The overall misclassification rate (percentage of observations classified incorrectly) is 41/200 = 20.5 percent. Of the 41 misclassifi-cations, 14 were patients with a clinical score ^0.75 who should have been discharged but were admitted, and 27 were patients with clinical score >0.75 who should have been hospitalized but were discharged. Thus, if hospitalization is considered the “disease,” the rule has a sensitivity rate of 79 percent, and a specificity rate of 80 percent. These rates use the same data to estimate and assess the rule, and for this reason, tend to be somewhat optimistic. The crossvalidated cost gives a more realistic estimate, and was 23 percent (Table 4). The cross-validated sensitivity and specificity values will be given in terms of their cross validated values. birth control pills
Analysis 1 assumes equal prespecified penalty for hospitalizing a patient who should have been discharged, and discharging a patient who should have been hospitalized. In order to decrease the chance of discharging a patient who should be hospitalized, the user can obtain a more sensitive rule, but with lower specificity. Analysis 2 (Table 4) specifies that the prespecified penalty of discharging home a patient who should have been admitted is twice that of hospitalizing a patient who should have been discharged.