PHQ-15: administration and interpretation in practice

The PHQ-15 is the third member of the PHQ family to reach a therapy practice, after the PHQ-9 and the GAD-7, and the one most often misread. Most summaries get the mechanics right — fifteen items, a 0-to-2 scale, a 0-to-30 range, cutpoints at 5, 10 and 15 — and then make the same clinical error: reading a high score as evidence that the client’s physical symptoms are psychogenic. It measures symptom burden, not its cause.

What the PHQ-15 is and what backs it

It is the somatic symptom scale of the Patient Health Questionnaire, the self-administered version of PRIME-MD that also produced the PHQ-9 and the GAD-7. The official instruction manual describes it as derived from the original PHQ studies and increasingly used to assess somatic symptom severity and possible somatization.

The original validation is Kroenke, Spitzer & Williams (2002), in 6,000 patients across general internal medicine, family practice and obstetrics-gynecology clinics. Its central finding was functional, not diagnostic: as somatic symptom severity increased there was a substantial stepwise decrement in functional status on all six SF-20 subscales, along with more sick days and higher healthcare utilization. That is where the 5, 10 and 15 cutpoints come from. The authors described it as a questionnaire that may be useful in screening for somatization and monitoring symptom severity: nothing more.

Administering it: a 0-to-2 scale and a four-week window

It is self-administered. The fifteen items cover pain, cardiopulmonary and gastrointestinal symptoms, sexual pain, one menstrual item, fatigue and sleep. We do not reproduce the form: download it from the rights holder’s own site.

Two differences from the PHQ-9 change the scoring.

On the PHQ-15 form the scale is 0 to 2, not 0 to 3, and it rates bother, not frequency. The official English anchors are “not bothered at all (0),” “bothered a little (1)” and “bothered a lot (2).” The maximum is 30, not 45, and no single symptom can move the total by more than two points: the score mostly reflects how many symptoms bother the client, not how much any one does. The manual sets out one exception: if you derive the fifteen items from the full PHQ, tired and sleep come from the depression module on the frequency scale and are recoded from 0–3 to 0–2, collapsing “more than half the days” and “nearly every day” into a single 2. Outside that recoding, scoring with frequency anchors does not produce a PHQ-15.
The recall window is four weeks, not two. The stem asks about “the past 4 weeks”; the depression module of the same questionnaire asks about the last two. Administered on the same day, the two instruments cover different periods — write that in the note. And because fatigue and sleep sit in both, giving them separately counts those symptoms twice.

The women-only item

Item d asks about menstrual cramps or other problems with periods, and is printed with a “WOMEN ONLY” flag. Neither the manual nor the official forms say what to do with that item for a male respondent. They do not fix whether the denominator drops to 28, whether the score is prorated, or whether the item is left blank, and we could not locate a source that settles it. So record whether item d applied, and do not compare a man’s 12 with a woman’s as though they measured the same thing.

Scoring and what the bands mean

Sum the fifteen items for a total of 0 to 30. The manual keeps the 2002 study’s cutpoints: 5, 10 and 15.

Score	Severity band
0–4	Below the first cutpoint
5–9	Low
10–14	Medium
15–30	High

Before you copy someone else’s table: the manual names only three bands — low, medium and high where it sets somatic symptom severity, mild, moderate and severe where it lines up the three family instruments — so the 0–4 label is ours, not theirs. It also adds the flag convention: 10 or greater is a “yellow flag,” 15 a “red flag.”

What a high score does not mean

It says nothing about the origin of the symptoms. The largest evidence synthesis on this instrument — Hybelius and colleagues (2024), 305 studies and 361,243 participants — uses “somatoform disorders” as an umbrella term for presentations marked by elevated somatic symptom burden “regardless of whether the etiology of the physical symptoms are fully understood.” The PHQ-15 measures that burden, not its cause. An 18 means the person reports being substantially bothered by many physical symptoms over the past month. It does not mean those symptoms are imaginary.

This is not a semantic quibble; it is patient safety. Treating a high PHQ-15 as proof of somatization is how cardiac, endocrine and autoimmune disease gets missed. Van Driel and colleagues (2018) make the point for older adults: symptoms that in younger people point to medically unexplained presentations may, in older people, originate in somatic disease.

It does not diagnose somatic symptom disorder. No cutpoint does, for a structural reason: the DSM-5 criterion B — disproportionate thoughts, persistent anxiety, excessive time and energy devoted to the symptoms — is not measured by any of the fifteen items. The PHQ-15 counts symptoms; the diagnosis assesses the response to them.

And it does not license you to rule anything out medically. The note should say that a physical-symptom inventory is not a medical evaluation, and should record referral to a physician where that applies.

How well it performs

Pooled internal consistency is good: alpha of 0.81 (95% CI, 0.80-0.82) in the 2024 meta-analysis, which adds a rarely cited detail: three items — menstrual problems, fainting spells and sexual problems — have item-total correlations below 0.40. Test-retest reliability was inconsistent (0.65 to 0.93).

