SEP-1: The 3-Hour and 6-Hour Bundle Elements as the Measure Actually Scores Them

A hospital can run a genuinely good sepsis programme — fast recognition, antibiotics inside the hour, a resuscitation team that shows up — and still post a poor SEP-1 score. The reason is almost never the care. It is that SEP-1 is a chart-abstracted composite measure with all-or-nothing scoring, a tightly defined clock start, and a set of data-element rules that decide what counts as documented. The abstractor is not scoring whether the patient was treated well. They are scoring whether the record, read strictly against one named version of a CMS specification, proves each element happened inside its window.

This page is written for that reader: the quality abstractor, the sepsis-programme coordinator, and the quality director who has to explain a fallout. It covers how the measure scores, where time zero comes from, which specification version governs which discharge year, and the abstraction rules that most often turn adequate care into a failed case. It is not clinical guidance. Every clinical value named below is named because CMS’s specification uses it as an abstraction threshold, and it is attributed as such — treatment decisions belong to the clinicians and to the sepsis guidelines your institution has adopted, not to a measure specification.

What SEP-1 is, and what it costs you

SEP-1 is the Severe Sepsis and Septic Shock: Management Bundle (Composite Measure). CMS adopted it into the Hospital Inpatient Quality Reporting (IQR) Program in 2015. It is abstracted by hand from the medical record against the Specifications Manual for National Hospital Inpatient Quality Measures, which CMS publishes on QualityNet.

Two facts about its current position are worth stating plainly, because they change how much attention the measure deserves.

First, SEP-1 is now the only chart-abstracted measure left in the Hospital IQR Program. CMS’s own release notice for Specifications Manual Version 5.18 describes the manual as containing “updates to chart abstracted measures (currently only the Severe Sepsis and Septic Shock Management Bundle).” Every other IQR measure has moved to claims, electronic clinical quality measures, registry feeds or survey. Whatever manual-abstraction capacity your quality department maintains, SEP-1 is now essentially the whole reason it exists.

Second, SEP-1 carries payment. It sits in the Safety domain of the Hospital Value-Based Purchasing Program, which means it feeds the Total Performance Score that converts domain performance into a payment adjustment. Under 42 CFR 412.165(a)(1), Table 1, a hospital needs a minimum of 25 SEP-1 cases in the performance period for the measure to be scored at all. Below that threshold the measure drops out of your domain; above it, every all-or-nothing fallout is a scored zero.

Which specification version applies to which discharge year

This is the single most common source of avoidable abstraction error, and it is the reason an undated description of the SEP-1 bundle is worse than no description at all. SEP-1’s data-element rules change materially between manual versions, and the version is keyed to the discharge date, not to the admission date, not to the date you abstract the chart, and not to the fiscal year.

Specifications Manual version	Applies to acute inpatient discharges	Source and date
V5.17a	1 January 2025 – 31 December 2025	CMS measure-maintenance contractor national provider call, 1 April 2025
V5.18	1 January 2026 – 31 December 2026	Listserv 2025-66-IP, 24 June 2025
V5.18a (addendum to V5.18)	1 January 2026 – 31 December 2026	Listserv 2025-96-IP, 19 September 2025
V5.19	1 January 2027 – 31 December 2027	eCQI Resource Center

Three practical consequences follow.

• A case discharged 31 December 2025 and a case discharged 1 January 2026 are abstracted under different rulebooks. If your abstraction team works a backlog across a year boundary, the version must be selected per case, not per abstraction session.
• V5.18a is an addendum, not a replacement. It was issued mid-period, in September 2025, for discharges that had not yet occurred. CMS’s notice states that it “is required for the Specifications Manual… Version 5.18a, for implementation with acute inpatient discharges effective January 1, 2026, through December 31, 2026.” Among its updates, Appendix A.1 reflects the FY2026 ICD-10 code updates effective for 1 October 2025 and forward discharges, and Table 4.01: Severe Sepsis and Septic Shock (SEP) was updated — that table drives code-based elements of the measure, so an abstraction workflow still pointing at the V5.18 code tables is pointing at a superseded list.
• V5.19 already exists and governs CY2027 discharges. Version changes are announced through the CMS IQR listserv and posted to QualityNet before the discharge period they govern begins; a sepsis programme that learns about a change from a fallout report is learning about it roughly a year late.

