New: independent coding checks for insurers and providers. See how Lens works →

Back

Sepsis Coding Guidelines: Sepsis-2, Sepsis-3 and Payer Clinical Validation

8 min read

The ICD-10-CM guidelines code sepsis from the provider's diagnosis. Payers validate it against Sepsis-3. Where the two split, how MS-DRG 871 gets downgraded, and the documentation that holds.

The ICD-10-CM sepsis coding guidelines and the clinical definition most payers audit against do not ask the same question. The guidelines ask whether the provider documented sepsis. Payers applying Sepsis-3 ask whether the patient had organ dysfunction caused by infection. When the answers differ, the claim drops out of MS-DRG 871 or 872 into a pneumonia or urinary tract infection DRG, and the hospital is left defending a diagnosis it coded correctly. This article sets out where the two definitions split, what the FY2027 guidelines require, how payers apply clinical validation, and what documentation holds up.

Sepsis-2 vs Sepsis-3: what each definition requires

Sepsis-2 came out of the 2001 international definitions conference and kept the 1991 framework: sepsis is infection plus a systemic inflammatory response (SIRS), severe sepsis adds organ dysfunction, and septic shock adds refractory hypotension. SIRS criteria include temperature above 38°C or below 36°C, heart rate above 90, respiratory rate above 20, and an abnormal white cell count.

Sepsis-3, published in JAMA in February 2016 by Singer and colleagues, redefined sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection. Organ dysfunction is operationalized as an acute rise of 2 or more points in the Sequential Organ Failure Assessment (SOFA) score. The task force dropped "severe sepsis" as redundant and found SIRS unhelpful for identifying sepsis, because so many hospitalized patients without infection meet it.

Element Sepsis-2 (2001) Sepsis-3 (2016)
Sepsis Infection plus 2 or more SIRS criteria Infection plus organ dysfunction (SOFA rise of 2 or more)
Severe sepsis Sepsis plus acute organ dysfunction Term retired; this is now simply sepsis
Septic shock Sepsis with hypotension despite fluids Vasopressors to keep MAP at 65 mm Hg or above, plus lactate above 2 mmol/L, despite adequate fluids
Screening tool SIRS qSOFA (2 of: respiratory rate 22 or more, altered mentation, systolic BP 100 or less)
Used by ICD-10-CM code structure, CMS SEP-1 quality measure Most critical care guidance, many commercial payer audits

The practical consequence: a Sepsis-3 "sepsis" patient is, in ICD-10-CM terms, a severe sepsis patient. A Sepsis-2 "sepsis" patient with SIRS and no organ dysfunction may not meet Sepsis-3 at all.

What the ICD-10-CM sepsis coding guidelines require

The coding rules sit in two places in the FY2027 ICD-10-CM Official Guidelines, effective October 1, 2026.

Section I.A.19 settles which definition the coder uses: neither. Code assignment is based on the provider's diagnostic statement that the condition exists, not on the clinical criteria the provider used to reach it. If the documentation conflicts, the coder queries.

Section I.C.1.d sets the sepsis rules themselves:

  • Sepsis: code the underlying systemic infection. If the organism isn't specified, assign A41.9 (sepsis, unspecified organism).
  • Negative or inconclusive blood cultures don't rule out sepsis. Query when the clinical picture supports it.
  • Urosepsis has no default code and is not a synonym for sepsis. Query.
  • Severe sepsis needs at least two codes: the systemic infection first, then R65.20 (severe sepsis without septic shock) or R65.21 (with septic shock), plus codes for the organ dysfunction. The organ dysfunction must be associated with the sepsis; if the link isn't clear, query.
  • R65.2- codes can never be principal.

Sequencing depends on timing and source. Timing is judged against the inpatient order, which is what the POA indicator records:

Scenario Principal diagnosis Then
Sepsis present on admission, meets principal diagnosis definition Systemic infection (e.g., A41.9) Localized infection, R65.2- if severe, organ dysfunction codes
Sepsis develops after admission for a localized infection Localized infection (e.g., pneumonia) Systemic infection, R65.2- if severe
Sepsis following a procedure Site-specific postprocedural infection code (T81.4-) T81.44 for sepsis following a procedure, organism code; T81.12- replaces R65.21 for postprocedural septic shock
Noninfectious process (e.g., trauma) leads to sepsis Noninfectious condition, if it meets the principal diagnosis definition Infection; R65.1- is not added when R65.2- is coded

The FY2027 guidelines don't change this framework. Check the bold-text revisions in the PDF each year anyway, and see our FY2027 ICD-10-CM update reference for this cycle's new codes.

