Model card — Smart NOA web simulator

Educational work by a pre-medical student — not medical advice. Not a clinical protocol, prescription, or patient handout, and not for patient care. See DISCLAIMER.

Generated from sim.parameterInventory() and engine/autorun-config.js on 2026-11-05 by node simulator/tools/model-card-tables.mjs so these tables match the code. Structure follows the ISPOR–SMDM Modeling Good Research Practices Task Force-7 report on model transparency and validation (Eddy et al., Value in Health 2012;15:843-850).

Intended use

Education only: an educational, label-calibrated anesthesia pharmacology simulator with fictional adult patients. Not for clinical decisions or dosing. Not a medical device. No device, monitor, pump or patient-data interface. Since November 2026 it also has palliative pain management and end-of-life care teaching modules (fictional patients; hour/day time scale; goals-of-care, proportionality and double-effect teaching; no outcome claims; no content about medical aid in dying, euthanasia or hastening death) — see "Palliative and end-of-life modules" below.

Verification and validation status

• Verification (does the code do what it is written to do): automated tests — 179 engine/library tests (patient derivations, dosing and contraindications, PK/PD sanity, physiology sanity, Python-controller parity, determinism, full GA and MAC runs, model-implementation checks, guided/automatic sequence behaviour, and the November 2026 additions: lidocaine pretreatment, IV fluids by type, antibiotic prophylaxis timing/alarms/checklist, technique templates in guided and full modes) and a 19-case browser smoke test.
• Implementation check of a published model: propofol uses the published Eleveld 2018 model; tests reproduce its reference-patient parameter values as recorded in independent open-source implementations. The PK coefficients were not compared with the original paper's Table 2 in this session (publisher pages unreachable) — see SOURCES.md.
• Face validity: no expert anesthesiologist review has been recorded.
• Timing cross-check (November 2026): every timed element was compared with FDA labels and guidelines — TIMING_EVIDENCE.md (one conflict fixed: MAC propofol now starts 4 min after midazolam; clinician-review checklist included, no review recorded yet).
• External validation: none performed. Simulator outputs have not been compared with observed patient data.
• November 2026 additions: antibiotic doses, windows and redosing intervals (Bratzler 2013, S46) and the lidocaine pretreatment dose (propofol label, S11) appear below as components prophylaxis and pretreatment; fluid label composition (S47, S48) and the fluid UI conventions under fluids; template definitions under template (no doses — filled by recommend()). Antibiotic pharmacology, propofol injection pain and type-specific fluid physiology are not modelled.
• Palliative and end-of-life modules (November 2026): tests/palliative.test.mjs checks every sourced rule (label starting doses and renal/hepatic/elderly adjustments, breakthrough 10–20%, the +50–100% titration suggestion, conversion ratios and the mandatory clinician-entered cross-tolerance reduction, renal warnings, the bowel-regimen reminder, CDC 2022 scope and the 50 MME/day pause), symptom-model sanity, naloxone increments, the six-item palliative-sedation gate, the proportionality alarm, no automatic escalation or acknowledgement, continued comfort opioids, determinism, all six example cases in guided mode, the banner/ethics text and a text scan for the hard exclusions; the browser smoke test runs both modules end to end. The symptom model is illustrative and not validated.
• Parts that cannot be validated with available sources: all rows below marked author estimate — unvalidated (effect magnitudes, physiology constants, wash-in time constants, fluid-response constants).

Parameter inventory (608 parameters)

Provenance counts: published_model: 29, author_estimate_unvalidated: 427, fda_label: 34, trial_or_physiology_study: 3, simulator_convention: 79, guideline: 26, clinical_practice_convention_owner: 2, project_config: 8.

Automatic demonstration and guided mode

Generated from engine/autorun-config.js and engine/autorun.js. Both modes are a scripted sequence of the same phase and airway actions as the manual Run-screen buttons (engine/case-actions.js); they are not a controller and not a clinical recommendation, and they have not been evaluated against clinical practice. They never give, change or stop a drug on their own (only the user's plan steps and the clinician's own clicks do — e.g. guideline prophylaxis, a sedation titration step, an advisory rescue bolus), never extubate unless sim.extubationReadiness() is met (the "Extubate anyway" override is never used), pause on a latched interlock hold or an unacknowledged HIGH-priority alarm without releasing or acknowledging it (the panel then shows a paused banner with the reason and the action needed, and Resume), offer the clinician two choices when MAC sedation stays below the plan target (titrate within the label range, or acknowledge and start the procedure at the current level; the full demonstration pauses there and never chooses), and end as soon as any manual control is used.

