Written and maintained by CASRAI Editorial Board
Last updated
An AHRQ Patient Safety Indicator (PSI) is not a clinical determination that a safety event happened. It is an administrative-data screen — a rule applied to the ICD-10-CM/PCS codes and Present on Admission (POA) flags already sitting on a hospital’s claims — that flags cases likely to represent a preventable complication. A flagged case is a lead for review, not a confirmed harm. That distinction matters for infection preventionists, patient-safety officers, quality directors, and risk managers because PSI rates feed public reporting, CMS Star Ratings, and the Hospital-Acquired Condition (HAC) Reduction Program — so a program that treats every flag as a proven event, or that can’t explain why a case was excluded, is working from the wrong starting assumption.
This guide covers how any individual PSI is structured, how AHRQ’s PSI 90 composite aggregates several of them into one score, the numerator/denominator/exclusion logic behind the most-cited component indicators, and where PSI mechanics actually diverge from the HAC payment provision they’re routinely confused with.
How Every PSI Is Built: Numerator, Denominator, Exclusions
Each PSI follows the same three-part structure, whether it’s reported as a standalone measure or rolled into a composite:
- Denominator — the population at risk: typically all discharges (or all surgical discharges, for the perioperative PSIs) meeting age, principal-diagnosis, and length-of-stay criteria for that specific indicator. A patient who was never eligible for the complication being measured — because they didn’t have the qualifying procedure, for instance — doesn’t enter the denominator at all.
- Numerator — discharges within that denominator population that also carry a qualifying secondary diagnosis or procedure code for the complication, subject to POA rules.
- Exclusions — cases removed from the denominator (not just the numerator) before the rate is calculated: common categories include a principal diagnosis that already represents the complication, transfers from another acute-care facility, obstetric or neonatal cases on adult-focused indicators, and very short stays where the complication couldn’t plausibly have developed during this admission.
POA does most of the load-bearing work inside the numerator. As covered in detail in CASRAI’s guide on the Present on Admission (POA) Indicator, a diagnosis coded Y or W is treated as already present at admission and is generally excluded from a PSI’s numerator; diagnoses coded N or U are the ones that can trigger a flagged case. This guide doesn’t re-cover the Y/N/U/W coding mechanics themselves — see that guide for the coding and documentation side. What matters here is what happens once a diagnosis clears that POA gate: it still has to survive the indicator-specific exclusion list before it counts.
The PSI 90 Composite: What It Aggregates, and Why
AHRQ publishes more than two dozen individual PSIs, but a hospital’s public quality profile is usually driven by PSI 90 (Patient Safety and Adverse Events Composite) — a single risk-adjusted score built from a defined subset of the individual indicators, not from all of them. PSI 90 exists because a hospital’s raw count on any one component indicator is often too small, in a given reporting period, to be statistically reliable on its own; combining several related indicators into one composite produces a more stable signal for comparing hospitals.
AHRQ constructs the composite using a harm-weighted, volume-weighted approach: each component’s contribution to the composite reflects both how severe the outcome it measures typically is and how frequently that outcome occurs across the reporting population, so a low-volume but high-harm indicator isn’t washed out by a high-volume, lower-harm one. The composite is then risk-adjusted for patient case mix and reliability-adjusted so that hospitals with smaller case volumes aren’t penalized or rewarded on the basis of statistical noise. AHRQ periodically revises exactly which indicators are included in PSI 90 and how they’re weighted — treat the component list below as the current, commonly documented composition and confirm it against AHRQ’s own technical specifications before using it for scoring, benchmarking, or public reporting purposes, the same way you’d confirm a HAC list before citing it as current.
