Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

NarxCare

The score you cannot see: an opaque prescribing-risk model

A proprietary model reads a patient's prescription-monitoring record and returns a secret 000-999 risk score that shows up in the patient header next to vitals and allergies — and can decide whether they get pain medication. Modeled on NarxCare. Nobody outside the vendor can see how the score is built, the patient cannot see or contest it, and when independent researchers rebuilt a version of the model its measured precision came nowhere near the vendor's claim. The question this round asks is not how to score better — it is which controls make a score contestable at all.

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 NarxCare-class opaque prescribing-risk score network: 5 components and 12 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: 3 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 proprietary, opaque prescribing-risk-score pattern documented in the NarxCare case file — not a reconstruction of the actual tool or its algorithm.

  • baseline

    This shape's defining feature is an absence: both inhibiting check pathways are drawn empty at baseline, because the model is proprietary and undisclosed so no independent audit or external validation of the deployed model is possible, and because the Food and Drug Administration (FDA) has not regulated it and patients cannot see or contest their scores, so no validation, re-authorization, or recourse channel exists that any actor in the deployment can reach.

  • assumed

    The score is advisory in name and documented de facto determinative in practice: the model-to-operator channel is drawn as strong anchoring while the operator-to-model override is drawn low, reflecting documented automation bias, the score's prominent placement, and fear of regulatory and criminal liability. No public override or denial rate exists, so the exact strengths are a modeling choice, not measured rates.

  • baseline

    The feature loop is present at baseline: PDMP dispensation records are the score's feature source, and because opioid-use-disorder treatment medication carries high morphine-milligram-equivalents, being in recovery treatment can itself raise a patient's own Overdose Risk Score — a documented perverse dynamic in the record-to-model loop.

  • baseline

    Peer coupling is drawn between the two gates: prescriber and pharmacist rely on the same single opaque score, so its blind spots are correlated across both decisions rather than idiosyncratic, and a high score cascades from one gate to the other.

  • assumed

    No overdose, suicide, or clinical outcome is modeled here. This Lab models institutional propagation only, and the patients this score sorts are not in the dynamics. The documented discrimination concern in the served population is qualitative and contested, is recorded in the case file, and is never computed from anything in this diagram.

What this example does not show

  • This Lab models institutional propagation only. It never models overdose, suicide, or any clinical outcome, and the patients this score sorts are not in the diagram — a score or a denial here is an institutional signal, never a person's care or harm. The documented patient harms, denials, and the score-to-denial-to-illicit-market dynamic are recorded in the case file, measured outside any diagram like this one.
  • The contested-accuracy figures are read carefully: the vendor's ~0.75 precision is self-reported on its own data, and the independent 0.01-to-0.32 precision comes from a 2026 simulation reconstruction that, lacking overdose-death labels, trained on proxy outcomes — so it is evidence that proprietary opacity prevents outside assessment, not a strict like-for-like refutation of the vendor's number. Both are carried as what they are in the case file.
  • The documented discrimination concern — that this kind of score is likely inflated for women and for Black, poor, uninsured, and rural patients through proxies such as cash payment and distance traveled — is an argued, contested concern, not a measured per-subgroup rate, and no disparity figure is asserted here; it is carried qualitatively in the case file as an external observation and is never computed from these dynamics.

Sources and evidence

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

  • NarxCare is a proprietary clinical-decision-support platform built by Bamboo Health that layers over state Prescription Drug Monitoring Programs and returns three Narx Scores plus a composite Overdose Risk Score (each 000-999) into the electronic health record, the PDMP portal, or pharmacy software, often in the patient header alongside vitals and allergies; adoption figures vary by what is counted (more than 40 states and territories run their PDMPs on Bamboo technology and five of the top six pharmacy chains use NarxCare, while the scoring module itself is switched on in more than 20 states). The vendor states the scores are intended to aid, not replace, clinical judgment and should never be sole justification for providing or refusing medication, but clinician and patient-advocacy sources document de facto determinative use — denials, forced tapers, and pharmacy refusals — driven by automation bias and fear of regulatory and criminal liability; patients cannot see, challenge, or correct their scores, the algorithm is proprietary and has not been independently validated for clinical care, and the FDA has not regulated it as a Software-as-a-Medical-Device, so contestation has instead run through FDA citizen petitions (one rejected on procedural grounds in 2023 and a second, docket FDA-2025-P-0701, pending since 2025 with more than 1,000 public comments).

    empirical
    • Vendor Bamboo Health, Inc., NarxCare Application Overview (Version 1.0, September 2023; hosted by the Idaho Division of Occupational and Professional Licenses) https://dopl.idaho.gov/wp-content/uploads/2024/07/2023.10.04.Bamboo-Health-NarxCare-Application-Overview.pdf
    • Academic Wang, Stofer, Chu, Huang, Li, Algorithmic opacity in opioid risk scoring and the need for transparent AI regulation (npj Digital Medicine, 2026; DOI 10.1038/s41746-026-02491-y) https://www.nature.com/articles/s41746-026-02491-y
    • Investigative Miller and Whitehead, Artificial Intelligence May Influence Whether You Can Get Pain Medication (KFF Health News, 2023) https://kffhealthnews.org/news/artificial-intelligence-pain-medication-narx-score/
    • Academic Buonora, Axson, Cohen, Becker, Paths Forward for Clinicians Amidst the Rise of Unregulated Clinical Decision Support Software: Our Perspective on NarxCare (Journal of General Internal Medicine, 2023) https://pmc.ncbi.nlm.nih.gov/articles/PMC11043299/
    • Academic Oliva, Dosing Discrimination: Regulating PDMP Risk Scores (California Law Review, 2022; Vol. 110) https://www.californialawreview.org/print/dosing-discrimination-regulating-pdmp-risk-scores
    • Investigative Pain News Network, Petition Asks FDA to Take NarxCare Off the Market (2023) https://www.painnewsnetwork.org/stories/2023/4/28/citizens-petition-calls-on-fda-to-take-narxcare-off-the-market-nbsp
    • Trade press Medscape, Hidden Formulas, High Stakes: The Fight to Regulate Clinical Decision Support Tools (2025) https://www.medscape.com/viewarticle/hidden-formulas-high-stakes-fight-regulate-clinical-decision-2025a1000cw3
    • Trade press Medscape, When an Algorithm Guides Pain Management: The Growing Backlash Against NarxCare Scores (2025) https://www.medscape.com/viewarticle/when-algorithm-guides-pain-management-growing-backlash-2025a100091n

  • A single automated rule set applied uniformly and without human review produced tens of thousands of correlated wrongful fraud determinations in the documented Michigan MiDAS case — one flaw repeating at caseload scale rather than averaging out.

    empirical
    • Government Michigan AG, settlement of civil-rights class action (Bauserman, 2022) https://www.michigan.gov/ag/news/press-releases/2022/10/20/som-settlement-of-civil-rights-class-action-alleging-false-accusations-of-unemployment-fraud
    • Investigative IEEE Spectrum, Michigan's MiDAS unemployment system: Algorithm alchemy that created lead, not gold https://spectrum.ieee.org/michigans-midas-unemployment-system-algorithm-alchemy-that-created-lead-not-gold

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