How the definition moves the MS-DRG

Sepsis principal diagnoses group to MS-DRGs 870 to 872 (septicemia or severe sepsis, split by mechanical ventilation over 96 hours and by MCC status). A pneumonia principal diagnosis groups to MS-DRGs 193 to 195, and a kidney or urinary tract infection to 689 or 690. The payment gap is large: in the FY2027 relative weights, MS-DRG 871 sits around 1.93 and MS-DRG 872 around 1.01. Where the same patient lands depends on whether sepsis survives review.

Worked example. A 68-year-old is admitted from the ED with community-acquired pneumonia. Temperature 38.6°C, heart rate 112, white count 15,200, blood pressure normal, creatinine at baseline, mentation normal, on 2 liters of oxygen with no change from home. The attending documents "sepsis due to pneumonia."

  • Coded per guidelines: A41.9 principal, J18.9 secondary. MS-DRG 871 or 872, depending on whether an MCC is documented. Correct coding: the provider stated the diagnosis.
  • Payer applying Sepsis-3: SIRS is met, but no organ dysfunction is documented. The payer removes A41.9, re-sequences pneumonia as principal, and pays MS-DRG 193 to 195.

Same patient, different record. Creatinine rises from 0.9 to 2.1 mg/dL and the attending documents "acute kidney injury due to sepsis from pneumonia." Now it codes A41.9, J18.9, R65.20 and N17.9. This patient meets Sepsis-2 severe sepsis and Sepsis-3 sepsis, the organ dysfunction is linked in the note, and the claim is much harder to deny.

The difference between the two charts isn't the coding. It's whether the record shows organ dysfunction and ties it to the infection.

Sepsis patients often arrive through the ED, and status matters before the DRG does. A patient who stays in observation never reaches an MS-DRG at all; our article on observation, inpatient and ED status covers how that decision is made. The two-midnight rule reference sets out Medicare's standard for it.

How payers apply clinical validation

Clinical validation is a separate process from coding. It asks whether the patient truly had the documented condition, and it is performed by clinicians, not by applying the coding guidelines. The ACDIS clinical validation white paper (2021 update) summarizes the AHA Coding Clinic position from Fourth Quarter 2016: coding must be based on provider documentation, and a reviewer's disagreement with the diagnosis is a clinical issue, not a coding error.

That distinction is exactly where sepsis denials live. The coder was right; the payer says the diagnosis wasn't clinically supported.

Commercial payers moved to Sepsis-3 years ago:

  • UnitedHealthcare began applying Sepsis-3 in 2018 and, from July 1, 2021, said it would review sepsis claims for Medicare Advantage and commercial members on a pre-payment or post-payment basis against its sepsis clinical guideline, most recently approved May 7, 2026.
  • Cigna announced in early 2020 that it would validate adult sepsis DRG claims against Sepsis-3 and adjust the DRG when sepsis isn't supported.

Traditional Medicare is in a different position. CMS's SEP-1 quality measure still uses a SIRS-based severe sepsis definition, and Medicare has not adopted Sepsis-3 for payment. That may change. The HHS Office of Inspector General's work plan item Medicare Inpatient Hospital Billing for Sepsis (OEI-02-24-00230), announced in March 2024, is analyzing 2023 inpatient sepsis claims and estimating what Medicare would pay if the narrower Sepsis-3 definition applied. As of October 2026 the OIG lists it as active, with no report published.

Where the disagreement is legitimate

Neither side is simply wrong. Payers argue that Sepsis-3 is the current clinical consensus and that SIRS-based sepsis sweeps in ordinary infections paid at a sepsis rate. Hospitals and many clinicians argue that the ICD-10-CM code set, the federal quality measure and many contracts still recognize the older definition, that early sepsis without measurable organ dysfunction is still treated aggressively, and that a payer can't retroactively impose a clinical definition its contract doesn't name.

