Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

ODMAP overdose spike alerts

Aggregate alert from a contested store: a nationwide overdose spike-detection network

This one does not score a person. Agencies enter suspected, unconfirmed overdose events into one shared national store, and a plain rule - a rolling 24-hour count against two standard deviations above the county's own 90-day average - fires an advisory alert, sometimes to the neighboring county so it can prepare. Modeled on the Overdose Detection Mapping Application Program (ODMAP) overdose spike alerts. The clever part is the neighbor edge; the hard part is everything around the store. It lives inside a federal drug-enforcement program, so public-health and law-enforcement read the same suspected data, and the policies both swear it is not an intelligence database and grant the host the right to combine it 'as the HIDTA sees fit.' The threshold learns from the store's own reporting, so a county that stops reporting quietly stops getting alerts. Every number in the annual report is the program grading its own homework. So the question is not whether the model is smart. It is who is allowed to read a public-health signal, what it can be combined with, and whether anyone independent ever checks that the alerts are right.

Stylized model of a documented deploymentBehavioral-health & crisis triage

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 ODMAP-class population-level overdose spike-alert network network: 7 components and 18 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: 2 assumed · 3 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 population-level, inter-agency overdose spike-alert pattern documented in the Overdose Detection Mapping Application Program (ODMAP) case file - not a reconstruction of the actual system, and not a claim about the accuracy of any real alert. The atlas-relevant object is the topology: reporting agencies write suspected, unconfirmed overdose events into one shared national store, and a purely rule-based threshold (a rolling 24-hour count against two standard deviations above that county's own 90-day mean, or a manual value) fires advisory alerts, including anticipatory alerts on neighboring counties. Any reading that implies a per-person overdose risk score misreads it - this is aggregate, unnamed event surveillance, deliberately unlike the atlas's patient-level NarxCare cell, which scores identified patients from prescription histories.

  • baseline

    The core governance dynamic is a contested link-authority tension on the shared store, drawn as the law-enforcement read edge plus the data-combination egress. The store is housed inside a federal drug-enforcement program, and its operating policies both disclaim intelligence-database status and grant the host rights to use the data 'as the HIDTA sees fit,' including combining it with other databases it manages for 'law enforcement and public health products' - both clauses are in the one source and both ship. A 2024 peer-reviewed stakeholder study documented divergent public-health versus public-safety data-privacy standards (drawn as the thin cross-sector friction check), and a 2025 peer-reviewed critical analysis argues the integration risks racialized surveillance and criminalization of people who experience overdose. That argument is a scholarly critique and a documented governance tension, not a documented misuse incident, and it is carried as contested.

  • baseline

    Three further structural facts are drawn on the map. First, a self-referential baseline: the threshold is recomputed from the store's own trailing 90 days (two standard deviations above the mean, updating every 90 days), drawn as the store-to-engine read, so under-reporting quietly lowers future alert sensitivity and the baseline can shift silently - the system does not notify a user when the threshold has changed. Second, coverage is a governed, spatially uneven quantity: only 1,362 of 5,605 approved agencies actually submitted events in 2025, roughly 24 percent, so the alert map is patchy by construction. Third, alert leakage by design: subscribers are added by email and 'do not need to be Overdose Detection Mapping Application Program (ODMAP) users,' drawn as the operator-to-external dissemination, in tension with the policy clause limiting spike-alert distribution to eligible agencies and users.

  • baseline

    The two defining absences are inactive check pathways. No independent evaluation of spike-alert accuracy or of the outcomes of alert-triggered response ever ran: nearly every quantitative figure (498,003 suspected events, 74,805 spike-alert notifications, 5,605 agencies, 37,700-plus users in 2025) is self-published by the program in its own annual report and manuals, with no audited false-positive or false-negative rate, so the independent-evaluation check is drawn empty. And no independent oversight body governs the law-enforcement read access or the data-combination boundary: the only standing governance is contractual and self-administered, so the authority-oversight check is drawn empty too. cross-model-check and oversight-cadence stand up the first; connection-auth and oversight-cadence stand up the second.

  • assumed

    Served people who experience overdose are not in the dynamics, and no overdose, mortality, or clinical outcome is computed from anything in this diagram; the Lab reads institutional propagation only, and a spike alert, a threshold, or a reported event here is an aggregate institutional signal, never a person. Nearly all quantitative figures are self-reported by the program with no independent audit and are labeled as such; the Overdose Detection Mapping Application Program's (ODMAP) own disclaimer is that events are suspected, unconfirmed, agency-defined, incomplete, and 'should not be generalized' beyond participating agencies, so nothing here estimates true overdose incidence. The racialized-surveillance and criminalization concern is a qualitative, contested scholarly critique, not a measured disparity, and is carried in the case file, never quantified here. The system is currently rule-based, not machine learning; an announced 2026 predictive-analytics extension is not yet deployed and ODMAP is not described as predictive in the present tense.

