Methods
Study Setting
This is a pilot study conducted in Edmonton. The City of Edmonton is located in the Province of Alberta, Canada, has a 2016 population of 932,546 and is approximately 30 km east-west and 35 km north-south, covering 700 sq km. This study relied on the standard geography from Census Canada known as Dissemination Areas (DA). It is also the smallest unit of the census data, and the use of DA in this study ensures that vulnerability is generated on precise levels. The following map shows how Edmonton is divided into 1,193 DAs (each grey block) with a population of approximately 400-700 people (Figure 2). After dropping the DAs with no information reported, Figure 3 shows the actual areas that were included in this study (901 DAs).
Figure 2. Map of Dissemination Areas (DAs) in the City of Edmonton.
Figure 3. Map of the Dissemination Areas (DAs) in the City of Edmonton that were used in this study.
Methods
Objective 1: Develop vulnerability indices using pre-determined dissemination area-level factors of exposures, sensitivity, and adaptive capacity.
Method 1: The index was developed using exposure (weather - annual mean temperature, annual cumulative precipitation; air pollution - nitrogen dioxide, fine particulate matter; Urban heat island index), sensitivity (age, marital status, immigration status, education, income, visible minorities) and adaptive capacity (active living environment; greenness; non-profit, health care facility, clinic, and EMS station density) using data for the period 2016-2020. Variables were selected based on modelling, expert opinion, stakeholder perspectives, and those identified from published literature. The unit of analysis was dissemination areas (DAs, the smallest census unit of 400-700 people).
Dimensionality reduction analysis using factor analysis was performed, and the resulting dimensions were combined using the below formula (from the previous theoretical framework) to generate a vulnerability index:
Vulnerability = (Exposure + Sensitivity) - Adaptive Capacity
Retained factors were mainly based on eigenvalues (>1). A clear break in scree plots and percentage variance is explained by the factors that were also taken into consideration where necessary. Factors were assumed to have an equal impact on vulnerability.
Objective 2: To assess if the vulnerable index is significant predictors of health events.
Method 2: We validated the vulnerable index using health events between 2015-2018 from Alberta Health services administrative databases. We will use ecological regression models (negative binomial regression) to assess the relationship between the vulnerability index as independent variables and respiratory (N=34,351), cardiovascular (N=90,817), mental health (N=88,037), and injury events (N=30,529) as the dependent variables. Multiple imputation methods will be explored for missing data. The unit of analysis was dissemination areas (DAs, smallest census unit of 400-700 people).