ECR-RS After Cardiac Arrest: A Floor Effect Worth Reading
- **Instrument:** Experiences in Close Relationships — Relationship Structures (ECR-RS), 9 items, 7-point Likert; items 1–6 = avoidance subscale, items 7–9 = anxiety subscale; four reverse-scored (items 1–4). No clinical cut-off in the original instrument.
- **Sample and reliability:** n = 123 out-of-hospital cardiac arrest (OHCA) survivors three months post-arrest, median age 59.9 (IQR 51–67), 84 % male. Cronbach's α = 0.88 total scale; 0.81 (avoidance) and 0.83 (anxiety); rising to 0.92 when items 5–6 were dropped.
- **Structure:** Exploratory factor analysis recovered the two-factor model (eigenvalues 4.61 and 1.18; 60 % variance explained), but items 4–6 loaded weakly or cross-loaded. KMO = 0.50 — marginal.
- **Validity signal:** Floor effects of 31 % (avoidance) and 72 % (anxiety) at the lowest possible score. Anxious attachment correlated weakly-to-moderately with HADS-A (ρ = 0.31) and HADS-D (ρ = 0.32); both subscales correlated negatively with SF-12 MCS (ρ = −0.24 and −0.28). Known-groups hypotheses (gender, partnership, education, psychiatric history) — not supported.
Most attachment questionnaires are validated on undergraduates and community samples. This Danish team — Hansen and colleagues at the University of Southern Denmark and Rigshospitalet — took a 9-item instrument that practitioners reach for in a few minutes and pushed it into a population where attachment is supposed to matter most: people whose hearts stopped, who survived, and who are now trying to live inside a relationship that has been quietly rearranged by mortality.
The ECR-RS held up partly. The two-factor structure recovered, internal consistency was acceptable, anxious attachment tracked the right way with HADS depression and anxiety scores. But the instrument also revealed something a Monday-morning clinician needs to hear: the scale floors out. Seventy-two percent of OHCA survivors picked the lowest possible answer on the anxiety subscale. Thirty-one percent did so on avoidance. Either this is the most securely attached cohort in the literature, or — much more likely — the instrument is failing to read what is actually happening relationally after a traumatic medical event.
How it works
Nine items, each rated 1–7. The respondent is asked about one specific relationship target (parent, partner, friend, "people in general"). Items 1–6 tap avoidance ("I don't feel comfortable opening up to this person", reverse-scored); items 7–9 tap anxiety ("I'm afraid that this person may abandon me"). Compute subscale means. Four reverse-scored items require attention in scoring sheets — easy to miss in a busy clinic. There is no published clinical cut-off; the scale is dimensional. Higher = more insecure on each axis. The instrument takes under three minutes to administer and is free.
In this study, the authors also classified participants into Bartholomew's four quadrants using a median split at 4 on each subscale. Most landed in "secure". The qualitative meaning of that quadrant placement is what the paper interrogates: secure attachment, or a survivor who has decided not to burden the person who watched them die?
In practice
Three things to take from this paper into your sessions tomorrow.
First, the instrument still earns its place for general adult attachment screening — α = 0.88 is solid, and the dimensional output is more clinically useful than forced-category typologies. If you want a quick read on a new patient's relational template, ECR-RS is reasonable. Free, fast, two scores.
Second, do not trust low scores in a medical-trauma population. A post-MI patient, a post-stroke patient, a cancer survivor, an ICU graduate — they may report "secure" attachment on this scale while simultaneously protecting their partner from every internal collapse. Item 5 ("I don't feel comfortable opening up to this person") had the highest endorsement of distress in this sample. Items 7–9 (the anxiety subscale) floored hard. If a survivor scores 1.0 on anxious attachment but their partner is in your other chair describing a man who no longer talks about anything, believe the partner.
Third, interpret cross-loaded items as clinical signal, not noise. The authors flag items 4–6 as psychometrically weak. From the chair, those items read as trust, emotional openness, and felt safety in communication — exactly the post-trauma relational fracture that often hides under a "we're fine" presentation. Use them as conversation prompts, not as data points.
A score of 1.0 on the anxiety subscale in a cardiac arrest survivor is not secure attachment — it is, more often, a survivor who has decided that loving someone now means not telling them anything.
Single-country sample (Denmark), 84 % male, predominantly partnered (94 %) and physically functional enough to complete an online survey three months out — non-responders were sicker. No Rasch or IRT analysis. The validity verdict applies specifically to OHCA survivors and should not be extrapolated to other medical-trauma populations without replication.