Table 1.
Performance of the Full versus Simple CSLR Model in Error Prediction. The Average Run Length (ARL) needed to detect test results with either a low or high bias for the full and simple CSLR models. The sensitivity of error detection for each test is shown in parenthesis. Specificity was fixed at 90%. All differences in ARL between the full and simple model were hightly statistically significant (P<0.0001).
| Low Bias | High Bias | ||||
|---|---|---|---|---|---|
| Simple Model |
Full Model |
Simple Model |
Full Model |
||
| Test | Error | ARL (Sens) |
ARL (Sens) |
ARL (Sens) |
ARL (Sens) |
| Na | ± 4 mmol/L | 27 (100) | 13 (100) | 42 (94) | 10 (100) |
| K | ± 0.5 mmol/L | 23 (100) | 13 (100) | 46 (92) | 15 (100) |
| Cl | ± 5% | 28 (100) | 4 (100) | 28 (100) | 8 (100) |
| HCO3* | ± 10% | 41 (99) | 14 (100) | 64 (73) | 21 (99) |
| Ca | ± 0.25 mmol/L | 26 (100) | 5 (100) | 55 (78) | 4 (100) |
| Alb | ± 10% | 60 (90) | 27 (99) | 129 (61) | 25 (98) |
| TP | ± 10% | 42 (99) | 15 (100) | 110 (67) | 18 (100) |
| ALP | ± 30% | 58 (80) | 16 (100) | 47 (94) | 26 (91) |
| ALT | ± 20% | 86 (69) | 27 (96) | 76 (81) | 33 (88) |
| AST | ± 20% | 66 (77) | 19 (100) | 62 (89) | 25 (96) |
| TB | ± 0.4 mg/dL | 36 (98) | 19 (100) | 31 (100) | 20 (100) |
| Creat | ± 0.3 mg/dL | 20 (100) | 13 (100) | 34 (99) | 21 (100) |
| Glu | ± 10% | 87 (68) | 31 (94) | 55 (96) | 40 (83) |
| BUN | ± 2 mg/dL | 49 (87) | 37 (86) | 63 (72) | 40 (84) |
| Mean | 46 (90) | 18 (98) | 60 (85) | 22 (96) | |
*
No available CLIA error recommendation.