How to document a suicide risk assessment

The single number that best explains why this note is hard to write comes from a follow-up study in Psychiatric Services (Louzon et al., 2016): 71.6% of the suicides during the study period occurred among patients who had answered “not at all” to item 9 on their most recent PHQ-9. The negative screen ruled nothing out. If the instrument cannot produce a verdict, the note cannot record one either.

A scope warning, and it is a real one: this guide does not teach you to assess suicide risk and contains no protocol. That protocol comes from your training and from the professional guidelines that apply where you practice, not from a blog. What follows is documentation: what your record should show about the assessment you did perform, how to set down the reasoning that connected what you found to what you decided, and how all of it fits into a progress note.

The note does not document a prediction

Most material on suicide risk is written as if the risk category were the clinical product. It is not. A 2018 meta-analysis in Psychological Medicine put the positive predictive value of a high-risk categorisation at 0.43% (95% CI: 0.014–1.3%) across six cohort studies, a result its authors read as suggesting that suicide risk models “are not a suitable basis for clinical decisions in inpatient settings”. The same group had earlier measured a sensitivity of 56%: nearly half the cases fall on the wrong side of the line.

The VA/DoD clinical practice guideline (version 3.0, 2024) states the consequence without drama: although suicide risk assessments have poor predictive value in identifying those at risk of dying by suicide, they can help identify factors that are contributing to a patient’s risk, which can facilitate treatment planning.

That yields the only useful definition of what your note does. It does not certify that you were right; it evidences that a reasonable clinical process took place. Reviewing US court decisions involving patient suicide, Obegi (2017) frames this as his fifth “probable standard”: the clinician documents the findings of the suicide risk assessment and, when substantial risk exists, the rationale for the selected course of treatment — because those decisions show that documentation is necessary to prove that reasonable care occurred. Necessary, not sufficient, and never a guarantee. Obegi is explicit that he is not writing a clinical practice guideline but legally informed reference points, and his frame is US tort law; as a writing criterion, though, it travels: write the note that lets another clinician — or you in two years — reconstruct what you knew, what you thought, and why you acted as you did.

What your note should be able to show

The APA Record Keeping Guidelines — adopted in 2007, expired by their own terms in February 2017, and never replaced — count among the purposes of a record to “effectively document any decision making, especially in high-risk situations”, and warn of the reverse: information written in vague or broad terms may not be sufficient if more documentation is needed. Five blocks cover what will be asked of you.

What you explored and what the client said. Not the summary (“ideation explored”), but the content of the inquiry and the response, with brief verbatim quotes where they carry clinical weight. A record that says you explored, without saying what you asked or what was answered, does not demonstrate the inquiry: it asserts it.
Which factors you canvassed, protective ones included. If your decision rested on protective factors, they have to be written down, or the decision has no visible support.
Your formulation and the risk level, with the reasoning. Recording the category alone is the habit the evidence above discourages; what gives that line value is the next one, the why. A risk level without reasoning is a label; with reasoning, it is an auditable decision.
The plan, and why it matches what you found. This is the joint most notes skip. Your judgment and your professional framework determine the plan; the note has to make explicit the relationship between the risk you formulated and the measures you took. If the plan is unchanged from last session, write why.
What happened in the session. The concrete actions: the safety plan produced, the conversation about access to means, the calls made, the referral offered or declined. If you contacted someone else, the note should say which limit of confidentiality you relied on — the one agreed in writing in the informed consent — and that what was shared was limited to what was germane to the purpose of the communication, which is what APA Ethics Standard 4.04(a) asks; the disclosure itself rests on 4.05(b). Statutory duties to warn or protect vary enormously by jurisdiction: verify what binds you where you practice, and, in Mexico, what a private practice owes the State and what it does not.

None of this holds if the note is written three days later: the usual documentation mistakes become expensive here first.

Follow-up is part of the note

Obegi’s sixth “probable standard” makes reassessment, not one-time assessment, the expectation: the clinician updates the risk estimate when there are clinically significant changes in the patient’s circumstances or condition, and reassesses at significant treatment junctures. In documentation terms that is three things that are almost never written down. The reassessment point: when you look at risk again, and what would bring that forward. The contact attempts, not only the successful ones — an unanswered call and its time are part of the process, and if they are not in the record they did not happen. And the coordination: who you referred to, whether the client accepted, and whether the appointment happened.

