Sign Up
Login
Try for Free
Login
The COWS (Clinical Opiate Withdrawal Scale) is an 11-item, clinician-administered opioid-withdrawal measure published by Wesson and Ling in 2003 and scored 0 to 48. Emergency, hospital, addiction-medicine, and opioid treatment program teams use it around buprenorphine initiation and withdrawal monitoring. This page covers how to document COWS assessments defensibly, with a fictional sample note.
Clinician-administered (nurses and prescribers); combines measured, observed, and patient-reported findings
Induction and observation teams, addiction consultants, ED and OTP programs, auditors and payers reviewing MOUD records
One flowsheet row per assessment plus a decision note · administration about 2 to 5 minutes
Clinician-administered withdrawal scale (11 unevenly weighted items, scored 0 to 48)
Buprenorphine initiation decisions, serial withdrawal monitoring, precipitated-withdrawal recognition, methadone-to-buprenorphine transfers
Express clinical-copy permission on the published version; not public domain; no law or regulation sets a COWS threshold
The Clinical Opiate Withdrawal Scale (COWS) is an 11-item, clinician-administered measure of apparent opioid-withdrawal severity, published by Donald Wesson and Walter Ling in the Journal of Psychoactive Drugs in 2003. The arithmetic matters more than most pages admit: the eleven item maxima sum to 48, not the 0-to-47 range some secondary papers repeat, and the items are unevenly weighted (four domains reach 5 points, one of them scored only 0, 3, or 5), so the total is an ordinal composite rather than an equal-interval scale. The published severity bands are 5 to 12 mild, 13 to 24 moderate, 25 to 36 moderately severe, and above 36 severe, which means severe starts at 37, a score of 36 is still moderately severe, and 0 to 4 sits below the published bands with no named category to invent. The developers derived those bands from clinical expertise, not statistics, and later validation work confirmed exactly that, so the bands describe severity and were never validated as treatment cutoffs.
Two more distinctions carry the documentation weight. First, licensing: the version distributed through federal resources carries express permission to be copied and used clinically, but that is not public-domain status; current instrument-registry records list a distributor with licensing conditions for some uses (funded research, commercial products, electronic implementations), so copy the authorized form for care and get rights clarification before republishing or embedding it in software. Second, scope: a COWS total estimates withdrawal severity at one moment. It does not diagnose opioid use disorder (that lives in the substance use assessment), establish which opioid was used or when, prove tolerance, guarantee any medication decision, or rule precipitated withdrawal in or out. And its alcohol-withdrawal neighbor, the CIWA-Ar, is a different instrument for a different syndrome; the scores are never comparable or interchangeable.
Emergency departments document COWS around ED-initiated buprenorphine, where the dominant pathway lineage starts at a score of 8 with objective signs; hospital and addiction-consult teams run serial scores through observed initiations, high-dose protocols, and low-dose overlap starts that deliberately use no threshold at all; opioid treatment programs fold it into individualized assessments, noting that the 2024 federal OTP rules require no COWS score for methadone or buprenorphine admission; and office-based and telehealth prescribers use it in observed and home-initiation workflows, where what was clinician-observed versus patient-reported must stay distinct. In Canada, provincial pathways set their own criteria (British Columbia's traditional pathway uses a score of at least 12; Ontario's ED toolkit uses at least 13), and Australia's national guideline and state rules run their own thresholds. The decision trace lands in the medication management note, acute deterioration escalates through a crisis note, and relapse-prevention planning picks up at discharge in a relapse prevention plan.
No statute prescribes a COWS note format, and no law sets a COWS threshold. What survives review is a record that makes the decision trace auditable: what was observed, when, under which named protocol, what was done, and what happened next. Each element below carries the pitfall that most often undermines it.
A time-stamped total with its component values. Record the exact clock time, the total over 48, and every component value, ideally in a flowsheet. Pitfall: a bare total. The items are unevenly weighted, two identical totals can reflect different clinical pictures, and inter-rater agreement on the severity category was only moderate in ED testing, so the components are what make the number defensible.
The objective substrate, separated from report. Preserve what was measured and observed: the actual pulse, visible sweating, pupil appearance in the prevailing light, observed tremor and restlessness, piloerection, tearing. The instrument scores only findings apparently attributable to withdrawal. Pitfall: confounders scored in. Exertion, ambient heat, stimulant effects, chronic pain, anxiety, allergy, and other withdrawal syndromes inflate totals, and the chart should say what was attributed away and why.
Last-use history, with its uncertainty. Substance and suspected substance, route, usual pattern, exact or approximate last use, and exposure to methadone, extended-release opioids, fentanyl, sedatives, or alcohol. Pitfall: charting an opioid-free interval as fact when the history is approximate; with fentanyl in the supply, elapsed-time rules are unreliable and the uncertainty itself belongs in the note.
