Haemoglobin Levels Measured by the Arterial Blood Gas Analyser Versus the Automated Haemoglobin Analyser in Critically Ill Patients

Abstract

Background: Haemoglobin (Hb) is the iron-rich protein component of red blood cells (RBC) that is responsible for at least 98% of systemic oxygen delivery (DO2). Haemoglobin measurements are often performed in the Intensive Care Unit ICU as a surrogate of DO2. Various methods of measuring haemoglobin by spectrophotometry, namely, the arterial blood gas analyser (ABG- A) and automated haemoglobin analyser (AHA), have been used with unknown levels of agreement in our setup. This study intended to establish the agreement between the two methods and their interchangeability. Objective: To assess the interchangeability of the haemoglobin results as measured by the arterial blood gas analyser versus the automated haemoglobin analyser at the Kenyatta National Hospital (KNH)Main ICU. Methods: This was a cross-sectional method-comparison study. Suitable patients had paired sampling for the two forms of haemoglobin estimation. The data was collected using a pretested data collection tool and stored in a password-protected Excel database. Continuous variables were tested for normality using the Shapiro-Wilk test and summarised into means, median, standard deviation, and interquartile ranges. Paired samples t-tests were used to compare means of parametric variables. Categorical variables were presented in tables or percentages. Bias and limits of the agreement were established using the Bland-Altman test. A mean difference of 0 indicated a lack of a fixed bias. A mean difference within 1.96 standard deviations (limits of agreement) indicated how far apart the measurements were in at least 95% of the samples analysed. Results: 132 paired samples from 73 patients were analysed, 62.8% of whom were male. The patients were between 18-77 years, with a mean age of 43.3.The mean Hb levels mean for ABGA was 11.35(+/-2.41SD, 95%CI of 10.93-11.76) whereas that of AHA was 10.88(+/- 2.20SD, 95%CI of 10.50-11.25). The mean difference was 0.47. The Bland Altman analysis indicated a strong agreement with the bias of 0.47 being close to zero and most data points falling within the 95% LOA. The correlation coefficient r=0.83(p-value 0.000) showed a strong association. The total allowable error for Hb was 4.7% Conclusion: There was a strong association and level of agreement between haemoglobin levels measured by the two methods, with the total allowable error falling within the United States Clinical Laboratory Improvement standards margins of +/-7%. Therefore, the two methods were interchangeable at the KNH Main ICU