An Examination of the Measurement Invariance and Construct Validity of the Housing Prioritization Tool

Background and Purpose: The vulnerabilities of people experiencing homelessness are complex and intersecting, including mental health challenges, substance use disorders, trauma, physical health issues, and a history of incarceration. These vulnerabilities necessitate interagency collaboration and individualized plans to address homelessness within each community. Historically, the VI-SPDAT was used to assess and prioritize services in most Continuums of Care (CoCs), but it was discontinued due to critiques of its validity and reliability. New tools have emerged to meet the U.S. Department of Housing and Urban Development's (HUD) requirement for a standardized assessment tool. This study explores the psychometric properties of the Housing Prioritization Tool (HPT) through factor analysis and invariance testing to determine its reliability in measuring vulnerability across gender and race.

Methods: Data was collected between November 2022 and February 2024 from a midwest county with around 500,000 residents. The study participants were at least 18 years old and experiencing homelessness at the time of the assessment. This cross-sectional descriptive analysis used a retrospective cohort of participants administered the HPT during the study period. Data was accessed through the Homeless Management Information System (HMIS). The sample size for this study was n=1180, providing more than adequate statistical power. Several analyses were used to examine the HPT using R (version 2024.04.1+748). An exploratory factor analysis (EFA) was conducted to describe and summarize the data by grouping correlated variables. A confirmatory factor analysis (CFA) was conducted on the second data set to confirm similarity to the factor loadings found in the EFA. Convergent and discriminant validity were measured to support the overarching construct validity of the assessment. Lastly, invariance testing was conducted to examine whether the measurement properties held when assessing vulnerability across race and gender.

Results: The exploratory factor analysis identified five factors within the HPT, accounting for 57% of the variance. Component one measured mental health and service use frequency, component two measured chronicity, component three measured physical health, component four measured experience with interpersonal violence, and component five measured criminal activity, income, and ER usage. The CFA confirmed the five-factor solution with adequate fit indices. The HPT demonstrated divergent validity but did not demonstrate convergent validity. The tool exhibited acceptable internal consistency overall, but individual domains ranged from acceptable to questionable. The analysis showed that the assessment tool did not exhibit full scalar invariance between the gender binary but did exhibit full scalar invariance across the racial binary.

Conclusions and Implications: The findings underscore the need for further research to validate the HPT to measure vulnerability within the homeless population. The factor loadings indicate that the hypothesized model demonstrates a good fit to the data but exhibits poor scalar variance across genders. Future research should seek to refine the model further according to theoretical constructs of vulnerability and establish a clear construct of vulnerability specific to the homeless population. Ethical considerations are crucial, as scores derived from this assessment influence decisions that significantly impact the lives of individuals and families facing homelessness or housing insecurity.