Agnew Relationship Measure – 5 (ARM-5)

The ARM-5 is widely used by clinicians engaging in Feedback Informed Treatment and is designed to be used close to the end of a therapy session at regular intervals (either every session, 2nd session or every 3rd session).

The scale is the 5-item version of the Agnew Relationship Measure (ARM; Agnew, Davies, Stiles, Hardy, Barkham, & Shapiro, 1998), and is a self-report measure designed to assess the client-therapist alliance.

Given that therapeutic alliance is among the best predictors of treatment success, this tool can help clinicians identify risk of dropout as well as track any ruptures/repairs in the alliance. The ARM-5 has been shown to be equivalent to other scales such as the Session Rating Scale (SRS) in terms of predicting client outcomes (Bouchard, 2018).

The ARM-5 has been shown to be equivalent to other scales such as the Session Rating Scale (SRS) in terms of predicting client outcomes (Bouchard, 2018). The ARM-5 assesses three dimensions of alliance known to be important for treatment efficacy:

1. Bond
2. Partnership
3. Confidence in therapy

Validity

Cahil et al. (2011) evaluated the ARM-5’s psychometric properties against the full 28-item Agnew Relationship Measure, and found it to have acceptable levels of internal consistency and alternative forms reliability. The ARM-5 was psychometrically approximate to the Core Alliance indexes on the longer form.

Used consistently as part of CBT for depression, the ARM-5 was shown to be predictive of client outcomes, where higher levels of alliance predicted later treatment gains/symptom reduction.

The Cahil et al. (2011) Cheffield Psychotherapy Project sample (n=1073) had a mean rating of 5.76 with a standard deviation of 0.91. Cahil et al. (2011) also was found the clients often rate therapists very highly, and the ARM-5 therefore suffers from ceiling effects.

Scoring and Interpretation

A total raw score ranging from 5 to 35 indicates the overall level of therapeutic alliance. In addition, scores are presented as “mean scores”, indicating the average responses (from 1 to 7). A percentile rank is presented using the mean and standard deviation from the Cahil et al. (2011) sample, indicating the level of alliance compared to a normative sample. The percentile should be interpreted with caution, however, given the data is significantly skewed with evident ceiling effects (max percentile = 91.4).

Research shows there is minimal psychometrically significant distinction between the three subscales (Bond, Partnership and Confidence). Therefore clinicians are encouraged to use the total alliance score, which is the most reliable and useful measure.

1. Bond (item 1) is the measure that encompasses the classic dimensions of client-therapist bond and feelings of therapist supportiveness.
2. Partnership (items 2 and 3) measures agreement on tasks, and agreement on goals.
3. Confidence (items 4 and 5) measures the client’s confidence in the treatment approach, as well as the perceived confidence the therapist has in their own techniques. Confidence has been identified as the strongest predictor of positive outcome, reflecting the therapist’s and client’s joint sense of progress and investment.