C-SSRS: what it is, who may use it, and what to document

Almost nobody comes to the C-SSRS out of curiosity. It shows up sideways — on a hospital form, in a study protocol, in a colleague’s note — with two questions: may I use it without permission or payment, and is the translation I downloaded right for my clients?

If you or someone close to you is at risk, contact the crisis line serving your location and your local emergency services.

A scope warning: there is no suicide risk management protocol here. You will not read “if the client says yes to question 4, then…”, and be wary of any blog that offers you one. Describing the instrument — its versions, its licence, its evidence, its limits — is in scope; deciding for you what to do at a given level of ideation is not. That line is the rights holder’s own. Its counterpart: any affirmative answer, or any disclosure of risk with or without an instrument, calls for an immediate clinical risk assessment under your own protocol and training.

What it is and where it came from

The Columbia Protocol — the name its rights holder uses publicly — is the Columbia-Suicide Severity Rating Scale. Its official page describes it as support for suicide risk screening through simple, plain-language questions that anyone can ask, whose answers help identify whether someone is at risk, the severity and immediacy of that risk, and the support they need. It came out of a 2007 US National Institute of Mental Health study of adolescents with depression.

The original validation study (2011, American Journal of Psychiatry) analysed three multisite samples of 124, 312 and 237 participants and reported convergent and divergent validity, sensitivity to change, and moderate-to-strong internal consistency on the intensity subscale. The detail most pages skip: its high sensitivity and specificity figures are for behaviour classification — against the Columbia Suicide History Form and an independent evaluation board — not for diagnosis and not for predicting a future suicide.

How it is built, and which versions exist

It measures four separate things: severity of ideation, on five levels; intensity — frequency, duration, controllability, deterrents and reason; suicidal behaviour, distinguishing actual, interrupted and aborted attempts and preparatory acts, with a separate check for non-suicidal self-injury; and lethality, only for actual attempts. An actual attempt does not require an injury: the potential for one is enough.

The five ideation levels, as the rights holder defines them on the full scale, are instrument content, not a traffic light for clinical action:

Level	What it describes
1	Wish to be dead
2	Non-specific active suicidal thoughts, no method or plan
3	Ideation with a method in mind, without intent to act
4	Ideation with some intent to act, without a specific plan
5	Ideation with a specific plan and intent

Administration branches: questions 1 and 2 are always asked, and if 2 is positive, 3, 4 and 5 follow. Recall windows depend on the version — the community screener asks about the past month for ideation and the past three months for behaviour — so recording which one you used separates an interpretable note from a useless one.

The setting picks the version, not preference. The toolkit builder filters by setting, population and language — English or Spanish — and delivers everything from the full scale to pocket cards, self-report screeners and children’s versions; research has five full-scale variants keyed to the study’s window. The modifications adjust triage or messaging, but the core questions do not vary.

There is no total score: there are bands

This is the most expensive misunderstanding. In standard clinical use the C-SSRS produces no sum. Its risk identification page explains that the “score” is low, moderate or high risk depending on where the affirmative answers fall, and that the most worrisome answers are the same in every setting — a recent yes to ideation items 4 or 5, or any behaviour in the past three months; its FAQ defines a positive screen as any high-risk answer. And there sits the sentence that underwrites the scope warning: “it is up to you to write your own triage next steps that agree with your policies and make use of the particular resources available to you.” Not even the rights holder publishes a universal protocol; a blog has no business inventing one.

Licensing: free in healthcare, licensed for clinical trials

The copyright belongs to the Research Foundation for Mental Hygiene, Inc., parent of the Columbia Lighthouse Project. The rights holder calls the protocol and its training free “for use in community and healthcare settings, as well as in federally funded or nonprofit research” (healthcare page), with no cost or licence to put the tools into an EHR and no special permission to use them in your setting.

The exception is drawn just as sharply: clinical trials do require a licensing agreement, issued per protocol, and the request asks for protocol number, phase, indication and number of administrations; the computer-automated patient-reported version is licensed through a commercial third party. We found no published fee schedule. Outside federally funded or nonprofit work, the research page asks you to contact the Project about whether your non-pharma academic study qualifies; and non-pharmaceutical commercial use is not addressed on any rights-holder page we could find. That logic is in the guide to psychological test licensing.

On training, two official statements coexist in tension, both verbatim. The site says training is not required and that no mental health experience is needed: it is free, in 30 languages, from about 20 minutes; for research it is required, with a certificate and re-training every two years. And yet the cover of the full scale warns that it “is intended to be used by individuals who have received training in its administration.” Download the forms from the rights holder’s own site: document-sharing repositories carry unauthorised copies with no verifiable provenance.

The Spanish you download may not be your country’s Spanish

Four distinct claims, and none of them implies the next.

