EMERGENZ MERIDIAN
← Platform Evidence →

Methodology

Published, versioned,
defensible.

Every weight, threshold, data source, and limitation is documented, versioned, and published under Creative Commons Attribution 4.0. Methodology v1.0 is the foundation. Empirical validation is the roadmap.

The Signal Doctrine

Signals describe.
They do not decide.

Every EMERGENZ Meridian output is a signal indicator — not a causal finding, not a verdict, not a recommendation for specific operational decisions. This is not a legal disclaimer. It is the conceptual framework governing every design decision in the platform.

A signal is a measurement derived from observable public data that describes a condition. A verdict assigns cause, blame, or responsibility. EMERGENZ Meridian produces signals. The interpretation of those signals belongs entirely to the humans and agencies that use the platform.

P.01
All inputs are public government data
No proprietary data sources. No personal health information. No individually identifiable records. Every data source has a public URL and a public license.
P.02
Every weight is documented and versioned
Component weights are published with full literature citations. Changes to weights increment the version number. Historical scores are archived with their version label.
P.03
Limitations are disclosed, not footnoted
Every score documentation leads with its limitations. Temporal validity concerns, geographic transferability constraints, and validation status are disclosed prominently and explicitly.
P.04
Architectural, not decorative governance
The signal disclaimer is in the API response, not the documentation. The PCSS pre-publication review is enforced at the endpoint layer. Governance that depends on human compliance fails; governance built into the system holds.

Mandatory API Signal Disclaimer

Every API response contains this object. Non-removable. Present in all response formats.

signal_disclaimer Required · All endpoints
// Injected at middleware layer
"signal_disclaimer": {
  "type": "signal_indicator",
  "not": ["causal_finding",
    "performance_evaluation",
    "operational_recommendation"],
  "methodology_version": "1.0",
  "validation_status": "literature_derived",
  "license": "CC-BY-4.0"
}
// Cannot be filtered or suppressed.

Scoring Architecture

How a score is computed.

1
Data Ingestion
16 ETL jobs ingest public data sources on scheduled cycles. All sources are public government datasets with documented APIs. Raw data stored in PostgreSQL+PostGIS with ingestion logging.
Sources: DataSF 311 Cases · SFFD Dispatch · SF OCME monthly reports · ACS 5-year · CDC SVI · CalEnviroScreen · SF Planning · SF DPH naloxone registry · VA vending locations · CFSRE/DEA alerts
2
Component Normalization
Each component variable is normalized to a 0–1 scale within the SF census block group population. Min-max normalization within the city boundary. Values beyond 3σ capped at 99th percentile.
3
Weighted Composite Score
Normalized components multiplied by v1.0 literature-derived weights and summed. Result is a continuous 0–1 composite. Score tiers applied at fixed thresholds: Low (<0.25) · Elevated (0.25–0.50) · High (0.50–0.75) · Critical (>0.75).
4
Versioning and Archival
Every score batch tagged with methodology version, computation date, and source data vintage. Historical scores archived under their prior version tag when methodology changes.
Schema: cbg_geoid · score_date · methodology_version · composite_score · tier · component_json
5
API Delivery with Mandatory Disclaimer
Pre-computed cached scores served via FastAPI. No live computation on user requests. signal_disclaimer injected at middleware. Public endpoints return neighborhood-level aggregates; credentialed endpoints return block group resolution.

Versioning Policy

Methodology changes
are not silent.

Minor version increments adjust individual weights within the existing framework. Major increments change score definitions or produce materially different outputs. Both are logged and archived.

VersionStatusWeight BasisValidationKey Changes
v1.0 Current Literature-derived. All weights traceable to peer-reviewed sources. Not yet empirically validated against SF mortality data. Initial release. 8 ACS components, 5 EES components, 3 NCS parameters, PCSS baseline, STI Tiers 2–3.
v1.1 Planned Empirically calibrated via logistic regression on OCME census-tract data. SF-specific validation of ACS and 311 weights against 2023–2024 mortality data. Weight recalibration for SF context. Columbus 311 transferability validation. SRO multiplier updated to post-2020 data if available.
v2.0 Future ML-based (LASSO / random forest) per Schell et al. 2022 methodology. Peer-reviewed validation study submitted for publication. ML weight derivation. STI Tier 1 wastewater. Social media signal integration.

Known Limitations — Disclosed, Not Footnoted

These are not edge cases. They are material qualifications applied consistently throughout every MERIDIAN output.

Critical
Columbus 311 Weights Applied to SF
Li et al. 2020 accuracy values are from Columbus, OH 2008–2017. SF has different geography, housing stock, and reporting culture. Transferability not validated.
→ Validated in v1.1 using SFFD 2023–2025 data
Critical
SRO Multiplier from 2010–2017
Rowe et al. 2019 covers 2010–2017, predating fentanyl dominance. The 19.3× multiplier may differ in the current era.
→ Revalidation sought via OCME data access partnership
Critical
No Platform Validation Study
EMERGENZ Meridian v1.0 has no published validation study. Component sources are validated individually; the composite architecture is not.
→ Retrospective validation study: Year 1 deliverable
Medium
ACS is a Lag Indicator
The 30-day rolling update cycle means ACS describes past conditions. Rapid enforcement displacement can shift risk faster than the update cycle captures.
→ STI provides real-time complement
Medium
Cychlorphine Potency is Estimated
The ~10× potency figure is from in vitro data, personal communication (CFSRE Jan 2026). No peer-reviewed clinical pharmacology exists as of May 2026.
→ Updated when peer-reviewed pharmacology published
Low
STI Tier 1 Not Operational
Wastewater-based detection requires a DPH partnership not yet established. Tiers 2 and 3 are operational at Phase 1 launch.
→ DPH outreach initiated; Phase 2 target

Licensing

Open methodology.
Replicable by design.

Platform code is released under Apache License 2.0. The methodology — scoring frameworks, component weights, documentation, and audit trail — is released under Creative Commons Attribution 4.0 International (CC BY 4.0).

Any jurisdiction can replicate the MERIDIAN scoring architecture using their own public data sources. All inputs are standard publicly available government indices (CDC SVI, Census ACS, CalEnviroScreen) or categories of data most large cities maintain. Attribution is required. Adaptation is encouraged.

Platform code
Apache License 2.0 — free to use, modify, distribute with attribution
Methodology
CC BY 4.0 — share and adapt with attribution. Board approval pending for final adoption.
Attribution
Organizational author: EMERGENZ Corporation — not attributed to individual staff.