DASS-21: scoring and interpretation in practice

The DASS-21 is one of the few widely used brief scales where the number you get from adding up the items is not the number you compare against the severity table: each subscale is summed, and that total is multiplied by 2. It is common for pages publishing the table to skip that step, and the error always understates. This guide starts there and continues with what almost nobody says: the bands are Australian and date from 1995, no DASS item touches suicidal ideation, and the licence permits less than people assume.

What the DASS-21 is and what backs it

The DASS is, in its authors’ words, “a set of three self-report scales designed to measure the negative emotional states of depression, anxiety and stress”; the full form runs to 42 items. The rights holder’s site also offers “a short version, the DASS21, […] with 7 items per scale.” A point of attribution, because the error is common: the primary source presents the short version as part of the DASS itself and points to the Lovibond & Lovibond manual (1995), not to Antony and colleagues (1998).

The original study (1995) evaluated the DASS in 717 people: Anxiety correlated 0.81 with the BAI and Depression 0.74 with the BDI. The reference validation of the short form is Henry & Crawford (2005), in 1,794 UK adults: the best-fitting model was quadripartite — a general factor of psychological distress plus specific factors — so each subscale also taps that general distress. Internal consistency in the original manual is .81, .73 and .81, per Bados and colleagues (2005).

And the frame that orders everything else: the DASS “is based on a dimensional rather than a categorical conception of psychological disorder,” with no implications for allocating patients to DSM or ICD categories. A score is not a diagnosis, and the people who built the scale say so.

The three subscales, and why Stress is the interesting one

Depression assesses dysphoria, hopelessness, self-deprecation, anhedonia and inertia; Anxiety, autonomic arousal and anxious affect; Stress, chronic non-specific arousal — difficulty relaxing, irritability and over-reactivity.

That third scale is the real differentiator from the PHQ-9 and the GAD-7, with a twist: per the rights holder, the DASS Anxiety scale “corresponds most closely” to the anxiety disorders with the exception of generalized anxiety disorder, while the Stress scale “corresponds quite closely to the DSM-IV symptom criteria for GAD.” If your client lives with sustained tension and irritability, the signal shows up in Stress. One caution: the three scales are moderately intercorrelated, with typical rs of .5 to .7, and are not pure measures.

Administering it

It is self-report; administering it needs no special skills, interpreting it needs training in psychopathology and assessment. The official site publishes no administration time: the “5 minutes” on blogs has no source.

Time window: the past week, rating each state 0 to 3. Changing it is possible, but the rights holder warns the data “will no longer be strictly comparable to the DASS normative data”; if you do, say so in the note.
Minimum age: 14. Below that comes the DASS-Y, for ages 8 to 17, a standalone instrument whose scores are not multiplied by 2.
Download it from the rights holder. It is common for the copies that top a search for a free DASS-21 PDF to come from user-uploaded repositories, with no provenance and sometimes a mis-transcribed table; the official form is free.

The score you record is health information: one more reason to have informed consent settled from the first session, under the data-protection rules where you practice.

Scoring: the step nearly everyone skips

The item map from the official template: Depression 3, 5, 10, 13, 16, 17 and 21; Anxiety 2, 4, 7, 9, 15, 19 and 20; Stress 1, 6, 8, 11, 12, 14 and 18.

Then the rule. The rights holder’s scoring template states it in one line: “For the DASS21, apply template to the sheet, sum the scores, and multiply the sums by 2.” The FAQ explains why: so scores “can be compared to the DASS normative data.” Those norms were built on the 42-item DASS. A fictional example:

Subscale	Raw sum	×2	Band
Depression	8	16	Moderate
Anxiety	6	12	Moderate
Stress	11	22	Moderate

Against the raw sums, that client reads Normal on all three. The bands, applying to the already doubled score (0 to 42 per subscale), come from the 1995 manual and are reproduced with that attribution in Tran, Tran & Fisher (2013):

Severity	Depression	Anxiety	Stress
Normal	0–9	0–7	0–14
Mild	10–13	8–9	15–18
Moderate	14–20	10–14	19–25
Severe	21–27	15–19	26–33
Extremely severe	28 or more	20 or more	34 or more

Two warnings from the rights holder. The cutoffs are published in the manual only, deliberately, “to avoid misinterpretation and reification of these (arbitrary) labels.” And the one that most changes clinical reading: “mild” means the person is above the population mean but still well below the typical severity of people seeking help — “it does not mean a mild level of a ‘disorder’.” For unanswered items there is no fixed standard: the rights holder offers, as a rule of thumb, tolerating up to 1 per 7-item scale.

What a DASS-21 score does not tell you

Nothing about suicide risk, the most important clinical point in this guide. The rights holder is explicit: “none of the DASS items refers to suicidal tendencies because items relating to such tendencies were found not to load on any scale,” adding that determining suicide risk in seriously disturbed people falls to the clinician. Unlike the PHQ-9, whose item 9 asks about ideation directly, a “clean” DASS-21 says nothing. If you adopt it as a routine measure, risk assessment happens separately, in the interview, every time. And if a client discloses risk at any point — whether or not the instrument flagged it — that calls for an immediate clinical risk assessment under your own protocol and training. This guide will not hand you a step-by-step protocol — that comes from the professional guidelines that apply where you practice — but two supports: make sure your client knows the crisis line serving their location and local emergency services, and note that how that assessment gets documented has a guide of its own.

