FFMQ-15 in clinical samples: short, sturdy, and finally validated for mental health service users
- The 15-item Five Facet Mindfulness Questionnaire (FFMQ-15) — a 12-fold compression of the 39-item original, three items per facet — held its five-factor structure in adults already receiving mental health services (n = 233; 115 English-speaking, 118 Chinese-speaking).
- Confirmatory factor analysis fit indices were acceptable: χ²/df = 1.99, CFI = 0.927, TLI = 0.904, RMSEA = 0.065 (90% CI 0.050–0.080), SRMR = 0.060 — all within conventional thresholds for a brief multidimensional self-report.
- Measurement invariance across English- and Chinese-speaking service users was non-significant — meaning scores can be compared across language groups without correction. This is the first time the FFMQ-15 has been formally validated in a Chinese-speaking clinical population.
- The 15-item version takes roughly 3 minutes to complete versus 15+ for the FFMQ-39 — a tradeoff that becomes defensible when a clinician needs to track mindfulness change across many sessions or many patients.
The FFMQ has been the de-facto mindfulness assessment for over fifteen years, but two practical problems have followed it everywhere. First, the 39-item version is too long for routine clinical use — patients balk by week three. Second, validation has clustered in non-clinical samples (students, community adults, meditators), leaving anyone working with actual mental health service users to extrapolate. This Hong Kong–led study, published in Healthcare in late January 2026, is the first to put the 15-item short form through a full CFA in adults who are already in mental health treatment, and to do so simultaneously in English and Chinese.
How the FFMQ-15 works
Five facets, three items each: Observing (noticing internal and external stimuli), Describing (putting words to inner experience), Acting with Awareness (versus autopilot), Non-judging of Inner Experience, and Non-reactivity to Inner Experience. Each item is rated 1–5 (never/rarely true → very often/always true). Total score range: 15–75. Higher scores indicate greater dispositional mindfulness. Some items are reverse-coded (you cannot just sum the raw responses).
The clinical interpretation is facet-by-facet, not just total. A patient with high Observing but low Non-judging — common in trauma populations and OCD — looks "mindful" on the total score while actually showing the rumination-prone profile that needs attention. Acting with Awareness drops first and most reliably under depressive load. Non-reactivity tends to be the slowest to move under MBCT and similar protocols, which is useful prognostic information mid-course.
The Wong et al. study used snowball sampling and online recruitment, so the sample is not a clinic-recruited consecutive series. But participants self-identified as receiving mental health services, and the CFA fit and invariance are clean enough that a busy outpatient practice can adopt the FFMQ-15 with reasonable confidence — particularly if you serve a bilingual population.
In practice
If you are already running MBCT, MBSR, MSC, or any third-wave protocol with a mindfulness component, swap the 39-item version for FFMQ-15 at intake, mid-protocol, and termination. Three time points × 15 items is roughly the same patient burden as one FFMQ-39, with the added benefit that compliance stops eroding by session six.
For trauma and dissociation work, watch the Observing × Non-judging interaction. High Observing without matching Non-judging often flags hyper-vigilant interoception — a profile where standard mindfulness instructions can backfire and where compassion-focused or somatic anchoring usually has to come first. The facet structure of the FFMQ-15 makes this distinction visible in 3 minutes.
For Chinese-speaking patients in EU/RU/US clinics, the measurement invariance result is the practical news. You can compare a Mandarin-speaking patient's pre/post scores against your English-speaking caseload without footnoting the comparison. Until this paper, that comparison required heavy caveats.
A 3-minute, five-facet measure that holds its psychometric structure in actual mental health service users — and across two of the world's most-used languages — is one of the small infrastructural wins that quietly upgrades a practice.
Snowball/online recruitment skews the sample toward higher-functioning service users and self-selected respondents — clinic-floor patients with severe pathology may show different factor performance. Internal consistency was reported at the model level rather than per-facet alpha, so the long-standing concern about three-item subscale reliability remains.