CMS held a SEP-1 Version 5.18a Review and Updates webinar on 19 May 2026 specifically to walk through the current guidance and its rationale (listserv 2026-41-IP, 15 May 2026). These review sessions are the closest thing to a plain-language change log the programme produces, and the slides are posted to QualityNet. If your programme does one recurring external thing, make it attending or reading the annual SEP-1 review deck.

The all-or-nothing logic, stated exactly

SEP-1 is a composite measure with all-or-nothing scoring. A case in the denominator passes only if every element applicable to that case was met inside its window. There is no partial credit, no weighting, and no element that matters more than another. Three elements met and one missed scores identically to zero elements met.

Two things follow that are worth saying out loud to clinical leadership, because they are frequently misunderstood:

1. Your SEP-1 rate is not a measure of how much of the bundle you deliver. A hospital delivering 90% of individual bundle elements across its septic population can post a composite pass rate far below 90%, because the failures are distributed across different cases. Reporting element-level compliance alongside the composite is the only way to show programme leadership where the actual gap is.
2. The cheapest improvement is usually documentation, not care. When element-level compliance is high and the composite is low, the constraint is almost always a small number of recurring abstraction technicalities — a missing order rationale, a time stamp that cannot be sourced, a note that contradicts itself. Those are fixable in weeks. Changing resuscitation practice is not.

Time zero: where the clock actually starts

Every window in SEP-1 is measured from a presentation time the abstractor derives from the record. This is the highest-leverage decision in the whole abstraction, because an error here shifts every downstream element at once. The measure uses two distinct clock starts.

Severe Sepsis Presentation Time

Severe sepsis is established, per CMS’s specification, from three things documented together within a defined window: a suspected or documented infection, two or more SIRS criteria, and at least one sign of organ dysfunction. The presentation time is the time at which the last of those criteria is met — not the time the physician wrote the word “sepsis,” and not the time a sepsis order set fired.

Two verified V5.17a abstraction rules illustrate how technical this gets:

• Suspicion of infection can be established by an order alone. Under V5.17a guidance, “a physician/APN/PA order for an FDA-approved biomarker test for sepsis detection can be considered documentation of suspicion of an infection, regardless of whether the result of the test indicates sepsis.” The order is the evidence; the result is irrelevant to that data element.
• Mechanical ventilation has its own timing rule. Where ventilation is the organ-dysfunction criterion, the abstractor uses “the time when mechanical ventilation was started, the earliest time directly associated with the patient being on mechanical ventilation, or the time when the mechanical ventilation changed from intermittent to continuous.” In CMS’s own worked example, an intubation flowsheet entry of “ET placement” at 1600 and a respiratory therapy note at 1730 reading “pt intubated and on a vent” resolves to 1730 — the time directly associated with being on the ventilator, not the airway placement.

Septic Shock Presentation Time

Septic shock is established either by persistent hypotension after fluid administration or by severe sepsis together with an initial lactate at or above the specification’s shock threshold. For that second route — what the specification calls Septic Shock Present criteria (b), severe sepsis present with an initial lactate of 4 or greater — V5.17a guidance is explicit: use the later of the severe sepsis presentation date and time, or the initial lactate level result date and time.

And the lactate result time itself has a documented source hierarchy. The primary source is the lactate result time from the lab; other time stamps intended to identify the lab result time are acceptable only where a policy, key or legend defines the terminology. CMS’s worked example: severe sepsis met at 1600, initial lactate 4.5, no lab “Result Time” available, but a lab report “Verified Time” of 1630 and a report legend stating that verified time stamps indicate the result date and time of labs. The answer is 1630 — the later time, sourced through the legend.

That legend requirement is a concrete, cheap institutional fix. If your laboratory information system emits a time stamp that is not literally labelled “Result Time,” getting a documented legend or policy in place that defines what that stamp means converts an unsourceable time into a defensible one for every future case.

