AES Properties

Reliability: Internal consistency has been found to be good for all three versions of the AES, with alpha ranging from 0.86-0.94 (Marin et al, 1991; Glenn et al, 2002). In a study involving stroke, Alzheimer’s disease (AD), depressed individuals, and elderly controls, test-retest reliability was good for all three versions (AES-S: r = 0.76; AES-I: r = 0.94; AES-C: r = 0.88) except for the group with AD, for which r = 0.44). (Marin, 1991) Inter-rater reliability was only tested for the AES-C and was found to be good (intraclass correlation coefficient = 0.94). (Marin, 1991)

Validity: Compared to elderly controls, differences were found with most of the experimental groups at p<.05 (Marin, 1991):

• AD: AES-S,I,C elevated
• Depression: AES-S,I,C elevated
• RH Stroke: AES-I,C elevated (not AES-S)
• LH Stroke: none elevated
• No significant difference between RH & LH (prediction had been RH>LH)

Most of these differences are in the direction expected. However Bonferroni corrections were not applied.

To test divergent validity with respect to depression, Marin et al administered the HRS-D and the Zung-D to the above groups with AD, stroke, and major depression. Results were mixed, with r ranging from 0.35 to 0.65 for the AES-S, I, and C. (Marin et al, 1991) Kant et al (1998) also presented evidence that the AES may not discriminate apathy from depression. 85% of subjects who were apathetic according to AES-S criteria also met the study BDI criteria for depression. Starkstein et al (2002) also presented evidence that divergent validity may be questionable for depression. Of 21 subjects rated as apathetic in their study, 15 were also rated as depressed by DSM-III criteria. As noted in the introduction, Glenn et al. (2002) adjusted Marin’s definition of apathy to include emotional causes because of the overlap found between apathy and depression.

Divergent validity with respect to apathy and severity of TBI has also been found to be questionable. No significant differences were found among ratings of mild, moderate, and severe TBI groups on the AES-S (N=45) or AES-I (N=37).

To test for convergent validity among the three versions of the AES, Marin et al. (1991) looked at intercorrelations among the three:

• AES-C & I: r = 0.62
• AES-C & S: r = 0.72
• AES-S & I: r = 0.43

In their study of subjects with TBI, Kant et al (1998) also found evidence that the comparability of the AES-S and AES-I may be questionable. In their study the mean AES-S score was 38.1 (SD 7.9) and the mean AES-I score was 50.5 (SD 6.6) (p<0.000001).

To test predictive validity, Marin et al. (1991) had subjects sit in a waiting room with a number of games that they were taught and then encouraged to play. There were no significant negative correlations between the number of games and AES scores. Correlations were low for percent of total time spent playing games:

• AES-S: r = -0.15
• AES-I: r = -0.33
• AES-C: r = -0.40

The response latency to replay games was only reported for one of three games, with r ranging from 0.26-0.45 for the three versions of the AES. For the level of difficulty chosen on a maze:

• AES-S: r = -0.31
• AES-I: r = -0.26
• AES-C: r = -0.37

There have been two attempts to establish cutoff scores for apathy on the AES. Glenn et al (1998) were unable to find a cutoff score that had both adequate sensitivity and selectivity in predicting clinicians’ designation of apathy either on a binary choice or a clinician scale. Starkstein et al (1992) looked at the ability of a neurologist to predict apathy using a cutoff score of > 14. Sensitivity was 66% and specificity 100% (N=12).