For clinicians in accredited settings, the cleanest one-sentence statement of the requirement comes from the Joint Commission’s R3 Report Issue 18 (updated 3 December 2025), whose NPSG 15.01.01 EP 4 asks organisations to “document individuals’ overall level of risk for suicide and the plan to mitigate the risk for suicide”. That element binds Joint Commission-accredited behavioral health care organisations, not independent practitioners anywhere — but it is a fair reference point for what a record should contain. SAMHSA’s SAFE-T pocket card — an assessment aid this guide neither reproduces nor endorses — makes the same point: documenting the process, including a follow-up plan, is its fifth step, not an afterthought. Any of this works better with a treatment plan carrying verifiable goals behind it. If you work remotely, also record where the client physically is and which emergency services correspond to that location — a point the teletherapy guide develops.

Safety plan versus “no-suicide contract”

Negatively, the evidence is clear: in the randomised trial that compared them head to head (Bryan et al., 2017, Journal of Affective Disorders), crisis response planning outperformed the contract for safety (the trial’s term for the no-suicide contract) in preventing attempts, resolving ideation and reducing inpatient hospitalisation — an estimated 5% attempted suicide with a crisis response plan versus 19% with a contract, a 76% reduction. If your chart contains a sheet where the client promises not to harm themselves, you are documenting the intervention that lost the comparison.

Positively, with a caveat: safety planning combined with telephone follow-up was associated with 45% fewer suicidal behaviours in Stanley et al. (2018), in JAMA Psychiatry, a cohort comparison across nine emergency departments (odds ratio 0.56; 95% CI: 0.33–0.95 over six months), yet the 2024 VA/DoD guideline holds in its Recommendation 7 that there is “insufficient evidence to recommend for or against” those interventions. The honest reading: the contract is discouraged by direct comparison; safety planning is standard practice with supporting, not conclusive, evidence.

What concerns this guide is the record. It should show whether a plan was made, how and with whom, what the client left the session with, what was discussed and agreed about access to means, and which crisis contacts were written into the plan. Name the crisis line serving your client’s location; in the United States that is the 988 Suicide and Crisis Lifeline, elsewhere whatever service actually answers there.

Where it goes in the chart

Jurisdictions differ on the form of the note, and the binding rules are the ones where you practice — verify yours. Mexico’s clinical-record standard is a useful worked example because it specifies contents rather than leaving them to habit: its progress-note clause (numeral 6.2) asks for the evolution of the clinical picture, relevant results of prior studies, diagnoses or clinical problems, prognosis, and treatment with its indications — which a SOAP note already produces — and its formal clause (5.10) requires date, time, full name of the author, and signature on every note. Mexican law adds a patient right to a treatment plan “reviewed periodically and modified according to the patient’s progress” (General Health Law, art. 74 Ter). How much of that norm actually reaches a private psychology practice is unpacked in the guide to Mexico’s clinical record standard.

Here is how a fictional, generic fragment reads inside that structure — the content of each field and the disposition it records come from your assessment, not from this example or from the format:

“Reports increased passive ideation over the past week, in their words, ‘wanting to disappear for a while’; denies plan, denies intent. Problem: depressive episode in follow-up; passive suicidal ideation without plan. Prognosis: guarded short-term; alliance intact, support network active. Plan: safety plan built in session, copy given; access to means discussed, storage agreed with a family member; crisis line entered in the plan; follow-up call agreed; next session moved up to 27/08. Rationale: ideation without plan or intent, two protective factors present, sustained adherence.”

What not to write, and what never to correct

Detail that does not serve the process. The note is not a transcript. Intimate detail that supports no decision does not improve continuity, and it widens what you would have to explain if the record is ever formally requested.
Judgments about the person instead of observations. “Manipulative”, “attention-seeking” and their relatives describe no behaviour, support no decision, and age badly in a record someone else will read.
The silent correction. If something was missing, add it as a new entry, with its real date and time, marked as supplementing an earlier note. Deleting or rewriting the original after an adverse event turns a documentation gap into an integrity problem.

A closing note on PHQ-9 item 9 — a free, public-domain instrument on the official site — because both halves belong together: a positive answer does raise relative risk, with cumulative one-year risk of attempt rising from 0.4% to 4% between “not at all” and “nearly every day” (Simon et al., 2013), while absolute risk stays low and sensitivity stays poor. Any answer other than “not at all”, or any risk disclosure with or without an instrument behind it, calls for an immediate clinical risk assessment under your own protocol and training, as both the PHQ-9 guide and the C-SSRS guide say — the latter describing that instrument and its limits without supplying a protocol; what this article adds is what should end up written about it.

Let the structure stop depending on memory

Writing this note well takes time and a clear head exactly when you have least of both. That is where a clinical platform helps with the mechanical part, and only that. gesell.ai keeps each client’s structured chart — intake, clinical history, treatment plan and notes — and drafts the note from your session record in SOAP, DAP, BIRP or GIRP format, for you to review and sign. The assessment, the formulation and the plan are yours: no tool produces them and none replaces them. What stops depending on your memory is whether the structure is complete.