Skip to content
HomePublicationsHomeless and High-NeedsThe Homelessness Prevention Unit: A Proactive Approach to Preventing Homelessness in Los Angeles County

The Homelessness Prevention Unit: A Proactive Approach to Preventing Homelessness in Los Angeles County

No Comments
Sun-filled photograph of downtown Los Angeles skyline during the day.

REPORT: The Homelessness Prevention Unit: A Proactive Approach to Preventing Homelessness in Los Angeles County PDF

PRESS RELEASE: New Report Offers In-Depth View of LA County’s Groundbreaking AI-Powered Homelessness Prevention Program

WEBINAR: The Homelessness Prevention Unit

Homelessness continues to be a pressing issue, and new approaches to addressing it are urgently needed. Programs offering one-time cash assistance ranging between $1,000 and $5,000 and short-term direct services have been shown to be effective at reducing homelessness. However, homelessness is a statistically rare occurrence, and it is therefore challenging to target these scarce prevention services to people who would otherwise experience homelessness if they did not receive the help. This report brings to light an innovative program that addresses that challenge by using predictive analytics to identify people who are at high risk of homelessness in Los Angeles County. Case managers use proactive outreach, tailored case management, and flexible cash assistance to help these individuals stabilize their situations and avoid homelessness.

The Homelessness Prevention Unit (HPU) was launched by Los Angeles County in 2020 and operates out of the Housing for Health division of the Department of Health Services (DHS) in close collaboration with the Chief Information Office (CIO) and Department of Mental Health (DMH), and in partnership with the California Policy Lab, which provides the predictive model. The California Policy Lab is evaluating the impact of the HPU through a randomized control trial, with results expected in 2027. The report (and the findings below) share preliminary findings about the HPU, who it is serving, what types of support are provided, and self-reported housing outcomes when people complete the program.

Key Findings

1. The predictive model identifies people at high risk of experiencing homelessness. Individuals on the high-risk list experience homelessness at a rate that is nearly 3.5 times higher than the rate of homelessness among the eligible population. In addition, all individuals on the high-risk list — whether they experienced homelessness or not — experience other adverse events, such as hospitalizations, mental health crisis holds, criminal legal involvement, and death, at a higher rate than the overall eligible population.

2. The model is equitable in the way it predicts risk of homelessness. We evaluated whether the model performs similarly for people from different races, ethnicities, and genders. This is important because if the model is less effective at predicting a person’s risk of homelessness because of their race, ethnicity, or gender, it could lead to a program that underserves people from that group. We found that the model was slightly better at predicting Black people who were at risk of homelessness. Because Black people are historically at greater risk of homelessness as a result of discrimination and systemic racism, we did not adjust the model to select fewer Black people for the program. Otherwise, we found no differences in the model’s performance across race, ethnicity, and gender.

Figure 2: False Negative Rates: the percentage of people who were not assigned to HPU among those who experienced homelessness, broken down by gender, race, and ethnicity
This bar graph illustrates the false negative rates—the percentage of individuals who experienced homelessness but were not assigned to HPU outreach—disaggregated by gender, race, and ethnicity using 2019 test data. The false negative rate is lowest for Black individuals at 56%, compared to 63% for White individuals, 65% for Hispanic/Latino individuals, and 65% for additional racial and ethnic groups. Gender disparities are also shown, with men having a false negative rate of 61% and women at 62%. The yellow bars represent 95% confidence intervals for each group, indicating the range of uncertainty around these estimates. The false negative rate for Black individuals is the only statistically significant difference across gender, race, and ethnicity groups.

Note: This figure uses test data on 47,582 individuals in the prediction eligible population (see section 4 for a comprehensive list of “prediction eligible” criteria) from 2019 so that 18 months of outcome data is available to evaluate the equity of the model. Yellow bars represent 95% confidence intervals. Also, people who identified as “Asian American,” “American Indian or Alaska Native,” “Native Hawaiian and Pacific Islander,” “Multiracial,” or “Other” are included in the “Additional Groups” category because they could not be meaningfully analyzed due to small sample sizes.

3. The predictive model identifies and the HPU serves individuals who are otherwise not accessing other homelessness prevention services. There is minimal overlap between the HPU high-risk list and the people enrolled in current prevention programs in LA County. There may be multiple reasons for this finding. The HPU model requires recent LA County service activity to be considered eligible, whereas current prevention programs are open to people without any County service history. The HPU model also excludes individuals with recent enrollments in homelessness services, while most other prevention programs do not.

Figure 3: In 2019, only 42 people overlap between the Homelessness Prevention Unit’s high-risk list and other prevention programs
This venn diagram shows the overlap between the Homelessness Prevention Unit (HPU) high-risk list and other prevention programs in 2019. Of the 10,672 individuals identified by the HPU high-risk list and 2,522 people enrolled in other prevention programs, only 42 people are included in both, highlighting the minimal overlap between these groups.
Note: This figure shows the HPU high-risk list from 2019 model validation data, as described in section 3 of the report. Other prevention enrollments are from 2019 HMIS data in the Information Hub, for people aged 25 and older in order to correspond with HPU eligibility criteria.

4. HPU participants have more intense service needs than those in other homelessness prevention programs. The HPU is serving a population with significantly higher service needs, including higher rates of service use indicating serious mental illness, substance use, risk of mortality, and prior involvement with the criminal legal system.

5. The HPU intervention model is more intense than most prevention programs (which typically offer cash assistance without customized case management) and is designed to serve participants with complex needs. The HPU provides an average of 6 months of case management catered to participants’ unique service needs; a 15 to 1 ratio of participants to case managers; an average of $6,469 of financial assistance used in a variety of ways to foster housing stability or return to work; and direct connections to other County supportive services like job training or mental health treatment.

Suggested Citation: Blackwell, B., Caprara, C., Rountree, R., Santillano, R., Vanderford, D., Battis, C. (2024). The Homelessness Prevention Unit: A Proactive Approach to Preventing Homelessness in Los Angeles County. California Policy Lab, University of California. https://capolicylab.org/the-homelessness-prevention-unit-a-proactive-approach-to-preventing-homelessness-in-los-angeles-county/



Stay Informed