PREDICTING RISK FOR HOSPITALIZATION WITH THE CHILD-BEHAVIOR CHECKLIST

Examined a sample of 1,071 cases between the ages of 6 and 17 years to determine whether the Child Behavior Checklist (CBCL) could discriminate ''hospitalized'' from ''never hospitalized'' children. Data were analyzed by linear discriminant function for the total sample and for the sample of boys ages 12 and over. Each sample was randomly divided in half to derive a prediction formula and then crass-validated on the other half of the sample. The predictive power of the samples ranged from 75.5% for the boys-12-and-over derivation sample to 674% for the cross-validation of the total sample. The results indicate that the CBCL distinguished the ''severity level'' of children with mental health problems more accurately than reported in earlier studies. However, the prediction level achieved is not sufficient for individual clinical decisions. The value of the CBCL rests more in its usefulness to program management.