The named protocol and its exact criterion. Name the pathway with its version or date, and paraphrase its criterion, including any objective-sign requirement. Authoritative pathways legitimately span a score of at least 6, at least 8 with objective signs, at least 12, and at least 13, and low-dose overlap protocols use no threshold gate at all, which the note should state plainly when that pathway is chosen. Pitfall: "COWS 12 required before buprenorphine" charted as law. No US, Canadian, or Australian statute or regulation sets any COWS threshold.
The decision, dose, and exact times. Initiate, defer, switch pathways, observe, or escalate, with the rationale; then medication, dose, route, and administration time, followed by serial scores with elapsed times. Pitfall: "serial COWS per protocol" with no clock times. The trajectory only means something against the dose clock.
Response, including any suspected precipitated withdrawal. Document worsening as a timed change from the pre-dose baseline: the numerical rise, elapsed minutes, new objective findings, vital signs and sedation, the differential (spontaneous progression, inadequate dosing, stimulants, acute illness), treatment, and response. Pitfall: a verdict from one number. Published definitions differ (a rapid rise above 5 with objective signs is one common description; others use at least 6 within an hour), so "suspected precipitated withdrawal" with the delta and reasoning is the defensible phrasing.
Disposition and the safety net. Observation status, naloxone provision, discharge medication or bridge, confirmed or attempted linkage, return precautions, and who owns follow-up. Pitfall: an induction note that ends at the dose. Reviewers trace the record to the handoff, not to the medication order.
COWS DOCUMENTATION BLOCK Date / exact time: [ ] Setting: [ ] Assessor: [ ] Mode: [in person / video; phone report is not an observed COWS] Total: [ ]/48 Components: [all 11 values in the flowsheet] Objective findings: [measured pulse / sweating / pupils / tremor / restlessness / piloerection / tearing, as observed] Attribution notes: [confounders considered and excluded] Last use: [substance; route; time +/- uncertainty; fentanyl, methadone, XR-opioid, sedative, alcohol exposure] Safety: [arousal, respiratory status, vitals, intoxication] Protocol: [name + version/date] Criterion: [paraphrase, incl. objective-sign requirement / low-dose: no threshold gate] Criterion met: [y/n] Decision + rationale: [ ] Medication: [drug, dose, route, exact time] Serial scores: [time -> total, elapsed from dose] Response: [improved / progressing / suspected precipitated withdrawal: delta, minutes, objective signs, differential] Interventions + response: [ ] Disposition: [observation / admit / discharge meds / naloxone / follow-up owner + date / return precautions] Clinician signature / credentials: Date:
Free to use and share, no signup. The PDF includes a one-page cheat sheet with element-by-element pitfalls and a pre-sign checklist; the DOCX is the blank documentation block, ready to adapt. Neither reproduces the scale itself.
Scenario: an observed ED buprenorphine initiation where the first assessment does not meet the local criterion and the second does. The protocol, threshold, and doses are invented to illustrate documentation structure; they are not treatment instructions. All details are fictional.
Patient: T.R., 31 · Setting: Emergency department, observed initiation · Clinician: E. Vance, RN / J. Malik, MD · Note date: 08/13/2026
09:10, first assessment: Reports last use of intranasal opioid believed to contain fentanyl approximately 20 hours ago (time approximate); denies methadone, extended-release opioids, alcohol, or sedatives. COWS 10/48, all 11 components in the flowsheet; objective findings limited to mildly enlarged pupils and facial moisture. Alert, respiratory rate 16, oxygen saturation 98% on room air. The ED's Observed Buprenorphine Pathway (revision March 2026, fictional) requires a COWS of at least 12 with at least one objective sign: criterion not met. Medication deferred; repeat assessment ordered; symptoms and plan reviewed with T.R.
10:00, second assessment: COWS 13/48 (components in flowsheet). Observed: facial sweating, enlarged pupils, frequent position changes, visible hand tremor, tearing. No sedation or intoxication; no alternate acute diagnosis identified. Criterion met. Buprenorphine/naloxone 4 mg sublingual administered 10:08 per the pathway.
Serial response: 11:10 (62 minutes post-dose): COWS 8, partial relief, tremor and facial sweating resolved, no abrupt worsening; second 4 mg dose at 11:18 per the same order set. 13:20: COWS 3, comfortable, ambulatory, no sedation or respiratory change at any point.
Disposition: Discharged with the pathway's bridge prescription, naloxone kit dispensed and teach-back completed, clinic appointment tomorrow 09:00 (confirmed by the ED navigator, named in the chart), and return precautions reviewed, including abrupt worsening after any dose. Follow-up owner: bridge clinic; ED to be notified of any bounce-back.
