Evaluating Systematic and Proportional Bias in Point-of-care Glucose Testing: A Correlation and Bland-Altman Analysis

Abstract

Point-of-care testing (POCT) for blood glucose provides rapid results and is widely used in clinical practice; however, concerns remain regarding its accuracy and agreement with laboratory auto-analyser methods. The purpose of this study was to compare glucose measurements obtained using POCT and a laboratory auto-analyser in a secondary healthcare centre within a low-resource setting. A comparison study was conducted using 120 paired blood glucose measurements obtained simultaneously by a POCT glucometer and a laboratory auto-analyser. Descriptive statistics were used to summarise glucose values. Pearson correlation and linear regression analyses assessed the relationship between methods, while agreement was evaluated using Bland-Altman analysis. Statistical significance was set at P < .05. The mean glucose concentration measured by POCT was 6.59 ± 2.27 mmol/L, while that measured by the auto-analyser was 6.33 ± 2.93 mmol/L. A strong positive correlation was observed between the two methods ( r = 0.971, P < .001). Linear regression analysis yielded the equation auto-analyser = 1.249 × POCT − 1.905 ( R ² = 0.943), indicating proportional bias. Bland-Altman analysis demonstrated a mean bias of −0.26 mmol/L, with 95% limits of agreement ranging from −2.03 to +1.50 mmol/L. Greater variability in differences was observed at higher glucose concentrations. POCT glucose measurements show excellent correlation with laboratory auto-analyser results but exhibit systematic and proportional bias, particularly at higher glucose levels. While POCT is suitable for rapid glucose assessment and monitoring, laboratory auto-analyser methods remain the preferred reference for diagnostic and critical clinical decision-making.