Flora I. Matheson

A simulation study of sample size for multilevel logistic regression models

BackgroundMany studies conducted in health and social sciences collect individual level data as outcome measures. Usually, such data have a hierarchical structure, with patients clustered within physicians, and physicians clustered within practices. Large survey data, including national surveys, have a hierarchical or clustered structure; respondents are naturally clustered in geographical units (e.g., health regions) and may be grouped into smaller units. Outcomes of interest in many fields not only reflect continuous measures, but also binary outcomes such as depression, presence or absence of a disease, and self-reported general health. In the framework of multilevel studies an important problem is calculating an adequate sample size that generates unbiased and accurate estimates.MethodsIn this paper simulation studies are used to assess the effect of varying sample size at both the individual and group level on the accuracy of the estimates of the parameters and variance components of multilevel logistic regression models. In addition, the influence of prevalence of the outcome and the intra-class correlation coefficient (ICC) is examined.ResultsThe results show that the estimates of the fixed effect parameters are unbiased for 100 groups with group size of 50 or higher. The estimates of the variance covariance components are slightly biased even with 100 groups and group size of 50. The biases for both fixed and random effects are severe for group size of 5. The standard errors for fixed effect parameters are unbiased while for variance covariance components are underestimated. Results suggest that low prevalent events require larger sample sizes with at least a minimum of 100 groups and 50 individuals per group.ConclusionWe recommend using a minimum group size of 50 with at least 50 groups to produce valid estimates for multi-level logistic regression models. Group size should be adjusted under conditions where the prevalence of events is low such that the expected number of events in each group should be greater than one.

Density, Destinations or Both? A Comparison of Measures of Walkability in Relation to Transportation Behaviors, Obesity and Diabetes in Toronto, Canada

The design of suburban communities encourages car dependency and discourages walking, characteristics that have been implicated in the rise of obesity. Walkability measures have been developed to capture these features of urban built environments. Our objective was to examine the individual and combined associations of residential density and the presence of walkable destinations, two of the most commonly used and potentially modifiable components of walkability measures, with transportation, overweight, obesity, and diabetes. We examined associations between a previously published walkability measure and transportation behaviors and health outcomes in Toronto, Canada, a city of 2.6 million people in 2011. Data sources included the Canada census, a transportation survey, a national health survey and a validated administrative diabetes database. We depicted interactions between residential density and the availability of walkable destinations graphically and examined them statistically using general linear modeling. Individuals living in more walkable areas were more than twice as likely to walk, bicycle or use public transit and were significantly less likely to drive or own a vehicle compared with those living in less walkable areas. Individuals in less walkable areas were up to one-third more likely to be obese or to have diabetes. Residential density and the availability of walkable destinations were each significantly associated with transportation and health outcomes. The combination of high levels of both measures was associated with the highest levels of walking or bicycling (p<0.0001) and public transit use (p<0.0026) and the lowest levels of automobile trips (p<0.0001), and diabetes prevalence (p<0.0001). We conclude that both residential density and the availability of walkable destinations are good measures of urban walkability and can be recommended for use by policy-makers, planners and public health officials. In our setting, the combination of both factors provided additional explanatory power.

Unwalkable Neighborhoods, Poverty, and the Risk of Diabetes Among Recent Immigrants to Canada Compared With Long-Term Residents

OBJECTIVE This study was designed to examine whether residents living in neighborhoods that are less conducive to walking or other physical activities are more likely to develop diabetes and, if so, whether recent immigrants are particularly susceptible to such effects. RESEARCH DESIGN AND METHODS We conducted a population-based, retrospective cohort study to assess the impact of neighborhood walkability on diabetes incidence among recent immigrants (n = 214,882) relative to long-term residents (n = 1,024,380). Adults aged 30–64 years who were free of diabetes and living in Toronto, Canada, on 31 March 2005 were identified from administrative health databases and followed until 31 March 2010 for the development of diabetes, using a validated algorithm. Neighborhood characteristics, including walkability and income, were derived from the Canadian Census and other sources. RESULTS Neighborhood walkability was a strong predictor of diabetes incidence independent of age and area income, particularly among recent immigrants (lowest [quintile 1 {Q1}] vs. highest [quintile 5 {Q5}] walkability quintile: relative risk [RR] 1.58 [95% CI 1.42–1.75] for men; 1.67 [1.48–1.88] for women) compared with long-term residents (Q1 to Q5) 1.32 [1.26–1.38] for men; 1.24 [1.18–1.31] for women). Coexisting poverty accentuated these effects; diabetes incidence varied threefold between recent immigrants living in low-income/low walkability areas (16.2 per 1,000) and those living in high-income/high walkability areas (5.1 per 1,000). CONCLUSIONS Neighborhood walkability was inversely associated with the development of diabetes in our setting, particularly among recent immigrants living in low-income areas.