This sample is fictional and for educational purposes. It does not describe a real patient or record; the protocol, threshold, doses, and scores are invented to show documentation structure and are not clinical guidance.
Writing these after every session? BastionGPT drafts complete notes from bullets, dictation, or a transcript.
Generate a note from bulletsThe US regulatory frame around COWS documentation changed almost completely between 2023 and 2026, and none of it created a COWS requirement. The MAT Act provision of the Consolidated Appropriations Act, 2023 eliminated the X-waiver and its patient caps (LAW), leaving ordinary DEA registration, state law, and the MATE Act's one-time eight-hour training attestation; the 2024 revision of 42 CFR Part 8 (effective April 2024, compliance October 2024) rewrote OTP admission, assessment, and take-home rules and requires individualized methadone dosing that considers withdrawal severity, yet never names COWS or any score for admission or medication (LAW); and telemedicine runs on two coexisting federal pathways as of August 2026: the finalized buprenorphine telemedicine rule and the broader fourth temporary extension of COVID-era flexibilities, which is scheduled to sunset December 31, 2026, while the special-registration framework remains a proposal, not law (all dated facts worth re-verifying before year-end). PDMP duties are state law with state-specific timing and exceptions, and payer rules are their own layer: Medicare's OTP bundles and office-based OUD codes require documentation supporting the billed service, and no national payer policy reviewed makes a COWS value a condition of payment, so a threshold should never be charted as a payer requirement unless the active plan policy actually says it.
Canada and Australia run the same pattern: national guidance stays threshold-neutral while named pathways set conventions. Canada's CRISM 2024 national guideline expressly leaves induction protocols out of scope; British Columbia's BCCSU pathway requires moderate withdrawal (a COWS of at least 12 in its published criterion) for traditional initiation while its low-dose pathway uses no gate, and Ontario's META ED toolkit uses at least 13. Australia's national guideline dates to April 2014 (a fact worth displaying, since it predates the fentanyl-era literature), describes a methadone-to-buprenorphine transfer pathway at a score of at least 13, and sits atop state-based LAW for authorization and records, with New South Wales' new opioid treatment standards taking effect November 5, 2026. On the measurement itself, the psychometrics counsel humility: the validation study centered on mild withdrawal, ED inter-rater agreement on the severity category was moderate at best with only the pulse item reaching substantial agreement, a one-factor model did not fit in a 516-participant analysis, and no minimal clinically important difference has ever been established, all of which is why the components, times, and observations, not the bare total, carry the record. If a translated version is used, name it; listed translations are not automatically validated.
The COWS was published by Wesson and Ling (2003) in the Journal of Psychoactive Drugs. The version distributed through federal clinical resources carries express permission to be copied and used clinically; that permission is not public-domain status, and current instrument-registry records list Mapi Research Trust as distributor with licensing conditions for some uses, including commercial and electronic implementations. BastionGPT is not affiliated with the instrument authors, the journal's publisher, or Mapi Research Trust. This page describes the instrument in original prose and reproduces no scale text, anchors, or scoring materials.
The numbers behind these errors deserve more respect than they get. In paired ED ratings, agreement on the published severity category was 67.5% with a weighted kappa of 0.55, and only the pulse item reached substantial agreement; the validation study's mean peak score was 7.6, meaning the instrument was validated chiefly for mild withdrawal; no minimal clinically important difference has ever been established; and published precipitated-withdrawal incidence spans roughly 0.76% in a 1,200-patient ED trial to 11.5% among hospitalized patients using fentanyl, depending on setting and definition. The BastionGPT Clinical Advisory Board sees the same errors most often in COWS documentation reviews:
BastionGPT is specifically trained, tuned, and clinically tested on behavioral health progress notes and screening documentation.
See how clinicians use it day to day on the AI therapy notes page.
Many BastionGPT users report saving more than 90 minutes per day on documentation.
HIPAA-compliant with a signed BAA on every plan. Your data is never used to train models. BastionGPT drafts, you review and sign.
Eleven unevenly weighted items sum to a total of 0 to 48 (some secondary papers say 47, but the published item maxima add to 48). The developers' bands are 5 to 12 mild, 13 to 24 moderate, 25 to 36 moderately severe, and above 36 severe, so severe starts at 37, and 0 to 4 simply sits below the published bands. Two caveats keep the number honest: the bands came from the authors' clinical expertise rather than statistical derivation, and the score describes apparent withdrawal severity at one moment. It is not a diagnosis of opioid use disorder, evidence of what was used or when, or a validated treatment cutoff.
