Population Cluster Analysis: Challenging Assumptions on Emergency Shelter Use for Better Person-Centred Approaches to Homelessness

In 2006, the RAFT’s emergency shelter for youth was in a situation that is familiar to many: demand for our beds was far surpassing our capacity to provide accommodation. This forced us to turn young people away daily. By 2007-2008, our 24 shelter beds were fully utilized most nights and we provided accommodations for approximately 500 unique youth between the two years.

The overwhelming demand for support jeopardized our capacity to help any of our young guests. We had to reduce the scope of our services to providing a bed, food, and building security. Each young person who accessed our shelter was experiencing some kind of emergency and those emergencies manifested in different ways, but our response was the same for everyone. In turn, this basic level of service compounded our core capacity problem by leaving many young people underserved and repeatedly cycling through our shelter. 

2006 was also the year we started looking at our shelter use data by population clusters to better understand the service needs and experiences of the youth accessing the RAFT’s shelter. The population cluster has helped our service provision better line up with our shelter’s core purpose: to provide emergency shelter.

Challenging Our Assumptions: Who Really Accesses The Emergency Shelter?

Through the population cluster analysis, we found that a number of our assumptions were wrong: 

  • We thought that most shelter guests lived in St. Catharine. In reality, over 50% considered other municipalities their home community.
  • We thought most of the youth in the shelter were choosing to live on the street, allowing them to skip school and have more freedom to use substances. In reality, over 90% were attending school immediately prior to their housing crisis, and substance use was often a means to build social connections with streetinvolved peers, not the reason for experiencing homelessness.

The population cluster analysis allowed us to provide better service, triage support for the individual, confront mistaken assumptions, and ultimately identify a pathway toward prevention programming.

Key Population Clusters: Episodic, Long Single Stays, and Bridging

Our population cluster analysis used data on how often and how long young people accessed our emergency shelter, their housing outcomes and service utilization, as well as a forensic review of client case notes to identify themes and common characteristics for each cluster. The following is a summary of findings on the population clusters that emerged through our analysis. 

Episodic (approximately 25% of shelter guests) 

  • Length of Stay: Brief – 7 days or less (50% of cluster); 30 days or less (73% of cluster)
  • Shelter Use Patterns: Significant time elapses between hostel use (1.5 - 5 years)
  • Predominant Reason for Accessing Shelter: Family/Relationship Breakdown (44%)
  • Service Utilization & Housing Outcomes: Little to no service involvement beyond using a bed – they may agree to use service but will not show up for meetings. Clients largely find their own housing (67% - Room rents, apartments, return to family/friend
  • Common Characteristics – Two Subgroups:
  • ‘Hedging’ Episodic: Half (50%) of our episodic guests completed an intake and never stayed a night, often securing bed in case their other arrangements fell through. This subgroup tends to first access shelter at a younger age (between 16-17 years of age).
  • Episodic due to Suspension, Service Restriction or Unsuccessful Housing Outcomes:Approximately 10%-25% of episodic population cluster 
  • Important Service Response for this Cluster: 
    • Shelter Diversion may be particularly effective with guests that fall in the ‘Hedging’ Episodic subgroup
    • Reducing suspensions and service restrictions by reviewing shelter policies to ensure that policies are not simply focused on controlling or managing behaviour or staff convenience. Declining suspensions and restrictions were correlated with better housing outcomes for young people.

Long Single Stays (approximately 20-30% of shelter guests)

  • Length of Stay: Long – 40 days or more
  • Shelter Use Patterns: Little or no history of previous shelter use; Little or no familiarity with the homeless-serving system; Lack of knowledge about housing/the housing market
  • Predominant Reason for Accessing Shelter: Disruption of Income – e.g., Losing OW; Bank account suspended; No income; Unable to access income due to immigration status, etc.
  • Service Utilization & Housing Outcomes: Increased length of stay when clients do not engage with prevention services but successfully housed once engaged.
  • Common Characteristics: Coming from out-of-region for school, work, family, or partner; Two or more people looking to be housed together; Newcomers; Recently incarcerated or hospitalized
  • Important Service Response for this Cluster: 
    • Income/employment supports
    • Securing proper documentation – ID, Health card, Citizenship, etc.
    • Clarifying immigration status 
    • Shelter diversion to reconnect with region of origin
    • Designating shelter workers to each guest (Prime Workers) to engage in housing-focused conversations every week/shift rotation and recording case notes for review by the Shelter Manager

Bridging (approximately 5% of shelter guests)

A blend of characteristics related to ‘Hedging’ Episodic and Long Single Stays populations

  • Length of Stay: Long shelter stays - 30 or more days
  • Shelter Use Patterns: History of shelter use similar to the Episodic population but with length of stays over 30 days
  • Predominant Reason for Accessing Shelter: In need of temporary accommodation while waiting to access treatment, transitional or supportive housing, etc.
  • Service Utilization & Housing Outcomes: Bridging guests are not looking for housing 
  • Common Characteristics: Looking for accommodation to bridge the time until they can access a program or service which includes residency  
  • Important Service Response for this Cluster: 
    • Ensuring guests have the proper paperwork and logistics to make the transition to their next location when it becomes available.
    • Having the guest commit to a plan for when they will be discharged from shelter.
    • Explore Shelter Diversion options for temporary accommodation.
    • Be ready to engage in housing searches if transition plan falls through

Conclusions: New Pathways for Person-Centred Care and Prevention

To provide better outcomes for our guests we need to confront certain truths about our shelter system: 

1. Shelters remain the most common service response to homelessness. For this reason, the shelter system has grown organically as it’s sought to address the various needs of guests. This organic growth and the addition of ancillary services (addictions, mental health, employment, life skills, etc.) has diluted our focus on the shelter’s original mandate of temporary emergency beds.  It is this truth, which the Housing First philosophy seeks to address by prioritizing housing over these ancillary services.

2.The system tends to view every guest through the lens of needs and consequentially our service response tends to categorize our guests based on these particular needs. For example, all guests that present with addictions are encouraged to participate in addiction services.

The examination of population clusters allows us to step back and view the entire population based on usage as opposed to individual needs. The analysis shows a clear difference between guests who are looking for short term shelter (under 30 days) versus guests who are looking for longer term shelter (45 days plus). Understanding these motivations for either short- or long-term stays allows us to provide a person-centered service. It can also provide information to support prevention and post-shelter support services which would allow guests the ability to avoid shelter or the need to return to shelter. Further, it allows us to better determine which guests would benefit and actively engage with our ancillary services.              

At the end of the day, our guests come to us because of the emergency beds we offer and not because of our excellent life skills course. The examination of population clusters based on bed use brings the management of the shelter in line with the expectations guests have when they access our beds.

To learn more, join us on September 23rd as we discuss the effectiveness of a Centralized Shelter Diversion program, in the Niagara Region in Ontario. Click here to register.