Choosing cutoff values for correlated continuous diagnostic data to estimate sensitivity and specificity