None, by law: no US, Canadian, or Australian statute or regulation sets a COWS threshold. What exists are protocol conventions that genuinely differ: common ED pathways start at 8 with objective signs, a 2021 federal quick-start guide suggested above 12 with a higher bar when fentanyl is suspected, British Columbia's traditional pathway uses at least 12, Ontario's ED toolkit uses at least 13, one Australian state pathway uses 6 with objective signs, and low-dose overlap protocols deliberately use no threshold at all. The defensible chart names the active protocol with its version or date and paraphrases its exact criterion; "COWS 12 required" with no source is the phrasing that fails review.
As a timed delta, not a verdict. Record the pre-dose score with its components and time, the dose and administration time, the onset of worsening, the repeat score with the numerical rise and elapsed minutes, the new objective findings (not just reported distress), vitals and sedation, the differential (spontaneous progression, inadequate initial dosing, stimulant effects, acute illness), the treatment with times, and the response. Published definitions differ: one common description is a rapid rise of more than 5 with objective worsening, another uses at least 6 within an hour, so "suspected buprenorphine-precipitated withdrawal, nine-point rise with new objective findings 25 minutes after the dose" is defensible where "COWS 18, precipitated withdrawal" is not.
No. The 2024 revision of the federal opioid treatment program regulations (42 CFR Part 8) requires individualized methadone dosing that considers withdrawal severity, opioid exposure, other substances, and treatment history, and it requires documentation supporting admission and treatment, but it never names the COWS or any score for admission or dosing. A program may adopt COWS locally to structure that assessment, which makes it an organizational convention worth following and citing as such. For methadone-to-buprenorphine transfers the scale is more directly relevant, and Australia's national guidance describes a transfer pathway using a score of at least 13, again a dated protocol convention rather than a universal rule.
The instrument sets no interval; the pathway does. Published and institutional examples range from reassessment 30 to 60 minutes after a dose to one-to-three-hour intervals scaled to severity, and fentanyl-era guidance leans toward more frequent checks without naming one mandatory schedule. What matters for the record is the actual clock time of every assessment and dose, so the trajectory reads against elapsed time, plus a stated reason when a reassessment was delayed, skipped, or accelerated. "Serial COWS per protocol" with no times is the version that cannot be audited.
By video, partially: a clinician can observe some findings (restlessness, yawning, tremor, sometimes tearing) and should document the modality and what was genuinely observable. By phone, no: audio cannot observe pupils, sweating, piloerection, tremor, or motor behavior, so a telephone contact should be charted as a patient-reported withdrawal update or run under an explicitly adapted protocol, never as a complete clinician-observed COWS. For home initiations, the note documents the selection rationale, the written protocol and its criterion, teach-back, naloxone, the emergency plan, and how each follow-up was actually conducted (phone report, video observation, or in-person exam). A structured self-report scale is the honest tool for patient-scored home monitoring.
It is deliberately mixed: direct measurement (pulse), observation (pupils, sweating, tremor, restlessness, piloerection), patient report (aches, anxiety, gastrointestinal symptoms), and blended items. That design makes it clinically complete and also explains the reliability data: in paired ED ratings, only the pulse item reached substantial agreement. The scoring rule that keeps it defensible is that higher ratings in most domains require some observable correlate, so document the objective substrate behind every nonzero item. When report and observation diverge, chart both neutrally: what was reported, what was and was not observed, the confounders, and how the discrepancy affected the decision; no published evidence supports labeling a patient as exaggerating from discordant findings alone.
Copying the authorized version for clinical use is expressly permitted: the federally distributed form states it may be copied and used clinically. That permission is narrower than people assume. It is not public-domain status: current instrument-registry records list a distributor (Mapi Research Trust) with licensing conditions for some uses, including funded research, commercial products, and electronic implementations, and registry metadata about the copyright genuinely conflict, which is itself a reason for caution. The safe pattern: use authorized copies in care, chart results freely, link to an authorized source rather than republishing the scale on a public site, and get rights clarification before embedding the instrument in software or a paid product.
Yes. Give it the assessment facts (times, totals and components, objective findings, last-use history with its uncertainty, the named protocol and criterion, doses and times) and it drafts the full decision trace: assessment blocks, criterion-met reasoning, serial response, suspected-precipitated-withdrawal documentation with the timed delta and differential, and the disposition with follow-up ownership, ready for your review. It can also check a finished note for thresholds charted as requirements, missing components or clock times, and verdict phrasing. BastionGPT is HIPAA-compliant with a signed BAA on every plan, and your data is never used to train models.
The instrument facts and compliance claims on this page trace to these sources, last verified August 2026:
Educational content, not legal or billing advice. Sample notes are fictional. Follow your organization's policies and your board, payer, and jurisdiction requirements.