Table 5 illustrates the cases on which the original
algorithm and the revised algorithm disagreed on ASD or
AD classification.
This table shows what non-ASD cases exceeded ASD or
AD cut-off (sensitivity) and what ASD cases did not reach
ASD or AD cut-off (specificity). The algorithms disagreed
on ten cases based on the ASD cut-off. Three cases
exceeded the cut-off for ASD on the original algorithm but
not on the revised algorithm, two with clinical ASD, one
with psychopathy. Conversely, seven cases exceeded the ASD cut-off on the revised algorithm but not on the original
algorithm, four with clinical ASD and three with
schizophrenia. Based on the stricter cut-off for AD, the
algorithms disagreed on eight cases, all of which exceeded
AD cut-off on the revised but not on the original algorithm.
Six of these eight had a clinical ASD classification, the two
others were from the schizophrenia group. This illustrates
the increase in sensitivity of the AD cut-off of the revised
algorithm compared to the original algorithm, but also
shows that the sensitivity and specificity of the ASD cut-off
remain fairly the same, even with other cases exceeding the
ASD cut-off.
The McNemar ch