Geographic Methods for Understanding and Responding to Disparities in Mammography Use in Toronto, Canada

OBJECTIVE: To use spatial and epidemiologic analyses to understand disparities in mammaography use and to formulate interventions to increase its uptake in low-income, high-recent immigration areas in Toronto, Canada.DESIGN: We compared mammography rates in four income-immigration census tract groups. Data were obtained from the 1996 Canadian census and 2000 physician billing claims. Risk ratios, linear regression, multilayer maps, and spatial analysis were used to examine utilization by area for women age 45 to 64 years.SETTING: Residential population of inner city Toronto, Canada, with a 1996 population of 780,000.PARTICIPANTS: Women age 45 to 64 residing in Toronto’s inner city in the year 2000.MEASUREMENTS AND MAIN RESULTS: Among 113,762 women age 45 to 64, 27,435 (24%) had received a mammogram during 2000 and 91,542 (80%) had seen a physician. Only 21% of women had a mammogram in the least advantaged group (low income-high immigration), compared with 27% in the most advantaged group (high income-low immigration) (risk ratio, 0.79; 95% confidence interval, 0.75 to 0.84). Multilayer maps demonstrated a low income-high immigration band running through Toronto’s inner city and low mammography rates within that band. There was substantial geographic clustering of study variables.CONCLUSIONS: We found marked variation in mammography rates by area, with the lowest rates associated with low income and high immigration. Spatial patterns identified areas with low mammography and low physician visit rates appropriate for outreach and public education interventions. We also identified areas with low mammography and high physician visit rates appropriate for interventions targeted at physicians.

Association of Neighborhood Walkability With Change in Overweight, Obesity, and Diabetes

IMPORTANCE Rates of obesity and diabetes have increased substantially in recent decades; however, the potential role of the built environment in mitigating these trends is unclear. OBJECTIVE To examine whether walkable urban neighborhoods are associated with a slower increase in overweight, obesity, and diabetes than less walkable ones. DESIGN, SETTING, AND PARTICIPANTS Time-series analysis (2001-2012) using annual provincial health care (N ≈ 3 million per year) and biennial Canadian Community Health Survey (N ≈ 5500 per cycle) data for adults (30-64 years) living in Southern Ontario cities. EXPOSURES Neighborhood walkability derived from a validated index, with standardized scores ranging from 0 to 100, with higher scores indicating more walkability. Neighborhoods were ranked and classified into quintiles from lowest (quintile 1) to highest (quintile 5) walkability. MAIN OUTCOMES AND MEASURES Annual prevalence of overweight, obesity, and diabetes incidence, adjusted for age, sex, area income, and ethnicity. RESULTS Among the 8777 neighborhoods included in this study, the median walkability index was 16.8, ranging from 10.1 in quintile 1 to 35.2 in quintile 5. Resident characteristics were generally similar across neighborhoods; however, poverty rates were higher in high- vs low-walkability areas. In 2001, the adjusted prevalence of overweight and obesity was lower in quintile 5 vs quintile 1 (43.3% vs 53.5%; P < .001). Between 2001 and 2012, the prevalence increased in less walkable neighborhoods (absolute change, 5.4% [95% CI, 2.1%-8.8%] in quintile 1, 6.7% [95% CI, 2.3%-11.1%] in quintile 2, and 9.2% [95% CI, 6.2%-12.1%] in quintile 3). The prevalence of overweight and obesity did not significantly change in areas of higher walkability (2.8% [95% CI, -1.4% to 7.0%] in quintile 4 and 2.1% [95% CI, -1.4% to 5.5%] in quintile 5). In 2001, the adjusted diabetes incidence was lower in quintile 5 than other quintiles and declined by 2012 from 7.7 to 6.2 per 1000 persons in quintile 5 (absolute change, -1.5 [95% CI, -2.6 to -0.4]) and 8.7 to 7.6 in quintile 4 (absolute change, -1.1 [95% CI, -2.2 to -0.05]). In contrast, diabetes incidence did not change significantly in less walkable areas (change, -0.65 in quintile 1 [95% CI, -1.65 to 0.39], -0.5 in quintile 2 [95% CI, -1.5 to 0.5], and -0.9 in quintile 3 [95% CI, -1.9 to 0.02]). Rates of walking or cycling and public transit use were significantly higher and that of car use lower in quintile 5 vs quintile 1 at each time point, although daily walking and cycling frequencies increased only modestly from 2001 to 2011 in highly walkable areas. Leisure-time physical activity, diet, and smoking patterns did not vary by walkability (P > .05 for quintile 1 vs quintile 5 for each outcome) and were relatively stable over time. CONCLUSIONS AND RELEVANCE In Ontario, Canada, higher neighborhood walkability was associated with decreased prevalence of overweight and obesity and decreased incidence of diabetes between 2001 and 2012. However, the ecologic nature of these findings and the lack of evidence that more walkable urban neighborhood design was associated with increased physical activity suggest that further research is necessary to assess whether the observed associations are causal.