One: a Mexico version is listed. The translations page states that the scale exists in more than 150 country-specific languages, and its list includes “Mexico/Spanish”, flagged as one of the languages with subtitled training.

Two: that validation is linguistic, not psychometric. The process — two forward translations, reconciliation, back-translation, clinician review — controls translation quality, not how the instrument behaves in a Mexican sample. The site notes that “many” of those translations have been linguistically validated — many, not all — and we could not determine whether the Mexican one is among them.

Three: what downloads free is not the Mexico version. Choosing “Spanish” in the toolkit filter delivers US Spanish — the full scale filed as US-Spanish 5.1 and the screener headed “Versión exploratoria - Reciente”; other translations are requested from the Project. On that same page it warns against using a translation validated for one country where it has not been validated: the warning and the default download do not agree. That free screener also contains at least one item with garbled wording — not cosmetic when the question asks about intent.

Four: there is Mexican psychometric evidence, but not of the official scale. The only Mexican study we located, Austria-Corrales and colleagues (2023), reports a Cronbach’s alpha of 0.814 and construct validity for three risk levels in 3,645 participants. But it administered an adaptation: six of twelve items, asked regardless of the first two answers, online and self-administered, in a non-probability sample — and the authors note that with a single measurement “it is not possible to assess predictive power.”

Hence a sentence almost no page writes: we located no Mexican diagnostic-accuracy study of the C-SSRS, none of the official clinician-administered version, and no Mexican norm naming it. Translation is not validation, and reliability is not diagnostic accuracy. Across the region the scope is uneven: Spain (2016; 467 outpatients) and Colombia (2025; 200 patients in Bogotá) used clinical samples; Argentina (2017), students.

What a result does not tell you

Start with the warning on the cover of the full scale: the questions “are suggested probes” and “the determination of the presence of suicidal ideation or behavior depends on the judgment of the individual administering the scale.”

The evidence supports that modesty. In a prospective Swedish study of 804 adults presenting after self-harm, the C-SSRS reached an area under the curve of 0.62 to 0.65 for predicting suicidal behaviour at six months — limited, in the authors’ own words. This is not peculiar to the instrument: Franklin’s 2017 meta-analysis, covering 365 studies, found prediction only slightly better than chance and no improvement across 50 years, and Carter (2017), reviewing risk scales, concluded that no “high-risk” classification was clinically useful.

The purpose is best stated by the Education Development Center’s Zero Suicide toolkit: screening “is not to predict suicide but rather to plan effective suicide care.” Administering a structured instrument documents a reasonable process; it does not predict an individual. That is the principle behind documenting a suicide risk assessment and behind treating a positive PHQ-9 item 9 as a trigger for clinical assessment, not one more line on a form. One widely repeated claim needs downgrading: Columbia says the FDA in 2012 declared its protocol the standard for measuring suicidal ideation and behaviour in clinical trials, but the FDA document says the C-SSRS “would be acceptable,” and is stamped throughout as a nonbinding draft.

What ends up in the note

Documentation rules are local, and the binding ones are where you practise — verify yours. In the US, the Joint Commission’s R3 Report Issue 18 asks accredited behavioral health care organisations to screen with a validated tool, to “document individuals’ overall level of risk for suicide and the plan to mitigate the risk for suicide,” and, when screening, not to change the wording of a validated tool’s questions. Mexico’s standard is a useful worked example: it specifies contents rather than leaving them to habit — down to prognosis, prior study results and the signature on every note — as the guide to Mexico’s clinical record standard unpacks.

Three elements make an administration reconstructable. Which instrument, in which version and window, and how it was administered: that says in one line what “risk scale administered” never says. The result as the instrument expresses it — the endorsed items and the band — with no invented summed score, because in standard clinical use none exists. And how it fits the clinical picture, the problem, the prognosis and what was decided.

A fictional, generic fragment inside a SOAP note; the clinical content comes from your assessment, not this example:

“O: C-SSRS administered (screener version, in interview; window: past month for ideation, past three months for behaviour). Endorses items 1 and 2; denies 3 through 5 and denies behaviour in the window. Band per the instrument: low risk. A: The result is integrated into the clinical assessment and does not replace it.”

The band is a finding, not an automatic disposition: the decision is documented separately, with its reasoning.

Keep the data point from vanishing between administration and note

A well-administered screen and a well-written note can be done on paper; what gets heavy is sustaining it across thirty clients and staying legible two years later. That is where a platform helps with the mechanical part: gesell.ai includes built-in validated scales with automatic scoring — PHQ-9 and GAD-7 among them — whose results sit inside the client’s structured chart, ready for the note, as decision support you evaluate. For an instrument you administer yourself, such as the C-SSRS, the structure holds the other end: a fixed place for version, window, result and reasoning alongside the history, on the same logic as sustained progress measurement. The assessment stays yours; what stops depending on memory is whether it is written down in full.