What this example does not show

  • Served people who experience overdose are not modeled here, and no overdose, mortality, or clinical outcome is computed from anything in this scenario; the Lab reads institutional propagation only, and a spike alert, a threshold, or a reported event is an aggregate institutional signal, never a person.
  • The racialized-surveillance and criminalization concern raised in the peer-reviewed critical analysis is a qualitative, contested scholarly critique of the law-enforcement and public-health integration, not a measured disparity or a documented misuse incident; both the program's own not-an-intelligence-database framing and the critique ship together, and nothing here quantifies any demographic effect.
  • Nearly every quantitative figure (the 498,003 suspected events, the 74,805 spike-alert notifications, the agency and user totals) is self-published by the program in its own annual report and manuals with no independent audit, and the Overdose Detection Mapping Application Program's (ODMAP) own disclaimer is that events are suspected, unconfirmed, agency-defined, incomplete, and should not be generalized - so nothing here estimates true overdose incidence.
  • The system is currently a rule-based threshold on aggregate counts, not machine learning and not a per-person risk score; an announced 2026 predictive-analytics extension is not yet deployed, so it is not described as predictive in the present tense, and this cell is deliberately distinct from the atlas's patient-level scoring cell.
  • A safe starting baseline is a property of this model, not a safety promise for any real deployment; the threshold arithmetic is exactly reproducible, but its fidelity as an overdose-spike signal is bounded by a self-reported, incomplete, self-referential event stream whose accuracy the public record does not establish.

Sources and evidence

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

  • ODMAP's shared overdose store is housed inside a federal drug-enforcement program, and its operating policies both state that ODMAP is neither an intelligence sharing database nor a pointer index records system and grant the host permission to use the data as the HIDTA sees fit, including combining it with other databases it manages for law enforcement and public health products; a 2024 peer-reviewed stakeholder study documented divergent public-health versus public-safety data-privacy standards, and a 2025 peer-reviewed analysis argues the integration risks racialized surveillance and criminalization of people who experience overdose, a contested scholarly critique of the link structure rather than a documented misuse incident.

    empirical
    • Government Washington/Baltimore HIDTA, ODMAP Operating Policies and Procedures (odmap.org, Rev. Sept 2022) https://www.odmap.org/Content/docs/training/general-info/ODMAP-Policies-and-Procedures.pdf
    • Academic Syvertsen, Looking into the black mirror of the overdose crisis: Assessing the harms of collaborative surveillance technologies in the United States response (Medical Anthropology Quarterly, 2025;39(1):e12875) https://pubmed.ncbi.nlm.nih.gov/39145768/
    • Academic Allen, Cohen-Serrins, ODMAP: Stakeholder Perspectives on a Novel Public Health and Public Safety Overdose Surveillance System (Journal of Public Health Management and Practice, 2024;30(6):E329-E334) https://pubmed.ncbi.nlm.nih.gov/39078392/

  • In 2025 ODMAP's pre-set county thresholds - a rolling 24-hour count against a threshold each agency sets or accepts, recommended by the system as two standard deviations above the county's own previous 90-day mean, a deterministic rule rather than a machine-learning model - fired 74,805 advisory spike-alert notifications from 498,003 suspected, unconfirmed overdose events that only about 1,362 of its 5,605 approved agencies actually submitted; these figures are self-published by the program in its own annual report and manuals, and ODMAP states its data are suspected, incomplete, not a system of record, and should not be generalized beyond participating agencies.

    empirical
    • Government Washington/Baltimore HIDTA, ODMAP 2025 Annual Report (odmap.org, 2026) https://www.odmap.org/Content/docs/ODMAP-Annual-Report-2025.pdf
    • Government Washington/Baltimore HIDTA, ODMAP Spike Alert Overview (odmap.org, 2026) https://www.odmap.org/Content/docs/training/general-info/ODMAP-Spike-Alert-Overview.pdf
    • Government Washington/Baltimore HIDTA, ODMAP Training Manual (odmap.org, October 2025) https://www.odmap.org/Content/docs/training/general-info/ODMAP-Training-Manual.pdf

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Deadline pressure — Statutory or managerial timeliness rules reward fast approval of machine output over slow disagreement.
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Behavioral-health & crisis triage domain page.

Levers available here and the patterns behind them

Documented case histories