Haematology and immunology
Haemoglobin, g/dL
Laboratory haematology analyser (reference) replaced by Point of care haemoglobin meter A (central laboratory setting). 60 paired results.
Compare it with the other comparisons in the explorer
Comparison line
The new system reads 4.7% higher than the old system on average. The comparison line has a slope of 0.99 (95% interval 0.946 to 1) and an intercept of 0.656.
Fig. 1. The comparison line against the line of no change, over the range of the decision limits (8, 12, 13 g/dL). Dots mark where each line meets each limit; the gap between them is the bias at that limit.
Crossing each limit
Results that moved across a decision limit when the new system replaced the old, against the number repeat testing on one unchanged system would move. The program reports the count across all limits together.
All decision limits together (8, 12, 13 g/dL)
Crossed after the change: 14
Expected from repeat testing alone: about 2
Grade and warnings
- Re-test band: Indicative: 60 pairs, below the 100-pair protocol minimum.
- Band half-width around each limit: 1.36%.
- Results re-tested on data the band was not fitted to, median: 13%.
- Share of the excess crossings caught on unseen data, comparison line and band together: 74%; comparison line alone: 64%. The excess is the crossings above what repeat testing alone would cause, so the share can exceed 100% or fall below zero.
- Repeat-test imprecision: measured, precision file (patient-sample replicates, 5 per sample), the larger of the two systems.
Source
Data: Huub Gelderblom (2016). Performance characteristics of three point of care hemoglobin meters in Durban-2.xlsx. figshare. doi:10.6084/m9.figshare.3119449 (opens in a new tab). Licence: CC BY 4.0 (opens in a new tab). Supplement to Jaggernath M, Naicker R, Madurai S, Brockman MA, Ndung'u T, Gelderblom HC (2016). PLoS One. doi:10.1371/journal.pone.0152184 (opens in a new tab); PMC4821624 (opens in a new tab) (title on the source page; it names the instruments). Changes: Paired results re-analysed by NakedSignal's program; instruments described generically; only derived results are shown. The deposit names the systems.
Show the numbers
| Field | Value |
|---|---|
| Paired results used | 60 |
| Comparison line slope | 0.990176 |
| Slope, 95% interval | 0.946253 to 1 |
| Comparison line intercept | 0.656095 |
| Mean difference, new minus old, as % of old | 4.691 |
| Decision limits (g/dL) | 8, 12, 13 |
| Crossed after the change | 14 |
| Expected from repeat testing alone | 1.59 |
| Band half-width (%) | 1.36 |
| Re-tested, unseen data, median (%) | 13.33 |
| Re-tested, unseen data, range (%) | 10 to 18.33 |
| Excess caught, line and band (%) | 74.21 |
| Excess caught, line alone (%) | 64.49 |