Phosphate concentrations change continuously throughout hospitalization; however, it is unclear which available phosphate measures are most clinically important for predicting hospital mortality. Therefore, we investigated phosphate concentrations in association with hospital mortality following admission to the intensive care unit. We retrospectively enrolled all adult patients receiving mechanical ventilation. Phosphate concentrations were divided into three categories: initially measured phosphate (iP); maximum–minimum phosphate values (ΔP); and phosphate arithmetic average (Pmean). In total, 175 patients were enrolled. The hospital mortality rate was 32.6%, and the most common primary diagnosis was respiratory failure. In multivariable logistic regression analyses, the odds ratios for hospital mortality in association with ΔP and Pmean values were 1.56 and 2.13, respectively (p < 0.0001). According to the obtained receiver operating characteristic curve, ΔP (0.75) and Pmean (0.72) each showed a fair predictive power for hospital mortality. In evaluating relative risks, we found that higher concentrations of Pmean and ΔP were each associated with a higher hospital mortality. ΔP and Pmean values were significantly associated with hospital mortality in critically ill patients, compared to iP. These findings showed that throughout hospitalization, it is important to reduce phosphate level fluctuations and maintain appropriate phosphate concentrations through consistent monitoring and corrections.