In this early released article and accompanying video abstract from Pediatrics, Dr. Matthew Westfall from the University of Colorado and colleagues from across the country used a new methodologic approach called “multiple systems estimation” to demonstrate that youth homeless experiences are much more common than appreciated (10.1542/peds.2024-067965).
Homelessness is a public health urgency: homeless youth have an all-cause mortality rate that is 10x that of those who are housed.
The authors found that 25% of Denver youth (ages 14–17 years) experienced homelessness in 2021. This shocking number is not a typo—the sheer magnitude of this public health crisis has likely been underestimated previously due to limitations in the usual methods of identifying youth who are homeless.
Most approaches use a straightforward technique called “point-in-time” (PIT) counts with street-level outreach and shelter data to count the total number of those who are homeless on a given winter night. Clearly, this method will miss youth who are sheltered outdoors, who are “couch-surfing” with friends, or living in cars or hotels, or who do not wish to be seen or found for any reason. The consequences of missing these youth are many and range from driving funding gaps to failures in outreach and service provision.
“Multiple system estimation,” or “capture-recapture analysis,” is an indirect estimation method that is useful for populations that are difficult to reach. Like the capture and recapture of fish in a lake—fish are caught, tagged, released, and resampled when recaught (the authors’ analogy is very helpful!)—multiple system estimation uses multiple databases and the demographic characteristics of the known population as information sources for the estimate.
The authors used these databases:
- Homeless Management Information System (HMIS) from the Metro Denver Homeless Initiative, which collects client-level data on housing services provisioning,
- McKinney-Vento data from the Denver Public Schools (DPS), federally required data on education and demographics on homeless youth, and
- Colorado’s Statewide Automated Child Welfare Information System (‘TRAILS’), from the child protection system.
The “known” homeless population included 14–17-year-olds found in one, two, or all three of the databases. The “unknown” homeless population (those not in any of the databases) were estimated by modeling the overlap of the “known” youth in the 3 databases. The “known” and “unknown” were then summed to estimate the total population of homeless youth in Denver city and county. Analyses were stratified by race, ethnicity, and gender. More detailed information is in the article’s methods section.
The prevalence of youth homelessness increased from 10% to 25% over the study period (2017–2020), potentially reflecting national trends and pandemic impacts. Known homeless youth counts were 1.7–3.2 times higher than PIT counts, and the total number of homeless youth was likely 6.5x higher than the known in the 3 databases combined.
These data shook me to the core. I wholeheartedly support the authors’ call to action: “We hope our findings serve as a cogent reminder for busy clinicians to screen for homelessness, housing insecurity, and other relevant social determinants of health among their patients.” Well said, and well done.