PSI 90’s Component Indicators
| PSI | Name | What it screens for |
|---|---|---|
| PSI 03 | Pressure Ulcer Rate | Stage III/IV or unstageable pressure injury developing during the stay |
| PSI 06 | Iatrogenic Pneumothorax Rate | Pneumothorax attributable to a procedure, not the admitting condition |
| PSI 08 | In-Hospital Fall with Hip Fracture Rate | Hip fracture coded as a complication of the inpatient stay |
| PSI 09 | Perioperative Hemorrhage or Hematoma Rate | Hemorrhage/hematoma requiring a specific treatment following surgery |
| PSI 10 | Postoperative Acute Kidney Injury Requiring Dialysis Rate | New dialysis-requiring AKI after elective surgery |
| PSI 11 | Postoperative Respiratory Failure Rate | Prolonged ventilation or reintubation after elective surgery |
| PSI 12 | Perioperative Pulmonary Embolism or DVT Rate | VTE developing after surgery, in surgical populations |
| PSI 13 | Postoperative Sepsis Rate | Sepsis coded as a complication of elective surgery |
| PSI 14 | Postoperative Wound Dehiscence Rate | Reclosure of a deep surgical wound in abdominopelvic surgery |
| PSI 15 | Abdominopelvic Accidental Puncture or Laceration Rate | Unintended puncture/laceration during an abdominopelvic procedure |
Two AHRQ indicators worth knowing even though they sit outside PSI 90: PSI 04 (Death Rate among Surgical Inpatients with Serious Treatable Complications), commonly called “failure to rescue,” and PSI 90’s predecessor-era components that AHRQ has retired or moved to standalone reporting over successive specification updates. If a component you expect to see isn’t in the table above, that’s a sign the specification has moved since this page was last checked against AHRQ’s technical documents — not necessarily an error in either source.
Numerator/Denominator/Exclusion Logic for the Most-Cited PSIs
The table above gives the general shape of each indicator. A few of the most frequently cited in hospital quality dashboards are worth walking through in more depth, because their exclusion logic is where most case-validation disputes actually happen.
PSI 03 — Pressure Ulcer Rate
Denominator: medical and surgical discharges age 18+ with a length of stay of several days or more (very short stays are excluded because a stage III/IV ulcer generally can’t develop and be documented within a day or two). Numerator: discharges with a secondary diagnosis of a stage III, stage IV, or unstageable pressure injury coded N or U for POA. Exclusions: a principal diagnosis of pressure ulcer, an admitting diagnosis that already documents the ulcer, and several other conditions AHRQ treats as confounding the measure (severe burns, certain exfoliative skin disorders). This is the PSI where POA miscoding does the most damage in either direction — a genuinely present-on-admission ulcer that gets coded N because the admission assessment wasn’t documented clearly inflates the rate for something clinical teams didn’t cause.
PSI 06 — Iatrogenic Pneumothorax Rate
Denominator: all medical and surgical discharges age 18+. Numerator: discharges with a secondary diagnosis of iatrogenic pneumothorax, coded N or U for POA. Exclusions: cases with a principal diagnosis of pneumothorax, trauma diagnoses, cardiac or thoracic surgery cases (where a pneumothorax is a recognized risk of the procedure itself rather than an unexpected complication), and pleural effusion procedures. The trauma and cardiothoracic-surgery exclusions are the ones most often misapplied when a case is being manually reviewed, because the exclusion depends on the procedure actually performed, not just the specialty of the admitting service.
PSI 09 — Perioperative Hemorrhage or Hematoma Rate
Denominator: surgical discharges age 18+ with an operating-room procedure. Numerator: discharges with a secondary diagnosis of postoperative hemorrhage or hematoma and a procedure code for treatment of that hemorrhage/hematoma (a diagnosis code alone isn’t enough — the indicator specifically requires evidence that the bleeding was significant enough to need its own intervention). Exclusions: a principal diagnosis of hemorrhage/hematoma, and cases where the treatment procedure occurred before or on the same day as the index procedure, which suggests the bleeding was addressed as part of the original operation rather than as a distinct postoperative event.
PSI 15 — Abdominopelvic Accidental Puncture or Laceration Rate
Denominator: discharges age 18+ with an abdominopelvic surgical, obstetric, or medical procedure. Numerator: discharges with a secondary diagnosis of accidental puncture or laceration during a procedure, again gated on POA (N/U). Exclusions: obstetric discharges are handled separately from the general surgical population, and a principal diagnosis already describing the injury removes the case. This indicator is intentionally broad across specialties (general surgery, urology, gynecology) rather than tied to one procedure type, which is why case-validation review for PSI 15 has to start from “what procedure was actually done” rather than assuming a single specialty owns it.
Validating a Flagged Case
A PSI flag is the start of a review, not the end of one. A defensible validation workflow, run by whoever on the patient-safety team owns PSI/HAC data integrity, generally covers:
- Confirm the POA value against the source documentation, not just the coded claim — the single most common source of a disputed flag is a POA value that doesn’t match what the chart actually supports, in either direction.
