Continuous Stats Engine — Precision Module
Vorian Connect continuously monitors QC data across all connected instruments. The precision module calculates rolling mean, standard deviation, and coefficient of variation for every analyte — then checks results against Westgard rules with analyzer-appropriate thresholds.
How It Works
- The stats engine runs as a sidecar observer — it reads QC results as they flow through the pipeline but never blocks or delays result delivery
- Rolling precision windows are computed per instrument, per analyte, producing mean, SD, and CV for every configured window
- Westgard rule violations are detected automatically by comparing the current precision window against the established baseline
- Analyte-specific CV thresholds are applied: 5% for chemistry analytes, 10% for immunoassays — no manual threshold configuration needed
- All precision results are stored in PostgreSQL with per-tenant isolation, indexed for fast retrieval by instrument, analyte, and time window
What's Included
- Rolling precision computation: Mean, sample standard deviation (n−1), and CV for every analyte window
- Westgard rule detection: Automatic comparison of current precision against baseline with human-readable violation descriptions
- Built-in analyte classification: 20 chemistry analytes (GLU, CREA, BUN, Na, K, Cl, etc.) use 5% CV threshold; all others default to 10% immunoassay threshold
- Per-tenant isolation: Stats tables use the same RLS policies as the rest of the platform — no cross-tenant data leakage
- Dashboard API: Precision results, violation reports, and trending data are available through the observability endpoints
Where It Fits
The stats engine shares infrastructure with the auto-verification pipeline but operates independently. It does not gate or modify results — it observes and reports. Lab managers and QC specialists view precision trends in the dashboard alongside message volumes, delivery status, and interface health. The engine writes to a dedicated rolling_stats table, keeping QC data separate from message logs while sharing the same tenant-scoped access controls.
Technical Details (for evaluators)
| Capability | Detail |
|---|---|
| Computation | Rolling mean, sample SD (n−1), CV% = (SD/mean) × 100 |
| Chemistry CV threshold | 5% — applied to 20 commonly ordered chemistry analytes |
| Immunoassay CV threshold | 10% — default for all analytes not in the chemistry set |
| Violation detection | Westgard rules: current window SD vs baseline SD |
| Storage | PostgreSQL with tenant-scoped RLS, indexed by instrument + analyte + time |
| Isolation | Sidecar observer — no impact on result delivery pipeline |
The precision module is Phase 1 of the continuous stats engine. Future phases will add moving average tracking, inter-instrument comparison, and QC lot-to-lot validation. All computation runs at the edge or in the gateway — no patient data is sent to external services for statistical analysis.