Diagnostic accuracy is a different story, and each figure reads against its own reference standard. For somatoform disorders, reported areas under the curve range from 0.63 to 0.79, and optimal cutoffs run from 6 in the general population to between 11 and 14 in psychosomatic clinics; the authors’ own wording: screening properties were suboptimal. Against a DSM-5 somatic symptom disorder interview in 372 psychosomatic outpatients, Toussaint and colleagues (2020) reported 0.70; combined with the SSD-12 it rose to 0.77, at 69% sensitivity and 70% specificity. Against the MINI, de Vroege (2012) found an optimal cut point of 9 with 56.5% sensitivity and 61.9% specificity.

Look at those “optimal” cutoffs: 6, 9, 11, 14. They travel badly between samples, and none should be read as a disorder threshold. For monitoring, the 2024 meta-analysis puts the minimal clinically important difference at 3 points for the SSS-8, and reports a single PHQ-15 estimate — 2.3, and excluding one item — so 3 points is a prudent floor, not an established threshold. The manual calls its responsiveness to change preliminary, and the authors’ own review (2010) calls it emerging but not yet definitive: a second-line choice for measuring progress in therapy next to the PHQ-9. The counterweight: a COSMIN-methodology systematic review (2017) recommends the PHQ-15 — or the 4DSQ somatization subscale — for somatization in primary care.

The Spanish versions: two exist, and neither is Mexican

If you serve Spanish-speaking clients, the language index on the official site is explicit: the PHQ-15 exists in exactly two Spanish variants, “Spanish for Chile” and “Spanish for the USA.” There is no Spanish-for-Mexico PHQ-15 — there is one for the PHQ-9, the GAD-7 and the PHQ-8 — and no Spanish-for-Mexico full PHQ to derive one from.

The two are not interchangeable: the Chilean form asks how much the problems “lo han afectado,” the US form how much “molestia” the client has had; where one says “articulaciones,” the other says “coyunturas”; where one says “relaciones sexuales,” the other says “penetración sexual.” Your legitimate options are the Spanish-for-Chile form, the Spanish-for-the-USA form, or deriving the fifteen items from a full-PHQ translation by the route the manual authorizes. Whichever you pick, record which one you used: two different wordings do not give strictly comparable scores.

Translation is not validation

The manual puts it plainly: most translations are linguistically valid, but, unlike the English versions, few have been psychometrically validated against an independent structured psychiatric interview.

The only published Spanish-language validations are from Spain. The DEPRE-SOMA study (2010) evaluated 3,362 psychiatric patients with depression and/or anxiety: an alpha of 0.78 and moderate-to-high correlations with the MADRS (r = 0.3-0.7). That is reliability and convergent validity, not diagnostic accuracy: there was no structured-interview reference standard. Robles and colleagues (2024) analyzed 1,495 people from the Spanish general population: factor structure and normative data.

We could not locate any Mexican or Latin American validation of the PHQ-15. We searched PubMed and Europe PMC; SciELO, Redalyc and LILACS were not reachable, so a validation published only there could have escaped us. What does exist: Brambila-Tapia and colleagues (2022) measured somatization with the PHQ-15 in 1,008 Mexican adults, citing the Spanish study as their validation reference, and scored it on a rescaled 1-to-3 metric. So we found no published Mexican reference value on the standard 0-to-30 scale. And in 172 patients from a treatment study for moderate-to-severe somatization, 68% of them Hispanic, Interian and colleagues (2006) found that within that group the instrument related less to medically unexplained symptoms and more to general psychiatric distress.

Licensing: public domain

The official site requires no permission to reproduce, translate, display or distribute the PHQ and GAD-7 screeners or their translations, and the manual uses the technical term: all the measures in its instrument table — the PHQ-15 among them — are in the public domain. Both Spanish forms carry that permission line, with slightly different wording — “mostrar” on the Chilean form, “presentar” on the US one — under attribution to Spitzer, Williams, Kroenke and colleagues and an educational grant from Pfizer Inc. Neither carries a copyright notice. The fine print sits in the terms of use: Pfizer disclaims all warranties, use is at your own risk, and the site asks that the clinician verify every response and make the definitive diagnosis on clinical grounds. None of that limits reproducing, translating or distributing the form. With a free original a click away, there is no reason to use thesis-repository copies, which circulate with the attribution footer cropped off. The wider map is in the guide to psychological test licensing.

Where the score lives in the note

Many jurisdictions require progress notes to record the relevant results of previously ordered diagnostic services, and to carry the date, time, name and signature of whoever wrote them; Mexico’s clinical record standard is one example, and you should verify what applies where you practice. A scale score fits there. A bare number does not:

“PHQ-15: 14. Somatization.”

Against a version that still means something six months later:

“PHQ-15, official Spanish-for-Chile version from phqscreeners.com; 4-week window. Score 14/30, medium severity band per the manual. Item d applicable. Reflects physical symptom burden, not a diagnosis or a medical evaluation; medical assessment of the gastrointestinal symptoms suggested.”

That belongs in the O section of your SOAP note. A number without the version, the language and the window it covers is a standing entry on the list of clinical documentation mistakes: it looks objective and nobody can reproduce it.

Make the instrument work for you

Scoring, recording the version and window, and putting the number where your record-keeping rules require it is easy with one client and heavy with thirty. That is where a platform helps: gesell.ai includes built-in validated scales with automatic scoring — the PHQ-9 and GAD-7 among them — in each client’s structured chart, with today’s score next to the previous ones. Always as decision support you evaluate: the scale contributes the data point; the clinical reading stays yours.