WAI-SR + TFI Cohesiveness in telehealth groups: a minimal two-scale process kit

If you run telehealth groups and want defensible alliance and cohesion data without burdening members with a 40-item battery, this paper gives you the minimal kit and benchmarks against which to score it. WAI-SR + TFI-COH together total 17 items, complete in under five minutes, and both scales held α≥.90 in a small, ethnically diverse, high-stress sample — the exact conditions where psychometrics usually fall apart. The CARE protocol is unusual (mentalizing-focused, parenting-specific), but the assessment package is not — these are off-the-shelf scales that any process researcher or quality-improvement clinician can deploy.

How the kit works

The Working Alliance Inventory–Short Revised (Hatcher & Gillaspy, 2006) collapses Bordin's three-factor alliance model into 12 items; the authors used only two of three subscales — Bond (4 items: warmth, mutual liking, trust) and Goals (4 items: explicit agreement on what therapy is for). Each item is rated 1–7 ("seldom" to "always"). Score the subscale by mean. Norms in psychotherapy outpatient samples place sub-scale means around 5.5–6.0; values below 4.5 by mid-treatment warrant a clinical conversation about rupture.

The Therapeutic Factors Inventory Cohesiveness subscale (Lese & MacNair-Semands, 2000; nine items) targets "I feel a sense of belonging in this group" and "Group members care about me." It is the only Yalom-derived cohesion subscale with published 1-week test-retest of .93 and α consistently in the .90–.94 range. For the full TFI-19 short form, add the other three subscales (instillation of hope, secure emotional expression, social learning) — but cohesion alone tracks a single mechanism that mediates outcome in nearly every group meta-analysis published since 2010.

Administer at sessions 1, 6, and 12 (or first / mid / final, whatever your length). The authors only used T1 and T12 — and explicitly flagged this as a limitation, because rupture-repair patterns happen between those points and you will miss them. Add a mid-treatment time point.

In practice

Three concrete moves for the next group you start:

Score, then look at trajectories, not absolutes. A WAI-SR Bond of 5.5 at session 1 looks fine until you see it drop to 4.8 by session 6 — that is a rupture, even though both scores would be "average" by population norms. Half the value of these scales is the slope, not the intercept.

Watch for the Goals ceiling. If a member endorses 7 across all four Goals items in session 1, treat it as an artefact of the intake conversation, not as alliance data. The CARE study showed 46% of caregivers maxed out Goals after session 1 — these scores are unmovable and tell you nothing. Bond and cohesion are more sensitive change metrics in early treatment.

Dose did not predict outcome. Attendance ranged 1–12 sessions in this sample, and number of sessions attended did not predict end-of-treatment alliance or cohesion scores. This is consistent with broader parenting-group literature and means: do not punish irregular attenders by assuming they have not bonded. The relational signal is in the rating, not the attendance log.