摘要Objective: To investigate the effect of the variability of Nas, Cls, Ks, lactate values and sodium effect (NaEffect), chloride effect (ClEffect), non-lactate strong ion difference (SIDnl) values calculated according to Stewart's approach on predicting intensive care unit (ICU) mortality. Methods: The study was conducted on 1539 patients, retrospectively. Serum Na (Nas), serum Cl (Cls), serum K (Ks), serum Ca (Cas), serum Mg (Mgs), lactate, pH values and SIDnl, NaEffect, ClEffect, APACHE Ⅱ (first, last), and SOFA (first, last) scores were recorded. Radiometer ABL 800 (Denmark) was used for blood gas analysis. The variability of each parameter was calculated. The effect of variability of each parameter on 30-day ICU mortality was analyzed. Results: The variability of lactate (P<0.001, OR=0.580, 95％ CI=0.505-0.652), pH (P=0.001, OR=0.004, 95％ CI=0.000-0.104), NaEffect (P<0.001, OR=0.550, 95％ CI=0.378-0.592), Ks (P<0.001, OR=0.385, 95％ CI=0.244-0.565) values were protective factors of ICU mortality and Cls value was a risk factor (P=0.004, OR=1.095, 95％ CI=1.024-1.164). Variability of ClEffect, SIDnl values did not affect ICU mortality. Conclusions: The variability of electrolytes is important. Electrolyte, effects, and lactate variability can guide treatment and fluid applications in ICU.