The upstream corollary: time zero is set by the record’s own trail of vital signs, labs and notes, so the quality of your deterioration-detection documentation directly determines how clean your SEP-1 clock is. Programmes that have already tightened early warning score capture and escalation documentation generally find their sepsis abstraction easier, because the recognition-to-escalation sequence is already time-stamped consistently in one place rather than reconstructed from scattered notes.

The bundle elements, as the specification frames them

CMS structures SEP-1 as timed bundles running from the relevant presentation time. As specified, the severe sepsis bundle requires, within 3 hours of severe sepsis presentation, that the record document: measurement of an initial lactate level; blood cultures drawn prior to the administration of antibiotics; and administration of broad-spectrum or other antibiotics.

Where septic shock is present, the specification adds fluid resuscitation with crystalloid at a target of 30 mL/kg, and a 6-hour set requiring a repeat lactate where the initial lactate was elevated and vasopressors for hypotension that persists after fluid administration.

Two ordering rules inside that structure produce a disproportionate share of fallouts:

• Blood cultures must precede antibiotics, and the record must show that they did. This is the element where measure compliance and other institutional priorities pull against each other most visibly — it interacts directly with antimicrobial stewardship programme expectations around culture collection and de-escalation, and with local practice on empiric coverage informed by your cumulative antibiogram. The measure does not adjudicate that tension. It only asks whether the documented collection time precedes the documented administration time.
• The repeat lactate is conditional on the initial one. It is required where the initial lactate was elevated; where it was not, the element does not apply to that case. Abstractors who treat the repeat lactate as universal will over-report fallouts.

Crystalloid: the target ordered volume, which is not simply 30 mL/kg

The fluid element is the one most often described wrongly in internal training material, because the measure does not ask “did the patient receive 30 mL/kg.” It compares what was given against a target ordered volume that the abstractor derives from the orders and notes. V5.17a guidance defines that derivation in detail, and these are the rules verified against CMS’s own worked examples:

• An order for less than 30 mL/kg can be the target volume, if it is justified in writing. A physician/APN/PA order for a lesser volume is acceptable as the target ordered volume where there is an order for that lesser volume as either a specific volume (for example, 1500 mL) or a weight-based volume (for example, 25 mL/kg), and physician/APN/PA documentation within a single source — one note or one order — giving both the volume to be administered and a reason for ordering less than 30 mL/kg. CMS lists non-exhaustive acceptable reasons: concern for fluid overload, heart failure, renal failure, blood pressure responded to a lesser volume, or a portion of the volume given as colloids (which additionally requires an order and documentation that colloids were started or given).
• “Within a single source” is the clause that fails hospitals. A volume in the order and the rationale in a separate progress note does not satisfy it. This is a documentation-design problem, not a clinical one, and it is solved by putting a required rationale field on the fluid order itself.
• Multiple lesser-volume orders are summed. Where there are multiple physician/APN/PA orders for lesser volumes with documented reasons, the abstractor uses the total of those lesser volumes ordered within the specified window of six hours prior through three hours after the triggering event.
• A later full-volume order overrides an earlier lesser one. If a lesser volume is ordered and there is physician/APN/PA documentation indicating the target ordered volume is 30 mL/kg within six hours after the lesser volume was ordered, the 30 mL/kg volume becomes the target.
• Ideal body weight can be the basis, if the physician says so. In CMS’s worked example, a 112 kg patient (30 mL/kg = 3360 mL) with an ideal body weight of 75 kg (30 mL/kg = 2250 mL) and a physician note reading “Bolus sepsis fluids 30 mL/kg per IBW due to obesity” yields a target ordered volume of 2250 mL.
• An unqualified order comment is not a rationale. In another worked example, an order for 1000 mL with the order comment “Monitoring fluid overload” did not establish a lesser target — the IBW note did. Compare with an order comment reading “Limit to 1000 mL due to CKD,” which did establish 1000 mL as the target. The difference is whether the text states a reason for ordering less, not whether it mentions a clinical concern.

Restating the practical rule: an order for less than 30 mL/kg is not automatically a fallout, and an order for 30 mL/kg is not automatically a pass. What determines the outcome is whether the order, the rationale and the administration record line up inside one sourceable trail.

The abstraction technicalities that fail good care

These are the recurring rules that decide cases where the clinical care was not in question.

