The limitation of CANS’ nominal scales to document change is rather self-evident. A child who is highly suicidal and may need two-on-one supervision to prevent serious self-harm would be rated as a three on the relevant CANS’ dimensions (sometimes labeled Suicide Risk and at other times Danger to Self). If treatment is partially successful and the child no longer needs around-the-clock supervision but is still actively suicidal and cannot contract for safety for more than a day at a time, this child would still be rated as a three and their improving clinical presentation overlooked by CANS. This is the primary psychometric problem when taking a continuous variable and splitting it into arbitrary buckets. The problem is compounded when one uses single-item scales. Lyons knew of this potential problem when he started developing CANS, writing: “single-item measures are not particularly sensitive to change” ([27], p. 3).

In Lyons’ initial first-author study using IL CANS-?? (Lyons, Griffin, Quintenz, Jenuwine & Shasha, 2003 [25]), he presents in Table 1 (p. 1631) the initial and discharge means and standard deviations of CANS vs. CAFAS scores, which represents the only known side-by-side comparison of CANS to another assessment tool. CAFAS total score effect size was 0.75. Calculated CANS domains ranged from 0.0 (Care Intensity) to 0.42 (Problems). This suggests that CANS has less than half the sensitivity to change as CAFAS, which itself has similar psychometric limitations (e.g., ordinal data).

Lyons acknowledged this limitation in 2004:

because of the item construction, the CANS is likely less sensitive to change, particularly over short periods of time, than other measures. ([27], p. 13)

Four years later, Lyons claimed it had good sensitivity to change without any additional supporting evidence (cf. [47], p. 806). No CANS’ Publication acknowledges the limitations found in 2003 ([25]). Further confirming this CANS’ limitations, several studies were unable to reject the null hypothesis using CANS ([H], [05]).