The contract is often the deciding document. If a commercial agreement specifies Sepsis-3 or the payer's clinical guidelines, an appeal arguing Sepsis-2 is weak. If it's silent, the hospital has a stronger argument that the coding guidelines and the provider's diagnosis govern. Have your managed care team pull the clinical validation and audit language from each major contract before building appeal templates.

How to document sepsis so it survives review

The goal isn't to make every infection into sepsis. It's to make the record show what the physician saw.

  1. Adopt one organizational clinical definition. The ACDIS paper recommends that medical staff, CDI and coding agree clinical indicators for high-risk diagnoses. For sepsis, most organizations now align with Sepsis-3, so documentation that meets it also meets the older definition.
  2. Document the organ dysfunction and link it. Name the dysfunction (acute kidney injury, acute respiratory failure, encephalopathy, thrombocytopenia, hypotension requiring vasopressors) and state that it's due to sepsis. The guidelines require that association before R65.20 is assigned.
  3. Show the numbers. Baseline and peak creatinine, platelets, bilirubin, Glasgow Coma Scale, oxygen requirement and mean arterial pressure are the SOFA inputs a reviewer will look for.
  4. Name the source. "Sepsis due to pneumonia" or "sepsis due to E. coli pyelonephritis," never "urosepsis."
  5. Document treatment intensity. Fluid boluses, broad-spectrum antibiotics within the hour, serial lactates, vasopressors and ICU transfer show the condition was treated as sepsis.
  6. Follow up "sepsis ruled out." If sepsis is suspected on admission and ruled out later, the discharge summary should say so. A sepsis code on a record that resolves to "no sepsis" is a validation finding waiting to happen.
  7. Query compliantly. Clinical validation queries should present the indicators for and against the diagnosis and allow "unable to determine" and "ruled out" as answers. A query that only asks the physician to confirm sepsis is leading.

Run a quarterly sample of your sepsis principal diagnosis cases against your organizational definition. Track the rate that would fail Sepsis-3 separately from the coding accuracy rate, because they measure different risks.

Where AI-assisted coding fits

Sepsis is a poor fit for coding without human review and a good fit for software-assisted review. The coding step is rule-based, but the risk sits in clinical validation, which is a judgment about the record.

Software can help by reading structured data alongside the note: flagging a sepsis diagnosis with no SOFA-relevant abnormality in the labs or vitals, a documented organ dysfunction that isn't linked to the sepsis, or a "rule out sepsis" that was never closed. Those cases go to CDI before the claim drops. That's how we'd expect an inpatient tool to be used: AI-assisted coding with human sign-off, and a credentialed coder or CDI specialist owning the final code. When evaluating any tool on this, ask for DRG-level accuracy rather than code-level accuracy; our methodology for measuring coding accuracy explains why that unit matters. For the wider picture of where automation currently stands, see how accurate AI medical coding is in 2026.

Frequently asked questions

Do the ICD-10-CM sepsis coding guidelines follow Sepsis-2 or Sepsis-3?

Neither, for code assignment. Guideline I.A.19 says codes are assigned from the provider's documented diagnosis, not from clinical criteria. The code structure still separates sepsis (A41.-) from severe sepsis (R65.20), which reflects the older Sepsis-2 framework.

Can a coder decline to code sepsis that doesn't meet Sepsis-3?

No. If the provider documents sepsis, the coder codes it. If the record conflicts with the diagnosis, the coder or CDI specialist queries the provider. Clinical validation is a separate process from code assignment.

Is "severe sepsis" still a valid diagnosis to code?

Yes. Sepsis-3 retired the term clinically, but ICD-10-CM still uses R65.20 for severe sepsis without septic shock. Assign it when the provider documents severe sepsis, or documents sepsis with an associated acute organ dysfunction.

Does Medicare use Sepsis-3 to deny claims?

Traditional Medicare has not adopted Sepsis-3 for payment, and SEP-1 uses a SIRS-based definition. Medicare Advantage plans and commercial payers may apply Sepsis-3 under their own policies. An OIG review of Medicare sepsis billing against Sepsis-3 was still listed as active in October 2026.

This article is general education on coding and documentation practice. It is not a coding determination or clinical guidance for any specific patient. Code assignment depends on the provider's documentation, the current ICD-10-CM guidelines and applicable payer contracts.

See it on your own charts.