Agreement between POC glucose meters and blood gas analyzers varies across ICU patient populations

Point-of-care (POC) glucose meters are widely used in intensive care units (ICUs) for rapid glucose monitoring. However, their accuracy may be affected by pathological and physiological factors in different critically ill patients.

Dataset 1 consisted of a three-year quality assurance (QA) practice (2022-2024) using non-ICU samples. Dataset 2 included paired glucose results from patients in general ICU, cardiac ICU, and neonatal ICU collected over six months (July to December 2024), with measurements obtained within a 20-min interval. Agreement of results between glucose meters and blood gas analyzers (BGAs) was assessed using Deming regression, ISO 15197:2013 criteria, and the Diabetes Technology Society (DTS) Error Grid.

In the QA dataset (n = 1476), glucose meter results showed excellent correlation with BGAs (r = 0.9930), with 92.6% (<5.0 mmol/L) and 88.2% (≥5.0 mmol/L) of results meeting ISO criteria. In ICU patients (n = 985), correlation was significantly lower (r = 0.8509), with slightly reduced agreement at higher glucose concentrations (82.5%). Performance varied across ICU populations, with the lowest correlation observed in the general ICU (r = 0.8068). Agreement at critically high glucose levels was markedly reduced in clinical settings compared with QA data (43.2% vs. 74.2%). DTS Error Grid analysis showed that 1.23% of the differences between glucose meters and BGAs fell within moderate- or high-risk zones.

The analytical accuracy of POC glucose meters is significantly reduced in ICU settings and varies among patient populations. Meter performance evaluation should be conducted under real clinical conditions, with particular attention to extreme glucose levels. The performance data should be shared with ICU physicians to support appropriate interpretation of glucose results.