653 - Evaluation of Agreement Between TSB- and TcB-Based Rates of Rise for Hyperbilirubinemia Screening

Publication Number: 1630.653

Elena Luo, MassGeneral Hospital for Children, Cambridge, MA, United States; Sara Cherkerzian, Brigham and Women's Hospital, Boston, MA, United States; Leela Sarathy, Mass General Brigham for Children, Boston, MA, United States

Background: Transcutaneous bilirubin (TcB) measurements have demonstrated a general correlation with total serum bilirubin (TSB) measurements and are accepted as a non-invasive option for hyperbilirubinemia screening. Calculating the rate of rise (ROR) of serum bilirubin levels over time has proven to be a useful approach for detecting hemolysis, one of the most common causes of pathologic hyperbilirubinemia, and for determining the need for further testing or treatment. However, to our knowledge there have been no studies investigating the agreement between TcB- and TSB-based rates of rise.

Objective: We conducted this retrospective cohort study to evaluate the agreement between TcB- and TSB-based rates of rise.

Design/Methods: Charts of level 1 newborns with at least one serum bilirubin measurement between January 2024 and January 2025 were extracted via automated report. Newborns with at least two corresponding TcB and TSB pairs (defined as measurements within 2 hours of each other) were selected for the calculation of ROR in mg/dL/hr. The agreement between the ROR of each method was assessed using the Bland-Altman analysis.

Results: Demographics are shown in Table 1. 69 neonates had at least two paired measurements of both TSB and TcB , for a total of 76 paired ROR values. 63 neonates had two paired ROR values, five had three pairs, and one had four pairs. Measurements were collected between 17.25 and 84.02 hours of life (HOL). ROR distributions of TSB and TcB had similar medians (0.18 mh/dL/hr) and IQR (TSB: 0.12-0.21, TcB: 0.13-0.24) (Table 2 ). Individual paired TSB and TcB measurements had a mean difference of 1.27, (95% limits of agreement (LoA) -1.25-3.79) across the range of values and standard deviation (SD, 1.26) and less than the SD of TSB distribution (SD = 2.09) (Figure 1A). Figure 1B shows the difference in TcB and TSB measurements by hour of life. There were 6 (7.9%) differences between TSB and TcB ROR that were greater than 0.2 mg/dL/hr (SD) and represented strong outliers in the Bland Altman plot of ROR measurements . Excluding these points resulted in a mean difference of 0.02, SD = 0.08 (v TSB ROR IQR = 0.09, 1C). Between 24-48 hours of life, ROR agreement had a mean difference of 0.03, SD = 0.08 (v TSB ROR IQR = 0.07, 1D ).

Conclusion(s): Individual TSB and TcB measurements demonstrated a strong correlation, and differences between measurements were within one standard deviation of the TSB distribution. However, the distribution of RoR differences indicated non-constant difference and variance across the range of values, together suggesting the TcB ROR as an unfavorable alternative to TSB.

Table 1. Demographics
Differences in categorical variables tested using Pearson chi square test, or 1 Fisher's exact test (where cell size < 5)

Table 2. Descriptive statistics of the TSB and TcB measurements


Figure 1. Agreement between TSB and TcB measurements using Bland Altman analysis accounting for paired (linked) repeated measures.
The Bland-Altman plot is a scatter plot where the difference between two measurements is plotted on the y-axis against the average of those measurements on the x-axis.
1A) Plot and analysis of paired TcB v. TcB measurements (n = 69 subjects (76 paired measurements), mean difference = 1.27, standard deviation of mean difference (SD) = 1.26 (95% LoA: -1.25, 3.79)). Constant difference and variance across the range of values indicated by differences on averages regression.
1B) Difference between TcB and TSB ROR over hours of life (hour of life calculated as the mid-point between the start and end measurement of the ROR calculation).
1C) Plot and analysis of paired rate of rise (ROR) values. Bland Altman assumptions of constant difference and variance violated given outliers both without and with transformation; 2B represents the plot with data points > |0.2| excluded (n = 6 (7.9%)). N = 64 subjects (70 paired measurements), mean difference = 0.02, standard deviation of mean difference = 0.08 (95% LoA: -0.15, 0.18).
Plot 1D represents the ROR measurements between 24 and 48 hours based on the midpoint between hours of life of the first and last values of the ROR calculation. Data points > |0.2| excluded due to violation of model assumptions (n = 52 subjects (55 paired measurements), mean difference = 0.03, standard deviation = 0.08, (95% LoA: -0.14, 0.19))