Skip to main content

Calibration Lab

Paired-reference verification for all eight measuring modules

Metrology status

readiness 0/100

Verified

0/8

Failing

0

Overdue

0

Each module is compared against a traceable reference, then judged on Bland–Altman bias and Arms against the tolerance its own published standard sets. A correction is offered only when the paired data supports one — never by default.

ANSI/CTA-2065 (wrist HR accuracy) · re-verify every 60 d · never verified

Task Force ESC/NASPE 1996 short-term HRV · re-verify every 90 d · never verified

ISO 80601-2-61 (A_rms ≤ 4 % for pulse oximeters) · re-verify every 90 d · never verified

ISO 80601-2-61 Annex (respiration rate ±2 br/min) · re-verify every 90 d · never verified

ISO 80601-2-56 (clinical thermometer ±0.3 °C) · re-verify every 120 d · never verified

ISO/CTA-2056 (step counting, ±10 % MAPE) · re-verify every 180 d · never verified

OIML R76 class III non-automatic weighing · re-verify every 180 d · never verified

Kyle et al. 2004 (ESPEN) BIA vs DXA, SEE ≈ 3.5 % · re-verify every 365 d · never verified

Why it is done this way

  1. 1 · Bias and precision are reported separately (Bland & Altman, Lancet 1986). A device can be perfectly precise and consistently wrong — averaging hides that.
  2. 2 · Accuracy is scored as Arms, the statistic ISO 80601-2-61 mandates for oximetry, applied to every module so the numbers stay comparable.
  3. 3 · Fits use Deming regression, not least squares: the reference is noisy too, and OLS would bias the slope toward zero (regression dilution).
  4. 4 · A slope is only fitted with ≥3 points spanning a real range; otherwise just the offset is corrected. Every point is timestamped and drives the re-verification clock.