Neighbourhood chronic stress and gender inequalities in hypertension among Canadian adults: a multilevel analysis

Background A growing body of literature regarding the contextual influences of ‘place’ effects on health increasingly demonstrates that living in neighbourhoods with high levels of deprivation is associated with worse cardiovascular outcomes; however, little research has explored whether neighbourhood deprivation has a differential impact on the cardiovascular health of men and women. The purpose of this study was to explore gender differences in the association between neighbourhood deprivation and the prevalence of hypertension among non-institutionalised Canadian adults. Methods Individual-level data from the Canadian Community Health Survey (2000–2005) were combined with area-level data from the 2001 Canada Census to assess the relationship between gender, neighbourhood deprivation and hypertension using multilevel regression. Results Of the 103 419 respondents, 20 705 reported having hypertension (17.6%). In multilevel models, neighbourhood deprivation was significantly associated with hypertension and this effect remained significant after adjusting for individual-level demographic, socioeconomic and lifestyle characteristics (OR 1.12, 95% CI 1.10 to 1.15). Neighbourhood deprivation appears to be a stronger predictor of hypertension among women, such that women living in areas of high deprivation were 10% more likely to report having hypertension in comparison with men living in the same neighbourhoods and with women living in the least impoverished neighbourhoods. Conclusions Although future research is needed to determine whether interventions at the area-level are effective in reducing inequalities in health outcomes across neighbourhoods, policies aimed at reducing area-level deprivation may have a differential benefit on the cardiovascular health of men and women.

Drinking in context: the influence of gender and neighbourhood deprivation on alcohol consumption

Background Findings from contextual studies have shown that living in both poor and affluent neighbourhoods increases the risk of drinking and drug use, but few studies have examined the connection between neighbourhood context and drinking from a gender perspective. Methods We investigated the association between gender, neighbourhood deprivation and weekly drinking behaviour (number of drinks) in a national sample of 93 457 Canadians using multilevel zero-inflated Poisson regression. A cross-level interaction between gender and neighbourhood deprivation was examined while controlling for other potential risk factors. Results 53% of Canadians reported having at least one drink in the last year (men=61%; women=46%). Among respondents who were drinkers, the average number of drinks per week was 6.4 with male drinkers reporting an average of 7.9 and female drinkers reporting an average of 4.6. Neighbourhood material deprivation was independently associated with weekly drinking. Findings from multilevel analysis showed a u-shaped curve between neighbourhood deprivation and drinking, but only for men. Men living in the poorest neighbourhoods drank more weekly (8.5 drinks) than men living in neighbourhoods of wealthy (4.5 drinks) and mid-range deprivation (3.7 drinks). No difference in drinking by neighbourhood material deprivation was observed among women. Conclusion Men, like women, experience gender-specific health difficulties (eg, alcohol-related problems) suggesting the need for a gendered focus on policies and services related to womens and mens health. The challenge for public health and primary care is to work together to target risk-taking behaviours among men through treatment, prevention and cultural/educational messages aimed at building healthy lifestyles.

A systematic review of randomized controlled trials of interventions to improve the health of persons during imprisonment and in the year after release.

We systematically reviewed randomized controlled trials of interventions to improve the health of people during imprisonment or in the year after release. We searched 14 biomedical and social science databases in 2014, and identified 95 studies. Most studies involved only men or a majority of men (70/83 studies in which gender was specified); only 16 studies focused on adolescents. Most studies were conducted in the United States (n = 57). The risk of bias for outcomes in almost all studies was unclear or high (n = 91). In 59 studies, interventions led to improved mental health, substance use, infectious diseases, or health service utilization outcomes; in 42 of these studies, outcomes were measured in the community after release. Improving the health of people who experience imprisonment requires knowledge generation and knowledge translation, including implementation of effective interventions.

Neighborhood context and infant birthweight among recent immigrant mothers: a multilevel analysis.

OBJECTIVES We compared the influence of the residential environment and maternal country of origin on birthweight and low birthweight of infants born to recent immigrants to urban Ontario. METHODS We linked delivery records (1993-2000) to an immigration database (1993-1995) and small-area census data (1996). The data were analyzed with cross-classified random-effects models and standard multilevel methods. Higher-level predictors included 4 independent measures of neighborhood context constructed by factor analysis and maternal world regions of origin. RESULTS Births (N = 22 189) were distributed across 1396 census tracts and 155 countries of origin. The associations between neighborhood indices and birthweight disappeared after we controlled for the maternal country of origin in a cross-classified multilevel model. Significant associations between world regions and birthweight and low birthweight persisted after we controlled for neighborhood context and individual characteristics. CONCLUSIONS The residential environment has little, if any, influence on birthweight among recent immigrants to Ontario. Country of origin appears to be a much more important factor in low birthweight among children of recent immigrants than current neighborhood. Findings of neighborhood influences among recent immigrants should be interpreted with caution.

Neighbourhood deprivation and regional inequalities in self-reported health among Canadians: Are we equally at risk?

Individual-level data from the Canadian Community Health Survey was combined with area-level data from the 2001 Canada Census to explore the relationship between neighbourhood deprivation and regional inequalities in self-reported health (n=120,290). While neighbourhood deprivation was a significant predictor of fair/poor health in all geographic regions (OR=1.11; 95% CI: 1.08, 1.14), living on the Atlantic and Pacific coasts exacerbated the detrimental effects of neighbourhood deprivation on the perceived health of respondents (OR=1.21; 1.28). By failing to explore regional variations in risk, we could fail to identify areas where provincial policies may interact with neighbourhood factors to reinforce health inequalities amongst deprived communities.