Results
Sample Characteristics
Table 1 gives the demographic characteristics of the sample. The 2 groups of patients did not differ significantly in age, sex, race or ethnicity, or diagnosis at the time of admission.
Number of Potassium Boluses
The number of potassium boluses received by each group was the primary outcome measure. A sideby- side boxplot (Figure 1) was used to compare the outcome measure for each group. This graphical approach shows median, spread, distributional shape, and any presence of outliers for each group's outcomes.
The shapes in Figure 1 appear to be slightly skewed to the right, because of the presence of outliers. The figure clearly shows that the typical number of boluses used for the treatment group (0.8; SD, 0.85) is lower than the number used for the control group (2.7; SD, 1.9). This difference was significant (t= 8.339; P< .001). Of note, the control group had an extreme outlier (maximum, 13), which is visually apparent in the figure. Data for this outlier were removed, and the parametric procedure was repeated. The difference was still significant (P< .001).
Table 2 gives the descriptive statistics for demographic and clinical variables. The total number of potassium boluses was 218 for the control group and 61 for the treatment group. The serum level of potassium at admission did not differ significantly for the 2 groups (t= 1.943; P= .05). The 2 groups did differ significantly in the serum level of creatinine at admission (t= 4.802; P< .001); however, the levels for all the patients were within the reference range. The mean length of stay in the ICU did not differ significantly between the 2 groups (t=1.381; P=.17).
A correlation matrix was constructed to determine if demographic and clinical variables were associated with the number of potassium boluses used. None of the variables correlated with the number of potassium boluses (P= .05).
Figure 2 shows the mean number of potassium boluses used by type of diagnosis for the control and treatment groups. The mean number of boluses for the treatment group was consistently lower than that for the control group for each diagnosis.