Conflicting documentation within one note

This rule appears across several data elements — Severe Sepsis Present, Initial Lactate Level Result, Initial Hypotension, Persistent Hypotension and Septic Shock Present — and its effect is consistent. Under V5.17a guidance, where a single physician/APN/PA note contains conflicting characterisations of a finding, the abstractor uses the criterion value if that note indicates the finding is both:

• normal for the patient, due to a chronic condition or medication, or due to an acute condition with a non-infectious source; and
• due to, or possibly due to, an acute condition, an acute-on-chronic condition, infection, severe sepsis or septic shock.

CMS’s own example is a note reading “Hypotensive post medications, possibly r/t sepsis.” The hypotensive readings are used. A hedged phrase written to convey diagnostic uncertainty pulls the case into the measure. Clinicians should not be asked to write less accurately to avoid this — but sepsis coordinators should know it happens, and should not be surprised by cases entering the denominator on the strength of a single “possibly.”

Where the initial lactate was drawn

V5.17a guidance directs selecting Value “1” for Initial Lactate Level Result where the initial lactate was obtained in the operating room, in interventional radiology, during cardiopulmonary arrest, or during procedural or conscious sedation. Sepsis recognised in a procedural area therefore does not abstract the same way as sepsis recognised in the emergency department, and a programme whose audit sample is drawn only from ED presentations will not see this class of case at all.

Transfer from another hospital or ASC

The transfer data element removes cases from the population, so mis-selecting it silently changes the denominator. V5.17a added guidance that the answer is “No” for Acute Hospital Care at Home Program transfers unless documented as an outside hospital. CMS’s worked answers also show that a transfer from a prison infirmary is “No,” while a patient received as a transfer from an outside emergency department is “Yes” even where the patient refused the ambulance and arrived by private vehicle — the mode of arrival is irrelevant to the element.

Sourcing time stamps at all

Recurring across the fluid, lactate and presentation-time elements is a single underlying requirement: a time used in abstraction must be traceable to a defined source. Medication administration record entries, lab result times, flowsheet entries and legend-defined stamps all qualify on their own terms. A time that exists only inside narrative prose, with no corroborating stamp, is the weakest evidence in the record.

Population, denominator and exclusions

SEP-1’s initial patient population is defined by ICD-10 coding and by the clinical criteria the abstractor establishes from the record. Coding is not incidental here: Table 4.01: Severe Sepsis and Septic Shock (SEP) was updated in V5.18a, alongside the FY2026 ICD-10 updates effective for 1 October 2025 and forward discharges.

One exclusion change is worth flagging specifically for anyone abstracting CY2026 discharges. CMS’s V5.18 release notice states that the manual “includes updates to chart abstracted measures (currently only the Severe Sepsis and Septic Shock Management Bundle) and the removal of the COVID-19 exclusion from the Initial Patient Population.”

That matters because V5.17a — which governs CY2025 discharges — carried detailed COVID-19 abstraction guidance, including a rule that documentation of COVID-19 as suspected or present should not be used where there is physician/APN/PA documentation within six hours following that the coronavirus or COVID-19 is not suspected or present. An abstractor who has internalised the CY2025 COVID-19 handling and applies it to a CY2026 discharge is applying a rule that CMS removed. This is precisely the class of error the version table above exists to prevent.

Other exclusion categories the measure applies — among them a directive for comfort measures only, transfer in from another acute care facility or ASC, and various administrative and clinical contraindications — are enumerated in the measure’s own specification. The authoritative, complete and current list of exclusions and their abstraction rules is in Version 5.18a of the Specifications Manual on QualityNet, and it should be read there rather than from any secondary summary, including this one.

What we verified, and what we could not

In keeping with the point this page makes about undated bundle descriptions, here is exactly where its statements come from.

Verified against CMS programme communications and dated primary notices: the version-to-discharge-period mapping in the table above; that V5.18 is the version for CY2026 discharges with V5.18a as its addendum; that the COVID-19 exclusion was removed from the Initial Patient Population in V5.18; that Table 4.01 was updated in V5.18a; that SEP-1 is currently the only chart-abstracted measure in the Hospital IQR Program; and the 25-case minimum at 42 CFR 412.165(a)(1).

