The Canadian province of British Columbia (BC) has been in a public health emergency for over three years because of a sharp rise in drug overdoses and deaths.
While most of the deaths have involved illicit fentanyl (an opioid substance), addiction to a wide variety of drugs and alcohol is high amongst particular communities. In the City of Vancouver’s 2019 Homeless Count, 69% of the 2,223 surveyed homeless people reported an addiction to at least one substance.
Step in the non-profit Streetohome, whose aim is to broker a comprehensive system response to homelessness. To address BC’s addiction epidemic, Streetohome has been exploring the possibility of introducing Recovery Community Centres (RCCs) to Metro Vancouver. RCCs are peer-operated centres that serve as local resources of community-based recovery support, and they’ve had great success across the United States.
As part of their mission, Streetohome asked me to perform some spatial analytics to identify potential locations for RCCs in Metro Vancouver. In essence, my task was to produce a map of addiction for the region.
Homelessness and addiction use data are difficult to come by, owing to a variety of privacy, reporting and data quality issues. Instead, I decided to use a ‘proxy’ dataset for estimating addiction: the location of Alcoholics Anonymous (AA) meetings. This dataset is freely available online and, usefully, AA is a core service offered in the RCC model.
The theory goes, by mapping the level of accessibility to AA meetings across the region, we can pinpoint exactly where addiction is being catered for, and where it isn’t. Obviously alcohol doesn’t capture the full spectrum of addiction, but it’s a start.
Collecting the appropriate data
To emphasise the transferability of this method to city regions other than Metro Vancouver, I decided to rely on open data and open-source methods only:
· The locations of all AA meetings across an entire week were scraped directly from the AA Metro Vancouver website.
· The layout of roads and sidewalks of the entire region were downloaded from OpenStreetMap, an open-source equivalent to Google Maps.
· The timetables and routes of the public transit network were downloaded from Translink (the regional transit operator), in the General Transit Feed Specification (GTFS) format.
· Shapefiles of the administrative boundaries of the region were acquired from MapIt.
· Census population counts from 2018, downloaded at the Dissemination Area level (the highest resolution openly available) from Statistics Canada.
Realistic routing to AA locations
In order to simulate realistic journeys through the region, several steps were taken:
· The Metro Vancouver area was first divided into 1-km sized hexagons
· A routing algorithm (described in full detail in this paper) was used to simulate the route between every hexagon’s centre point (known as a ‘centroid’) to every AA location in the region.
· This algorithm was run assuming transportation by (a) car only, and (b) public transit only. OpenStreetMaps provided the actual street layouts for driving, and the GTFS data provided real timetabling of transit. Essentially, this is the same as running thousands of ‘A-to-B’ trips on GoogleMaps — but it doesn’t require a Google API key, which can become costly!
Calculating the ‘catchments’ of AA locations
Once the routing was complete, a ‘cumulative opportunity’ analysis was performed: put simply, I summed the total number of AA locations that could be reached within 30 minutes from every hexagon centroid. 30 minutes was chosen as a rough estimate of how long an individual might realistically want to travel to get to an AA meeting.
As shown in Figure 2, the greater the number of meetings a resident can reach from the hexagon in which they live (coloured in yellow), the higher their ‘addiction centre accessibility’.
It’s possible to calculate how many people fall within different categories of accessibility, based on census data of where people live. By summing up the number of residents in hexagons with similar levels of access, we find that over 1.3 million people (55% of the population) can access at least 50 AA meetings from their home by car. However, zero people can access that many meetings by public transit — so there’s a huge disparity between transport modes. In fact, almost 2 million people (79% of the population) can reach fewer than 25 AA meetings by transit, and 200,000 people can’t reach a single meeting by transit.
To get a better idea of overall accessibility using all available transport modes, accessibility scores by car and transit were weighted and then combined. Transit was assumed to be twice as importance as cars in the weighting, because many vulnerable populations accessing AA services are likely to do so by transit. The combined, weighted score for each hexagon was converted to an ‘accessibility index’ between 0 (worst) and 1 (best).
Figure 4 shows that the greatest overall accessibility to AA meetings is in East Vancouver, West Burnaby and North Surrey.
Going one step further, I highlighted the hexagons with the highest combined scores, and the hexagons with the lowest combined scores. Figure 5 shows that the greatest supply of AA meetings — and possibly, by extension, the highest demand for them — is around East Vancouver (coloured in green), including the Downtown Eastside where the concentration of homelessness is highest. Conversely, there is a far higher diversity of locations throughout the region with zero access to a single AA meeting (coloured in red).
It’s all well counting the number of AA locations somebody can reach from their home, but this doesn’t account for the number of other people they compete with for those services. Just like schools, whose catchments are based on the number of children in the surrounding neighbourhood who might wish to enrol, AA services can only cater to a certain number of people at a time.
To account for competing demand for AA resources, the catchment population of each AA location (determined from the census) was combined with the accessibility scores using the method described in this paper. The resulting metric has no real units, but effectively represents a ‘per capita’ measure of AA accessibility.
Yet again, Figure 6 shows a hotspot around East Vancouver, implying there is a high supply of AA meetings compared to the number of people living in the surrounding neighborhoods.
What do the maps say?
There is some difficulty in interpreting these maps. Should an RCC be set up where accessibility to AA meetings is highest, presuming that a greater number of meetings equates to a greater level of addiction rates or ‘demand’? Or rather, should we target areas where accessibility is lowest, on the basis that there is an unacceptably low ‘supply’ of service?
There is no right answer at this stage — more research is needed to identify potential users’ needs across Metro Vancouver.
* * * * * * * * * * * *
The code necessary to run the OpenTripPlanner algorithm is available on my Github page.