Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

VI-SPDAT

The standard nobody validated: a homelessness triage score

A yes/no interview becomes a 0-to-17 score, and the score decides who reaches the top of the housing line. Modeled on the Vulnerability Index-Service Prioritization Decision Assistance Tool (VI-SPDAT). Watch the thing that never happened: this score became the coordinated-entry standard across dozens of states without anyone validating that it worked, and its intended human override quietly became the whole decision.

Stylized model of a documented deploymentHousing & homelessness services

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the VI-SPDAT-class self-report triage score network: 5 components and 10 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 5 assumed · 2 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

  • assumed

    This models the self-report coordinated-entry triage pattern documented in the Vulnerability Index-Service Prioritization Decision Assistance Tool (VI-SPDAT) case file — not a reconstruction of the actual tool. It is drawn historically: OrgCode began phasing the tool out in December 2020 and ended support at the close of 2022.

  • baseline

    This shape's defining feature is an absence: the independent cross-check of the model is drawn empty at baseline, because no validation, reliability, or equity check ran on the instrument as it became the coordinated-entry standard across at least 39 states, and the builder itself said it was never designed with a racial or gender equity lens.

  • assumed

    The intended human override is drawn as an inactive peer-review step, present on the map but empty: OrgCode labeled the tool 'Decision Assistance, not Decision Making,' but its central concern at withdrawal was that many communities relied solely on the score, so the correction channel that could catch a biased rank sits open for a lever to fill, not active.

  • assumed

    The tool has no administrative-data feed: its only input is the self-report interview, so there is deliberately no record-to-model feature loop. The modeled memory loop is store-to-operator — the biased score persists in the shared Homeless Management Information System (HMIS), builds the prioritization list, and carries into downstream referrals (record contamination), with re-administration a second documented feedback.

  • baseline

    The bias mechanism is drawn on the interview-to-model input edge, where it is documented to live: a deficit-based questionnaire on which people of color more often answer 'No,' scoring as less vulnerable — the bias is in the input elicitation and item design, not in a hidden learned weight, which is why improving the model moves it least.

  • assumed

    The community prioritization list is drawn as a mediating artifact on the model → staff pathway, reflecting a system whose output is a ranked list. It carries no flow of its own and does not affect the dynamics.

  • assumed

    The documented racial scoring disparity is recorded externally in the case file, not computed here. This Lab models institutional propagation, not demographics, and estimates no differential harm to the people experiencing homelessness this score rations scarce housing among.

What this example does not show

  • Bias propagates here the way institutional failures do — through interviews, records, and prioritization lists. The Lab models institutional propagation only: no demographics, and no differential harm to the unhoused people this score ranks. The documented racial scoring disparity (four Continuums of Care, 2019; a large 2020 community sample; Fresno-Madera, 2024) is recorded in the case file and measured outside any diagram like this one; note that the newest single-community study's aggregate tier-assignment tests were not statistically significant.
  • A housing allocation is a one-shot rationing decision, not an error that spreads — so the modelable surfaces here are the validation and override checks that never engaged and the biased score's persistence on the prioritization list, never a claim about any individual's outcome.
  • This is a historical shape (2013-2022): the tool was withdrawn by its own builder and is being replaced community by community. It is preserved as the sector's defining example of a decision-assistance score used as a decision-maker, not as a live system.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • The VI-SPDAT was the dominant U.S. homelessness triage assessment for roughly a decade, adopted in at least 39 states and the District of Columbia by 2015, before its own creators announced its phase-out in December 2020 on equity grounds; a 2019 commissioned racial-equity evaluation across four Continuums of Care found race predicted 11 of 16 subscales and that people of color received statistically significantly lower prioritization scores.

    empirical
    • Advocacy National Alliance to End Homelessness / Homelessness Research Institute (Joy Moses and Ann Oliva), Looking Back at the VI-SPDAT Before Moving Forward (2022) https://endhomelessness.org/wp-content/uploads/2022/08/NextGenTools_VISPDATBrief_08-30-22.pdf
    • Vendor OrgCode Consulting (Iain De Jong), A Message from OrgCode on the VI-SPDAT Moving Forward (2020) https://www.orgcode.com/blog/a-message-from-orgcode-on-the-vi-spdat-moving-forward
    • Academic C4 Innovations (Wilkey, Cannon, Donegan, Yampolskaya), commissioned by Building Changes, Coordinated Entry Systems: Racial Equity Analysis of Assessment Data (2019) https://homelesshub.ca/resource/coordinated-entry-systems-racial-equity-analysis-assessment-data/

  • The VI-SPDAT showed poor test-retest reliability, with most participants scoring higher on re-administration, and poor inter-rater reliability, with scores varying by interviewer and site; its predictive validity for housing outcomes was mixed across studies, positive for the youth version, null for single adults in one study, and positive in another community sample.

    empirical
    • Vendor Bitfocus, Going Beyond the VI-SPDAT: Deficiencies of the VI-SPDAT (2021) https://www.bitfocus.com/blog/deficiencies-of-the-vi-spdat
    • Advocacy National Alliance to End Homelessness / Homelessness Research Institute (Joy Moses and Ann Oliva), Looking Back at the VI-SPDAT Before Moving Forward (2022) https://endhomelessness.org/wp-content/uploads/2022/08/NextGenTools_VISPDATBrief_08-30-22.pdf
    • Academic Shinn and Richard, Allocating Homeless Services After the Withdrawal of the Vulnerability Index-Service Prioritization Decision Assistance Tool (American Journal of Public Health, 112(3):378-382, 2022) https://pmc.ncbi.nlm.nih.gov/articles/PMC8887175/

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Austerity & recovery incentives — Cost-cutting and overpayment-recovery targets tilt the system toward denial and enforcement errors.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Housing & homelessness services domain page.

Levers available here and the patterns behind them

Documented case histories