Two further limits: the DASS is transparent with no lie scale, and it is a state measure, not a trait measure — its authors warn a high test-retest correlation is not desirable.

The Spanish evidence, and what is missing in Mexico

If you see Spanish-speaking clients, this section matters most. The rights holder hosts four Spanish translations free: Daza and colleagues, Bados and colleagues (Spain), García Valenzuela (Latin America) and IQVIA (es-US). None is “the official one”: the rights holder warns “we can’t guarantee the quality of the translations.”

Translation is not validation. Only two have a validation study: Daza and colleagues (2002), in a Hispanic clinical sample in the United States, and Bados and colleagues (2005), in 365 Spanish undergraduates, where the three-factor model fit “acceptable but not good enough.” And a distinction nobody draws: reliability and factor structure are not diagnostic accuracy. In Spanish there is plenty of the first and almost none of the second.

In Mexico the sequence is consistent and the conclusion uncomfortable:

Gurrola Peña and colleagues (2006), 200 adults in the Toluca valley, usually cited as the Mexican validation: of the 21 items, 14 survived.
Salinas-Rodríguez and colleagues (2023), 1,613 adults: best fit for a bifactor model, no concurrent or predictive validity assessed.
Salinas-Muñoz and colleagues (2024), 135 undergraduates in Veracruz: three factors, and no cutoffs established.
Domínguez-Lara and colleagues (2024), 509 adults: the best version is a unidimensional DASS-12 of general distress — the construct the K-10 measures head-on —, and Cruz-Flores and colleagues (2024), in 2,315 workers, also fit better with 12 items.
Silva Castillo and colleagues (2025), 1,251 undergraduates at a single private Mexican university: gender invariance and a dominant general factor; the authors name what is missing, norms and ROC-based cutoffs.

Say it plainly: we could not locate any Mexican study establishing the DASS-21’s sensitivity and specificity against a diagnostic interview. The Mexican evidence also repeatedly finds the 21-item three-factor structure needs trimming to fit there. That does not invalidate the instrument as a distress measure; it argues against reading the three-point profile as a fine-grained map.

The only Spanish-language study with ROC-derived cutoffs we could locate is Chilean: Román, Santibáñez & Vinet (2016) proposed 6 for Depression, 5 for Anxiety and 6 for Stress — on the raw 0 to 21 score, not the doubled one, so putting them in the Australian table puts two scales in one column. Their reference standard was a therapist checklist, not a structured interview. The strongest accuracy evidence against SCID comes from Vietnam (Tran and colleagues, 2013): 79.1% sensitivity and 77.0% specificity in 221 women, at a cutoff of 34 or more on the total of all three doubled subscales (0 to 126), not on any single one.

Licensing: what you may and may not do

The questionnaire “is public domain, and so permission is not needed to use it,” and may be copied without restriction. The download page adds the limit: “the scales may not be modified or sold for profit.” Three consequences:

You may charge for the assessment; you may not sell the scale. The FAQ: “While it is not permissible to sell the DASS scales themselves to another person, it is quite acceptable to charge patients for assessment that includes the DASS.”
Online administration depends on who the site is open to. Open to the public — a practice website — no; restricted to a defined group such as enrolled patients, with results returned to the clinician rather than the respondent, yes. Automated interpretation must not be fed back to the respondent either, being “misleading and potentially dangerous.”
The manual is not free. It costs AUS$55 and ships only on paper. The instrument is free; the full normative interpretation is not — the same free/freely-available/public-domain distinction the instrument licensing guide sorts out.

We do not reproduce the 21 items here, though the licence would allow it: the rights holder asks that the instrument circulate among professionals rather than clients or the general public, and we would rather you download it from the source.

Where the score lives in the note

Many jurisdictions regulate what a clinical record must contain, and the details differ — verify the rules where you practice. In Mexico, for instance, NOM-004-SSA3-2012 binds establishments in the public, social and private sectors, consulting rooms included, and names the psychologist among the professionals whose findings are recorded; numeral 6.2.3 requires the progress note to carry relevant results of previously requested diagnostic and treatment services, and 5.10 requires date, time, full name and signature (Diario Oficial). Its wording addresses physicians; psychology practices adopt it as the reference standard, as the Mexican clinical record standard guide sets out.

“Assessment: DASS-21, Spanish Latin American translation, one-week window, self-administered. Raw D=8, A=6, S=11; converted (×2): Depression 16, Anxiety 12, Stress 22. Against the prior administration (D 22, A 16, S 30), a drop on all three, largest on Stress. The DASS-21 does not assess suicidal ideation; explored in interview.”

Record which translation you used — the four differ in wording — and put the raw sum beside the converted one, so whoever reads the chart neither repeats nor skips the step. A band crossing belongs in the O section of your SOAP note and feeds how you measure progress.

Scoring, comparing bands and writing it down is easy with one client and heavy with thirty: gesell.ai includes built-in validated scales with automatic scoring, PHQ-9 and GAD-7 among them, integrated into the structured chart, so today’s score sits next to the previous ones. Two things hold with any tool: the output is decision support you evaluate, and with the DASS-21 digital administration must respect the rights holder’s own restriction.