Verified against CMS’s measure-maintenance contractor guidance for V5.17a (the 1 April 2025 national provider call, Severe Sepsis and Septic Shock: Management Bundle (Composite Measure) Version 5.17a Review and Updates): every specific abstraction rule and worked example quoted above — the biomarker-order rule, the mechanical-ventilation timing rule, the conflicting-documentation rule, the septic-shock presentation-time “later of” rule and lactate result-time source hierarchy, the initial-lactate location rule, the transfer examples, and all of the crystalloid target-ordered-volume rules.

Not independently verified against V5.18a: whether each of those V5.17a abstraction rules carried forward unchanged into the version that governs CY2026 discharges. qualitynet.cms.gov serves its content through a client-side application that returns no retrievable document text to an automated request, so the V5.18a manual and its release notes could not be read directly for this page. CMS did hold a Version 5.18a Review and Updates webinar on 19 May 2026 and posts those slides to QualityNet. Before relying on any specific rule above for a CY2026 discharge, confirm it against the V5.18a release notes and that webinar deck. Treat the V5.17a rules here as the shape of the reasoning the measure applies, and the current manual as the authority on its present wording.

Running a SEP-1 abstraction programme that survives validation

A practical sequence, in the order that recovers the most fallouts per hour of effort:

1. Pin the version per discharge date. Put the governing manual version on the abstraction worksheet itself, derived from the discharge date, before any element is abstracted. This single control eliminates the whole class of cross-year errors.
2. Re-abstract every fallout independently before accepting it. A meaningful share of apparent failures are abstraction disagreements, not care failures. A second abstractor working blind to the first result is the standard way to find them.
3. Classify confirmed fallouts by element and by cause. Separate “element not delivered” from “element delivered but not documented to specification” from “time could not be sourced.” These three have completely different remedies and completely different owners.
4. Fix the documentation architecture, not the individual. A rationale field on the fluid order, a lab report legend defining time stamps, a sepsis note template that keeps volume and reason in one source — each of these removes a recurring failure permanently. Where a fallout does trace to individual practice, a just culture algorithm is the right instrument for separating system design from individual accountability; an abstraction miss is very rarely a discipline matter.
5. Feed it into the formal quality structure. SEP-1 performance and its improvement work belong in the documented improvement cycle, not in a parallel spreadsheet — the QAPI plan and performance improvement project write-up is where a surveyor will look for evidence that the programme is governed rather than merely tracked.
6. Assign ownership explicitly. Sepsis programmes commonly sit across emergency medicine, critical care, pharmacy, the laboratory and quality. Where it interacts with infection prevention practice, the infection preventionist role is usually the natural convenor; what matters is that one named role owns the composite result and the abstraction rules, not that it is any particular role.
7. Watch the listserv, not the fallout report. Specification changes are announced ahead of the discharge period they govern. Subscribing to the CMS IQR listserv and calendaring the annual SEP-1 review webinar is the difference between adapting in advance and discovering a change in arrears.

SEP-1 sits alongside the other measures that translate hospital performance into payment and public reporting — see the patient safety cluster for the wider set, including how the Hospital Readmissions Reduction Program applies a different penalty arithmetic to a different population.

Frequently asked questions

What is SEP-1?

SEP-1 is CMS’s Severe Sepsis and Septic Shock: Management Bundle (Composite Measure), a chart-abstracted, all-or-nothing composite measure in the Hospital Inpatient Quality Reporting Program that also feeds the Safety domain of Hospital Value-Based Purchasing. It scores whether the medical record documents each applicable bundle element within its window, measured from a presentation time the abstractor derives from the record.

Which Specifications Manual version applies to 2026 discharges?

Version 5.18, released 24 June 2025, applies to acute inpatient discharges from 1 January 2026 through 31 December 2026, together with its addendum Version 5.18a issued 19 September 2025 for the same discharge period. Version 5.19 applies to CY2027 discharges.

What are the 3-hour and 6-hour SEP-1 bundle elements?

