Validation of Hemoglobin and Hematocrit Measurements from a Dialysis Machine Sensor Compared to Laboratory Analysis

Abstract

Background: Continuous monitoring of hemoglobin (HB) and hematocrit (HCT) during hemodialysis could improve fluid management and patient safety. The Fresenius 5008 dialysis machine includes an ultrasound-based sensor that estimates HB and HCT values, though its accuracy compared to standard laboratory measurements remains unclear. Methods: This exploratory observational study assessed the agreement between sensor-derived and laboratory-derived HB and HCT values in 20 patients at the start of hemodiafiltration. A total of 177 paired blood samples were collected. Results: Sensor values significantly underestimated laboratory HB (9.61 vs. 11.31 g/dL) and HCT (27% vs. 34%) (p < 8 × 10^-25). Correlations were strong for both parameters (HB: r = 0.788; HCT: r = 0.876). Regression analyses revealed consistent proportional bias. Applying a fixed correction of +1.69 g/dL for HB and +7.55% for HCT eliminated the statistical differences and reduced intercepts in regression models. Bland-Altman plots confirmed improved agreement post-correction. Albumin levels correlated modestly with error magnitude. Conclusions: HB and HCT values from the Fresenius 5008 sensor are strongly correlated with laboratory data but are systematically underestimated at treatment start, likely due to hemodilution. Applying fixed correction factors improves accuracy and supports the sensor's use for real-time monitoring.

Keywords: blood volume monitoring; hematocrit; hemodialysis; hemoglobin; sensor validation.

Figure 1

Linear regression: The graphs (A,B) display the linear regression of hemoglobin (HB, on the upside) and hematocrit (HCT, on the downside); values are corrected using the proposed adjustment factor. Each black dot represents a paired data point (monitor vs. laboratory values), showing a clear linear distribution. The line indicates the regression line, while the shaded gray area represents the 95% confidence interval. Dashed lines mark the 95% prediction limits. HB values are expressed in g/dL, HCT values are expressed in %. The x-axis represents HB or HCT measured using the dialysis sensor, while the y-axis shows the values measured by the laboratory.

Figure 2

Bland–Altman graphs: The graphs (A,B) present the Bland–Altman plots (hematocrit (HCT) on the upside and hemoglobin (HB) on the downside) for values that have been corrected using the adjustment factor. As delineated in the paper, the adjustments are validated by the mean difference being near to zero (continued line) and the majority of plots being between the 95% confidence interval (discontinued line).