A-333 Comparison between measured total carbon dioxide and calculated bicarbonate in hospitalized patients

Abstract

Abstract Background Bicarbonate (HCO3-) is a key indicator of anion excess or deficit and important clinically for assessing acid-base balance. For patient care, HCO3- may be reported as a calculated value from blood gas analysis or measured on automated chemistry analyzers. When measured, the final result includes HCO3-, carbonate (CO32-) and carbamino compounds and thus is reported as total CO2 (TCO2). Anecdotally, some patients have discrepancies between HCO3- and TCO2. However, there is limited data in the published literature assessing agreement between the two methods. Our study aims to compare the calculated HCO3- values with measured TCO2 obtained from simultaneously collected clinical specimens. Methods This study included all results from physician ordered HCO3- and TCO2 between 01/2019 to 12/2020 retrospectively retrieved from the laboratory information system (Cerner). Specimens were collected by nursing staff within 5 minutes of each other. TCO2 was assessed on a Roche cobas 702 (CV = 4.2%) according to the manufacturer’s instructions using an enzymatic phosphoenolpyruvate carboxylase method in lithium heparin tubes. HCO3- was calculated using a Radiometer ABL800 blood gas analyzer from arterial or venous blood gas samples according to the manufacturer’s instructions. Briefly, HCO3- is calculated using the Henderson-hasselbalch equation and the measurement of PCO2 (CV= 1.75%), and pH (CV = 0.06%). Deming regression was performed. The CLIA total allowable error of 4 mmol/L and a reference change value (RCV, Cv Individual = 4%) of -13 and + 14 for TCO2 was used to assess agreement. All data was analyzed in R. Results There were 101,895 paired HCO3- and TCO2 results during the study period. The values ranged from 4 to 56 mmol/L and the mean was 25.7. The mean difference between the calculated HCO3- and measured TCO2 was 1.33 mmol/L. Relative to HCO3-, there were 2,119 (2.1%) TCO2 results with a negative bias exceeding 4 mmol/L and 427 (0.4%)[FC1] with a positive bias exceeding 4 mmol/L. Twenty-two samples exceeded the positive RCV of 14% and 22 samples exceeded the negative RCV. Deming regression revealed a slope of 1.09 (95% CI: 1.087-1.092) an intercept of -2.19 (-2.18 to -2.06), and a Pearson correlation of 0.960 (0.959 to 0.960). We also analyzed differences between HCO3- and TCO2 in arterial and venous samples. The Pearson r for venous samples was 0.956 (0.955 to 0.957) and for arterial samples was 0.966 (0.966-0.966). Deming regression revealed a slope of 1.099 (0.195 to 1.104) and 1.071 (1.068 to 1.074) for venous and arterial samples respectively. The intercept for venous samples was -1.441 (-1.559 to -1.324) and for arterial was -2.005 (-2.072 to -1.938). Conclusions There is bias between HCO3- calculations and TCO2 measurements in simultaneously drawn samples. A small proportion of samples (2.5%) exceed the CLIA total allowable error between the results but very few exceed the RCV. This may due to differences in matrices, with different correlations and slopes observed between arterial and venous specimens. Further studies are required to distinguish the causes of the observed differences between HCO3- and TCO2.