Variables of interest

All relevant psychiatric admissions were aggregated at the dissemination area level, which in Canada corresponds roughly to the size of a neighbourhood and includes 400 to 700 persons [16]. We used the Statistics Canada’s Postal Code Conversion File to link admissions to dissemination areas in the Census data [17].

To obtain data on deprivation, we used information contained in the Ontario Marginalization Index, which explores multiple dimensions of marginalisation, such as residential instability, material deprivation, dependency, and ethnic concentration [18]. In turn, area-level deprivation was ascertained from the material deprivation score, which measures the inability of individuals and communities to access and attain basic material needs [18], and is made up of the following indicators: proportion of the population considered low-income; proportion of the population aged 15 and older who are unemployed; proportion of the population receiving government transfer payments; proportion of the population aged 20 and older without a high school diploma; proportion of households living in dwellings that are in need of major repair; and proportion of families who are lone parent families. This measure was available for 2006 and 2011. We used the 2006 value to derive the deprivation measure for 2006, 2007 and 2008, and the 2011 value to derive the deprivation measure for 2009 and 2010. The 2006 version was estimated using data from both the Canada Census short- and long-form questionnaires.¹ In 2011, the federal government replaced the mandatory long-form census with the National Household Survey, which does not require mandatory reporting. The voluntary nature of this survey introduces the possibility of non-response bias among respondents. Therefore, the 2011 update does not use data from the National Household Survey but instead uses alternative data sources, such as Statistics Canada 2011 Canada Census Profiles data, the Registered Persons Database, the Immigration, Refugees and Citizenship Canada data, Statistics Canada Family Tax Return File and the Municipal Property Assessment Corporation data, to replace indicators formally based on the Census long-form questionnaire (this change in methodology appears to have had minimal impact on the construction of the score; nonetheless, caution should be applied when interpreting the results, namely when examining changes from 2008 to 2009).

We derived a series of variables for each dissemination area, which have been found to be risk factors for SMI-related hospital admissions and potential drivers of inequality [15]. To the extent possible, we tried to employ the same variables used by White and colleagues, which in turn were informed by their literature review (where this was not possible, we employed similar variables known to be associated with psychiatric admissions among individuals with SMI) [15]. We used the Registered Person Database to obtain data on core explanatory variables – total population aged 15 and older, and percentage of males and females by 5-year bands from 15 to 19 to 60–64 and wider age bands thereafter (65–74 and 75+). Demand-side (i.e., need) variables included SMI prevalence per 1000 individuals aged 15 and older (which was calculated using the hospitalisation databases), percentage of individuals identified as immigrants, and percentage of individuals identified as refugees (where these last two variables were determined through the Immigration, Refugee and Citizenship Canada database). Supply-side variables included an indicator of rural residency (where rural communities were defined as those with a population of 10,000 or less through the use of the Postal Code Conversion File), average minimum distance (in kilometres) to an acute care provider for individuals aged 15 and older, average minimum distance (in kilometres) to a mental health care provider for individuals aged 15 and older, general practitioner (GP) density per 1000 individuals aged 15 and older, and psychiatrist density per 1000 individuals 15 and older. The average minimum distance for all the provider variables was estimated using an “as the crow flies” distance method, which calculates great circle distances (in kilometres) from one place to another using latitude and longitude. We used each patient’s postal code and the postal code of the nearest hospital/provider within the regional health authority of residence, and estimated the shortest straight-line distance between the two. GP and psychiatrist densities by regional health authority were estimated using data in the ICES Physician Database.

Table 1 provides the descriptive statistics at the dissemination area level. Our sample included 97,926 admissions distributed across 96,834 dissemination areas. On average, about one SMI-related psychiatric admission occurred in each dissemination area over the study period.

Descriptive statistics at the dissemination area level, April 2006 to March 2011

Legend: SD Standard deviation, min. minimum, max. maximum, GP General practitioner

Note: a higher value of the material deprivation score indicates higher level of deprivation