the cross-classified data structure (i.e. siblings in the same families attending different schools and living in different neighbourhoods at age 15).

RESULTS

In the crude model, an increase of 1 SD in neighbourhood deprivation was associated with a 57% increase in the odds of being convicted of a violent crime (95% CI 52%-63%). The effect was greatly attenuated when adjustment was made for a number of observed confounders (OR 1.09, 95% CI 1.06-1.11). When we additionally adjusted for unobserved familial confounders, the effect was no longer present (OR 0.96, 95% CI 0.84-1.10). Similar results were observed for substance misuse. The results were not due to poor variability either between neighbourhoods or within families.

CONCLUSIONS

We found that the adverse effect of neighbourhood deprivation on adolescent violent criminality and substance misuse in Sweden was not consistent with a causal inference. Instead, our findings highlight the need to control for familial confounding in multilevel studies of criminality and substance misuse.