Anion gap (with albumin correction)

Base anion gap = sodium − (chloride + bicarbonate). When potassium is supplied, the potassium-inclusive anion gap = (sodium + potassium) − (chloride + bicarbonate); it runs about 3.5–5 units higher and uses a higher reference interval, so the two forms are shown separately. When albumin is supplied, the albumin-corrected anion gap = anion gap + 2.5 × (4.0 − albumin in g/dL): each 1 g/dL of albumin below the 4.0 g/dL baseline adds 2.5 mEq/L back to the gap (Figge 1998), unmasking a high-anion-gap acidosis that hypoalbuminemia would otherwise hide. Values are reported in mEq/L (numerically equal to mmol/L for these monovalent ions).

The anion gap is a diagnostic/classification index, not a graded severity or outcome score — do not read it as a mortality or acuity score. It has NO interpretation bands here because reference intervals are strongly method-dependent (flame photometry gave ~12 ± 4 mEq/L; modern ion-selective electrodes shifted it down to ~6 ± 3, and verified lab intervals range widely, e.g. 10–18 mmol/L K-exclusive in Chionh 2022): always classify against the reporting lab's own reference interval, not a fixed cutoff. The potassium-inclusive form runs ~3.5–5 units higher and requires a correspondingly higher reference interval — never compare a K-inclusive value against a K-exclusive range. The albumin correction uses the Figge 1998 slope of 2.5 mEq/L per 1 g/dL below a 4.0 g/dL baseline (equivalently 0.25 per g/L below 40 g/L); a small number of sources use a 2.3 coefficient or a 4.5 g/dL baseline (results are clinically near-identical) — this implementation uses 2.5 and 4.0, the most common clinical form. Correcting for albumin increases sensitivity, not specificity, and is an adjunct to — not a replacement for — direct measurement of lactate, ketones, etc. Pediatric-first flag: the 2.5 coefficient and the reference intervals were derived predominantly in ADULTS (Figge's cohort was adults); they are applied to children by convention, not from pediatric-derived data. Total CO₂ from a basic metabolic panel is used interchangeably with HCO₃ (a ~1–2 mmol/L offset, conventionally ignored). [NEEDS SOURCE]: the na (100–180), cl (70–130), hco3 (3–45), k (1.5–9) mEq/L and albumin (1.0–6.0 g/dL) input limits are engineering input-validity bounds, not thresholds from a specific publication. Garbage-in caveat: spurious electrolyte values (pseudohyponatremia, bromide interference with the chloride assay) distort the AG directly.