Problem Statement: 

Substance use disorders continue to disproportionately impact vulnerable populations, particularly in urban areas where systemic inequities limit access to care. In San Francisco, treatment and recovery services are vital yet unevenly accessible across neighborhoods. Through a network analysis, this project will provide insights in spatial accessibility to recovery centers, focusing on communities identified as vulnerable through the Healthy Places Index (HPI).

Using ArcGIS Pro’s Network Analysis tools, this study will attempt to answer three key questions:

• Which vulnerable communities are the furthest from substance abuse recovery centers?

• How many vulnerable neighborhoods are outside of a 10-minute walking distance?

• Where is the best location to add a new recovery center to improve access for vulnerable populations currently without immediate services?

Assumptions

Closest Facility

To identify the closest substance use recovery centers to vulnerable communities, HPI-weighted census tract centroids were input as incidents, and known recovery centers were input as facilities. The Closest Facility tool was run using driving time as the travel mode, with the number of facilities set to 1. No cutoff value was applied. This configuration ensures each point connects to its nearest facility based on real-world travel routes. The resulting routes layer was used to determine which vulnerable communities are furthest from recovery centers.

Service Area

To evaluate walkability, recovery facilities were used as input points in the Service Area tool. A cutoff of 1,000 meters was selected, based on common benchmarks for reasonable walking distances in urban areas. Since many vulnerable residents may not have access to a car, walking accessibility was prioritized. The resulting service areas help determine how many vulnerable communities fall within this walkable range.

Optimal Location

The Location-Allocation tool was used to identify the best site to add a new facility to improve access for underserved communities. Demand points were represented by HPI-weighted census tract centroids, using a reclassified positive-value version of the HPI scores (original scores ranged from -1.2 to 0, and the tool does not accept negative weights).

The 11 existing recovery facilities were input as facilities, and several hypothetical candidate locations were digitized. The number of facilities was set to 12 so the tool could allocate one additional facility. Driving time was selected as the travel mode, and no cutoff was applied. The output reveals where an added facility would most improve accessibility for vulnerable populations.