Every social worker and case manager knows the routine. You sit down with a client navigating one of the hardest moments of their life, open your Homeless Management Information System (HMIS), and ask a series of deeply personal questions about their health conditions, housing histories, income, and domestic violence history.
Over my years in frontline case management, I lost count of how many people asked me variations of the same fair question: "Why do you need all this information? How is this actually going to help me?"
My answer was always a promise: "The data we collect today will be used to help serve you better, and if not you directly, it will help us make the system better for the next person who walks through our doors."
For a long time, fulfilling that second half of the promise felt like an uphill battle. We collected a ridiculous amount of data for our HMIS-funded programs. Still, much of it sat in a static database, used primarily for retroactive compliance reporting rather than real-time clinical decision-making.
That changed when I had the opportunity to work alongside Salt Lake County and Social Data Analytics LLC (SDA) to pilot HAST™.
What is the HAST™?
HAST™ was developed by Social Data Analytics LLC (SDA). This article/consulting work references and uses the HAST™ name and methodology with SDA's permission.
HAST™ is an evidence-based, data-driven decision-support tool that is designed to improve prioritization while reducing assessment burden. It is developed under the leadership of Professor Rhema Vaithianathan and the SDA team.
Rather than relying solely on point-in-time vulnerability indexes that can be subjective or easily outdated, the HAST™ analyzes years of historical, anonymized HMIS data to predict an individual's risk of chronic homelessness, emergency shelter dependence, or failed housing placements.
Fulfilling the Promise of Frontline Data
When we deployed the HAST™ model in Salt Lake County, it felt like I personally finally took the first step in fulfilling my promise to each person I served directly. We collectively honored the tens of thousands of hours of work from every frontline worker who ever entered data into HMIS.
By analyzing patterns across more than a decade of historical client trajectories, the machine learning model could identify subtle combinations of risk factors that humans cannot spot in a 30-minute intake. It allowed case managers to:
Intervene Earlier: Identify individuals at high risk of long-term homelessness on Day 1 of shelter entry rather than after months of system involvement.
Target Scarcity Fairly: Ensure high-intensity housing resources (like Permanent Supportive Housing) are prioritized for those least likely to resolve their housing crisis without deep support.
Honor Client Contributions: Turn 15 years of client disclosures into a protective shield for future individuals entering the crisis system.
Reduce Data Collection Time: Reducing the time it took to gather the necessary data lowered the burden of a housing assessment and made it possible to connect people who would otherwise resist completing a lengthy assessment.
Garbage In, Garbage Out: Why Data Quality is Everything
Predictive models like the HAST™ are powerful, but they carry a critical caveat: an algorithm is only as good as the historical data powering it.
If intake records are incomplete, if exit destinations are logged as "Unknown," or if assessment workflows vary wildly across providers, the predictive model's accuracy degrades rapidly. Worse, poor data quality can introduce or amplify systemic biases in resource allocation.
Before an organization, county, or Continuum of Care (CoC) can successfully deploy advanced machine learning, it must first establish rigorous, compassionate data hygiene across its system:
| Focus Area | What Good Data Quality Looks Like |
|---|---|
| Completeness | Minimizing missing fields in core universal data elements (UDEs). |
| Consistency | Standardizing assessment protocols across all participating agencies. |
| Timeliness | Entering shelter entries, exits, and services within 24–48 hours of occurrence. |
| Accuracy | Verifying the data entered into the system matches reality. |
| Relevancy | Including only what is necessary to do the work in the data. |
Is Your System Ready for Predictive Analytics?
Implementing a tool like the HAST™ isn't just about deploying code; it's about building a solid data foundation and ensuring that every piece of data collected serves a clear, compassionate purpose for the people you serve.
Whether you're exploring predictive modeling or simply want to improve the quality and focus of your current HMIS data, Sharpe Homeless Data Consulting helps bridge the gap between frontline practice and advanced analytics.
Services Offered:
Data Scope & Purpose Alignment: Streamline your data collection to focus only on what improves outcomes, avoid unnecessary data creep, and reduce client burden during intakes.
Data Quality Evaluation & Cleaning: Root out systemic data quality errors once and for all and implement practical, lasting fixes.
Training & Standardization: Align frontline workflows across agencies so your data stays clean, accurate, and consistent, no matter who collects or enters it.
Predictive Tool Readiness Strategy: Evaluate your system's readiness for tools like the HAST™ and build an ethical, effective implementation roadmap.
Ready to unlock the true potential of your HMIS data? Reach out today to schedule an initial conversation about your system's data quality and readiness.
