Background: The serum anion gap (serum [Na+]-Cl-]-[CO2]) is still the first-line approach to metabolic acidosis. However, while it is generally acknowledged that hypoalbuminemia mandates a downward adjustment of the expected anion gap, a specific correction factor for the anion gap in the face of low serum albumin has never been demonstrated. Methods: We reviewed initial laboratory data from 432 consecutive patients admitted or transferred to the medical intensive care unit at Nassau County Medical Center over a 6-month period and correlated the serum albumin with the anion gap and the serum [tCO2] using multivariate analysis. We looked at the anion gap as a function of (albumin), the difference between normal and actual serum albumin, defined as 4.0 - measured serum albumin g/dl. We also assessed [tCO2] as an independent variable. Results: For patients with normal or high serum tCO2, the ratio of change in anion gap ( anion gap) to (albumin) was 1.46 and 1.45, respectively. For patients with serum tCO2 <22 mEq/l this ratio was 1.89. In the latter group, anion gap was best predicted taking both (albumin) and serum tCO2 into account: anion gap = 36.2 - serum tCO2 - 2.3* (albumin) (r = 0.71, p < 0.0001). Conclusion: For intensive care patients with normal or high serum tCO2 (>21 mEq/l) a simple bedside adjustment of the anion gap by subtracting 1.5 times the difference between measured serum albumin and the 'normal' level of 4.0 g/dl gives a close estimate of the actual anion gap. For intensive care patients with serum tCO2 <22 mEq/l, correction of the anion gap is well predicted by adding about twice the (albumin) to the calculated gap.