While either mode runs, an on-screen banner (exact text: BANNERS in engine/autorun.js) states that it is a scripted educational sequence, not a controller or a clinical recommendation. Steps taken are logged with the prefix AUTO: (log kind auto), and the debrief and the downloaded session JSON (automaticMode, summary.automatic) record the mode, outcome, pauses and where/why it stopped.

GA sequence

1. Transfer to OR — gate “Transfer to OR” (guided: you press it; full demonstration: passes automatically); pre-op holding: enabled 5 min after any premedication bolus has ended (countdown shown).
2. OR setup → preoxygenation — automatic in both modes; after OR setup (2 min, monitors applied): preoxygenation FiO2 1.
3. Start induction — gate “Approve & start induction” (guided: you press it; full demonstration: passes automatically); enabled after 3 min of preoxygenation and at least 10 min after the premedication ended (countdown shown); bag-mask ventilation, then induction plan steps run at their offsets.
4. Secure airway — gate “Intubate (secure airway)” (guided: you press it; full demonstration: passes automatically); enabled when intubating conditions are met; tube + mechanical ventilation.
5. Start maintenance — gate “Enter maintenance mode” (guided: you press it; full demonstration: passes automatically); maintenance plan steps run at their offsets.
6. Start surgery — gate “Start surgery (incision)” (guided: you press it; full demonstration: passes automatically); pre-incision verification, no time-based start: airway + ventilation, processed-EEG index in the GA band, analgesia per plan, neuromuscular block adequate, antibiotic in window (required), WHO time out confirmed by the clinician (the full demonstration waits for the state items and pauses for the clinician items; it never confirms them).
7. End surgery — gate “End surgery” (guided: you press it; full demonstration: passes automatically); clinician-chosen end of surgery.
8. Begin emergence — gate “Enter emergence / extubation mode” (guided: you press it; full demonstration: passes automatically); emergence plan steps run (agents off, antiemetic, reversal).
9. Extubate — gate “Extubate” (guided: you press it; full demonstration: passes automatically); enabled only when sim.extubationReadiness() is met.
10. Observe after extubation — automatic in both modes; observation, then complete.

MAC sequence

1. Transfer to OR — gate “Transfer to OR” (guided: you press it; full demonstration: passes automatically); pre-op holding: enabled 5 min after any premedication bolus has ended (countdown shown).
2. Start sedation — gate “Approve & start sedation” (guided: you press it; full demonstration: passes automatically); after OR setup (2 min, monitors): supplemental O2 FiO2 0.4, then sedation plan steps run at their offsets.
3. Start procedure — gate “Start procedure” (guided: you press it; full demonstration: passes automatically); pre-procedure verification: plan sedation target held for 10 s with all sedation plan steps run, antibiotic in window if indicated, and the WHO time out confirmed by the clinician (the full demonstration pauses for it).
4. End procedure → recovery — gate “End procedure” (guided: you press it; full demonstration: passes automatically); ends the procedure and starts recovery (plan stop steps run).
5. Recovery — automatic in both modes; complete when sedation level is minimal or awake.

Every condition and timing (29 entries; also listed in the parameter inventory as component autorun):

Timing rules and template timing

Generated from engine/timing.js (TIMING_CONFIG, the plan timing validator, live timing checks, pre-incision verification and state-gated repeat boluses) and from the timingRationale of every technique-template step (data/templates.js). A rule marked simulator convention has no source for its number; where a label statement is cited in its text, the convention is anchored to that statement, not taken from it. Clinical practice convention (owner) is practice stated by the project owner, not a published source.

Minimum interval from the end of a premedication bolus to "Start induction": 10 min (pre-op holding 5 min + OR setup 2 min + preoxygenation 3 min). Plan drugs in the induction/sedation phase never start while a premedication bolus is pending or running.

Template step timing (why each step is where it is)

Same-phase loading → maintenance: buildPlan() and the plan editor start the maintenance infusion at loading start + loading duration (mac_dex: dexmedetomidine maintenance at +10:00; mac_propofol: propofol initiation at +4:00, maintenance at +9:00); the plan editor warns when a user schedules them overlapping.

Every timed element was cross-checked against FDA labels and guidelines in November 2026: see TIMING_EVIDENCE.md (what each source says, agreement, classification, the one conflict fixed, and a clinician-review checklist — no clinician review recorded yet).