As specified, within 3 hours of severe sepsis presentation the record must document an initial lactate measurement, blood cultures drawn before antibiotics, and administration of broad-spectrum or other antibiotics. Where septic shock is present, the specification adds crystalloid fluid resuscitation at a 30 mL/kg target, and within 6 hours a repeat lactate where the initial lactate was elevated and vasopressors for hypotension persisting after fluid administration. The authoritative enumeration and its current abstraction rules are in Version 5.18a of the Specifications Manual on QualityNet.

What is time zero in SEP-1?

There are two. Severe Sepsis Presentation Time is when the last of the three severe sepsis criteria — suspected or documented infection, two or more SIRS criteria, and at least one sign of organ dysfunction — is met. Septic Shock Presentation Time is established either by persistent hypotension after fluids or, where severe sepsis is present with an initial lactate of 4 or greater, by the later of the severe sepsis presentation time and the initial lactate result time.

Does a fluid order for less than 30 mL/kg automatically fail the measure?

No. Under the V5.17a rules, a lesser volume can be the target ordered volume where a physician, APN or PA orders that specific or weight-based lesser volume and documents both the volume and a reason for ordering less than 30 mL/kg within a single source. Acceptable reasons named by CMS include concern for fluid overload, heart failure, renal failure, blood pressure responding to a lesser volume, or part of the volume being given as colloids. Confirm the current wording against V5.18a before relying on it for a 2026 discharge.

Why does our SEP-1 score look worse than our bundle compliance?

Because the measure is all-or-nothing at the case level. Element-level compliance and composite pass rate are different statistics, and high element compliance spread across different failing elements in different cases produces a much lower composite. Reporting both is the only way to locate the actual gap.

Is SEP-1 still the only chart-abstracted measure in the IQR Program?

CMS’s Version 5.18 release notice, dated 24 June 2025, describes the manual as covering “chart abstracted measures (currently only the Severe Sepsis and Septic Shock Management Bundle).” Confirm current status against the latest manual release notice, since programme measure sets change with each annual IPPS rule.

How many SEP-1 cases does a hospital need for the measure to count?

Under 42 CFR 412.165(a)(1), Table 1, the minimum case count for SEP-1 in Hospital Value-Based Purchasing is 25 cases in the performance period. Below that, the measure is not scored for that hospital.

Sources

• CMS, Specifications Manual for National Hospital Inpatient Quality Measures, QualityNet — the authoritative source for the SEP-1 measure specification, its data elements, exclusions and release notes. Note that this site serves content through a client-side application and returns no retrievable document text to automated requests; it must be read in a browser.
• CMS Inpatient VIQR Support Contractor, listserv 2025-66-IP, Release of Specifications Manual for National Hospital Inpatient Quality Measures, Version 5.18, 24 June 2025.
• CMS Inpatient and Outpatient Healthcare Quality Systems Development and Program Support, listserv 2025-96-IP, Addendum Release of Specifications Manual… Version 5.18a, 19 September 2025.
• CMS Inpatient and Outpatient Healthcare Quality Systems Development and Program Support, listserv 2026-41-IP, SEP-1 Webinar: Severe Sepsis and Septic Shock: Management Bundle (Composite Measure) Version 5.18a Review and Updates, 15 May 2026.
• CMS Behavioral Development and Inpatient and Outpatient Measure Maintenance Support Contractor, Severe Sepsis and Septic Shock: Management Bundle (Composite Measure) Version 5.17a Review and Updates, national provider call, 1 April 2025 — the source of every V5.17a abstraction rule and worked example quoted above.
• eCQI Resource Center, National Hospital Inpatient Quality Measures updated Specifications Manual — source for Version 5.19 applying to discharges 1 January 2027 through 31 December 2027.
• eCFR, 42 CFR 412.165, Table 1 to paragraph (a)(1) — minimum case counts for Hospital VBP measures, including 25 cases for SEP-1.

This page describes a CMS quality-measure specification and how it is abstracted. It is written for quality abstractors and sepsis-programme coordinators, and it is not clinical guidance for the treatment of sepsis. Clinical values named above are named because CMS’s specification uses them as abstraction thresholds. Abstraction rules attributed to Version 5.17a were verified against CMS’s 1 April 2025 measure-maintenance guidance and govern CY2025 discharges; CY2026 discharges are abstracted under Version 5.18a, which should be consulted directly on QualityNet.