Predicting Ionized Hypocalcemia in Critical Care: An Improved Method Based on the Anion Gap

Background Low ionized calcium (ICa) is prevalent in critical care patients. It is poorly detected by the popular indirect method, which corrects serum total calcium (TCa) for change in albumin. That correction (cTCa) ignores any concomitant change in the anion-complexed fraction of TCa. We tested whether the diagnosis of low ICa can be improved by further correcting for calcium complexation, represented by the anion gap (AG) or its components—sodium, chloride, and total carbon dioxide (tCO2). Methods We retrospectively studied all patients in our intensive care units between 2009 and 2011 with ICa measured on arterial (n = 310) or venous (n = 462) gas panels within 19 min of a comprehensive chemistry panel. Logistic models to predict low ICa and linear models to estimate ICa were derived in the arterial group and validated in the venous group, using either AG (AG model) or its components (Ion model) as predictors, adjusted for TCa and albumin. Results AG and its set of components were each highly significant independent predictors of low ICa. On validation, the logistic Ion model was better than the logistic AG model (ROC curve area ± SE: 0.92 ± 0.02 vs 0.89 ± 0.02; P = 0.008), which, in turn, was far better than cTCa (0.81 ± 0.03; P = 0.0006); the hypocalcemia rates predicted by the models showed good fit with the observed rates. Linear estimates of ICa were too imprecise for clinical use. Conclusions The adjustment of TCa for AG or for sodium, chloride, and tCO2 markedly improves the diagnosis of low ICa. This finding may be useful in guiding ICa testing.