Palliative and end-of-life modules

Generated from engine/palliative.js palliativeParameterInventory() (data: data/palliative.js; sources S100–S120 in SOURCES.md). Educational simulation with fictional patients; not clinical decision support. Opioid starting doses come only from FDA labels; titration, breakthrough, conversion, CDC, naloxone and palliative-sedation values from the cited guidelines/frameworks; every other number is a labelled simulator convention or an illustrative author estimate. The symptom model (pain, dyspnoea, agitation, secretions, nausea, constipation, RASS, respiratory rate) is illustrative and not validated; no survival or mortality is modelled and no outcome claim is made. Palliative sedation is gated by six documentation items, every titration needs the clinician's approval, the engine never escalates or acknowledges anything on its own, and comfort opioids continue (they are not used to sedate). Excluded by design: any content about medical aid in dying, euthanasia or hastening death.

78 parameters. Provenance counts: fda_label: 11, guideline: 21, simulator_convention: 12, author_estimate_unvalidated: 34.

Advisory haemodynamic rescue

Generated from engine/rescue-config.js (RESCUE_CONFIG; logic in engine/rescue.js; card ui/advisory.js). Owner-approved feature (November 2026). On MAP below the trigger the Run screen (and the guided/automatic panel) shows an advisory that suggests one label-sourced rescue bolus; the clinician gives it with one click ("Give suggested bolus", logged ADVISORY accepted) or dismisses it (logged). Card text: "Advisory only — the simulator never gives a vasopressor by itself." The simulator never gives the suggested drug by itself — not in manual, guided or automatic-demonstration mode — and the advisory never acknowledges alarms or releases interlock holds. Doses are not written in the advisory module: they are read from the drug library rescue regimens (phenylephrine [S18], ephedrine [S19]) through recommend() and the label ceilings. MAP above the high threshold gives teaching text only (no drug, no dose). Epinephrine is never suggested for routine hypotension. Every non-drug checklist line is teaching text without doses: “Check anaesthetic depth and the volatile / hypnotic concentration — reduce if deeper than needed.” “Assess volume status and fluids given; consider a fluid bolus (Fluids panel).” “Look for surgical causes (bleeding, caval compression, retraction) and positioning — tell the surgeon.” “Rule out sensor artefact: check the cuff / transducer level and the waveform before treating a number.” The debrief and the session JSON (summary.advisories) list every advisory shown, accepted and dismissed, with times. Not validated; not clinical decision support.

16 entries (also in the parameter inventory as component rescue):

Advisory: depth index below the GA band

Generated from engine/depth-advisory.js (card ui/advisory.js depthAdvisoryCard; settings in engine/timing.js TIMING_CONFIG). Owner-approved feature (November 2026), chosen instead of recalibrating the illustrative depth model, which reads about 34–39 at label-default maintenance doses. Text only: no drug, dose, rate or percentage is suggested; it is not an alarm (nothing is added to the alarm list, guided and automatic modes never pause for it, it never acknowledges anything and never changes a dose). Shown on the Run screen and in the guided/automatic panel when the simulated index stays below the GA band lower bound for the sustain time during GA maintenance after incision with the airway secured, after loss of consciousness, and not after every maintenance hypnotic is stopped once surgery has ended; never in MAC/sedation, before loss of consciousness or in emergence. Before incision the pre-incision item (acknowledge an index below the GA band) remains the gate. The only control is "Dismiss" (logged). If the MAP < 65 haemodynamic advisory is shown at the same time, each card names the other. Checklist: “Review the volatile concentration (end-tidal / MAC) or the propofol infusion rate against the plan.” “Consider haemodynamics: if the patient is hypotensive, anaesthetic depth may be contributing.” “Confirm the planned depth target and the current surgical stimulus.” “Check for sensor artefact (electrode contact, signal quality) before acting on a number.” Footnote: “Advisory only — the simulator never changes a dose by itself. Simulated index (illustrative); the 40–60 band is the protocol target used in B-Unaware/BAG-RECALL [S52][S53].” The debrief and the session JSON (summary.depthAdvisories) list every card shown and dismissed, with times, and the total time below the band. No outcome effect of deep anaesthesia is claimed. Not validated; not clinical decision support.

Rationale text for each entry: see the timing-rules table above and the parameter inventory (component timing); timing evidence row 65 (TIMING_EVIDENCE.md).

Sources

Full citations, what each supports, and how each was checked are in SOURCES.md (43 entries; plus S100–S120 for the palliative and end-of-life modules).

AI assistance

Code, parameter selection and source gathering for this simulator were produced with AI assistance (Anthropic Claude, via Claude Code, November 2026) under the owner's direction. Sources were checked through web-search records of labels and publications as recorded in SOURCES.md; independent human verification of every source by the owner is pending. AI was not used to generate data presented as observed. The owner is responsible for the content.