A six-item ruler for first-episode psychosis: the COMPASS-6 holds its shape
- In 3187 people in coordinated specialty care for first-episode psychosis, the brief clinician-rated COMPASS-6 showed the cleanest fit of all models tested, separating positive and negative symptoms along two clear axes.
- The longer COMPASS-10 held a three-factor structure (positive, negative, affective), but the trimmed six-item version returned the strongest fit indices, suggesting that a shorter ruler measured the construct more sharply, not less.
- Both scales reached good internal consistency by McDonald's omega, and both correlated with an independent symptom index, supporting convergent validity in a real-world clinical sample rather than a research-only cohort.
- The instrument was built for community clinics: a clinician can complete it in minutes, which addresses the practical gap that long research batteries rarely survive into routine early-psychosis care.
Early-psychosis services run on a paradox. The first episode is the moment when accurate, repeated symptom measurement matters most, yet it is also the setting least able to absorb a forty-minute research interview. Most validated psychosis scales were built inside trials, where a trained rater and an unhurried hour are available. In a busy coordinated specialty care clinic, those instruments quietly fall out of use, and the clinician is left tracking recovery by impression. The COMPASS scales were designed to close that gap, and this study asks the only question that matters before adoption: does the short version actually measure what it claims to measure?
The team evaluated two clinician-administered tools – the ten-item COMPASS-10 and a conceptually derived six-item COMPASS-6 – in 3187 individuals with first-episode psychosis treated across the Early Psychosis Intervention Network in the United States. The sample is the headline. This is not a tidy university cohort but a large, demographically mixed clinical population (mean age 21.4; 54.7% male; 43.5% White, 30% Black or African American), assessed in the community settings where the instrument would actually be used. Validation in the deployment environment is rare and valuable.
Confirmatory factor analysis tested a three-factor model (positive, negative, affective symptoms) for the COMPASS-10 and a two-factor model (positive, negative) for the COMPASS-6. Both reached generally acceptable fit, but the COMPASS-6 returned the best fit indices of the set. That result runs against the intuition that more items buy more precision. Here the shorter scale carved the symptom space more cleanly, likely because each retained item was chosen to load on a clear factor rather than to pad a subscale. Internal consistency was good for both, and both correlated – at small-to-medium strength – with the Modified Colorado Symptom Index, the expected pattern for convergent validity between two instruments that overlap but are not redundant.
What the editorial eye should hold onto is the structure, not just the fit numbers. A two-factor positive/negative split is the workhorse architecture of psychosis assessment, and a six-item tool that reproduces it stably can serve as a longitudinal tracker: score it at intake, score it again at follow-up, and the change is interpretable because the underlying factors are stable. That is what makes a screener clinically useful rather than merely brief.
Why brevity earned its place here
Short scales usually trade precision for speed; this one did not. By starting from a conceptual model of which symptoms to retain, the COMPASS-6 kept the two-factor signal intact while shedding the items that blur it – a reminder that item selection, not item count, drives measurement quality.
Where it fits in the digest's theme
This issue circles psychosis and the social brain. A measurement instrument is the unglamorous infrastructure beneath that science: claims about social-cognitive deficits or negative-symptom trajectories only mean something if the symptom axes are measured consistently. A stable, deployable positive/negative ruler is exactly the foundation such work rests on.
A six-item scale that reproduces the positive/negative structure of psychosis is not a compromise; it is a tracker a real clinic can actually use twice.
This was a validation in United States coordinated specialty care, so generalisation to other health systems and to chronic, non-first-episode psychosis is untested. Convergent validity rested on correlation with a single comparator index, and the design did not test measurement invariance across the demographic subgroups or longitudinal stability of scores. Clinician-rated tools also carry rater variance that a factor analysis cannot fully address.