- Check the case against the indicator’s exclusion list before assuming the flag is a true positive — principal-diagnosis exclusions and procedure-timing exclusions (as in PSI 09 above) are the ones most often missed on a first pass.
- Pull the clinical record for a true clinical-validity review, separate from the coding review — a case can be coded correctly and still represent a legitimate clinical question about whether the event was actually preventable, which is a root-cause-analysis question, not a coding one.
- Route confirmed true positives into the same event-review process the rest of your safety program uses, rather than treating PSI validation as a standalone reporting exercise disconnected from clinical improvement work.
How PSI Differs From a HAC
“PSI” and “HAC” get used almost interchangeably in casual conversation, but they’re two different mechanisms that happen to share the same POA data:
- A PSI is a measurement definition. AHRQ defines the numerator/denominator/exclusion logic, and PSI rates are used for public reporting (Care Compare), AHRQ’s own quality reporting, and as an input to other programs. A PSI, by itself, doesn’t set a hospital’s payment.
- The HAC payment provision (DRA 2005 §5001(c)) is a claim-level payment rule, not a rate: it strips a specific hospital-acquired diagnosis of its ability to raise a claim to a higher-paying DRG when that diagnosis is coded N or U for POA on CMS’s designated HAC list. It runs directly off the POA field CASRAI covers in the linked POA guide, independent of any PSI calculation.
- The HAC Reduction Program (HACRP) is a third, separate mechanism: it combines the PSI 90 composite with NHSN healthcare-associated infection standardized infection ratios into a Total HAC Score, and applies a 1% total Medicare payment reduction to the worst-performing quartile of hospitals each year. This is the point where “PSI” and “HAC” genuinely do intersect — PSI 90 is literally one of the two inputs — but it’s still a distinct program from the DRA 2005 claim-level provision above, and conflating the two is a common, avoidable error on a quality dashboard.
In short: every HACRP score depends partly on PSI 90, but not every PSI-flagged case has anything to do with a payment reduction, and the DRA 2005 HAC payment provision doesn’t reference PSI logic at all — it runs on the HAC list and the POA field directly.
Frequently Asked Questions
Does a PSI-flagged case mean a preventable safety event definitely happened?
No. A PSI flag means the claim met the indicator’s numerator/denominator/exclusion logic based on coded data — it’s a screen built to catch likely cases, not a clinical determination. Confirming an actual safety event requires the clinical-validity review described above.
Is PSI 90 the same thing as “the HAC score”?
Not exactly. PSI 90 is one of two inputs into the Hospital-Acquired Condition Reduction Program’s Total HAC Score (the other being NHSN infection SIRs) — so PSI 90 heavily influences HACRP performance, but HACRP is a broader program and the DRA 2005 HAC payment provision (the CC/MCC DRG rule) is a separate mechanism entirely that doesn’t use PSI 90 at all.
Why does one PSI sometimes dominate a hospital’s PSI 90 score?
Because the composite is harm- and volume-weighted rather than a simple average, an indicator with both meaningful case volume and a high harm weight can move the composite more than several lower-weighted indicators combined — which is why patient-safety teams generally track component-level rates, not just the composite number, to know where an improvement effort will actually move the score.
Can a hospital appeal or correct a PSI-flagged case?
PSIs are calculated from submitted claims data, so the practical lever is coding and documentation accuracy before and at the time of claim submission — correcting a POA value or exclusion-relevant code after the fact, through the normal claims-correction process, is what changes whether a case counts, rather than an appeal against the PSI logic itself.
Do all AHRQ PSIs roll into PSI 90?
No. AHRQ maintains more PSIs than the composite uses — PSI 04 (failure to rescue) is a well-known example of an indicator reported and used independently rather than as a PSI 90 component. Always check AHRQ’s current technical specifications for composite membership rather than assuming every PSI number is included.
Related CASRAI Resources
For the coding mechanics behind the POA field this guide builds on, see Present on Admission (POA) Indicator. For the separate NHSN surveillance pipeline that runs on its own definitions rather than claims/POA, see CLABSI and CAUTI. For the broader payment context PSI 90 feeds into alongside HACRP, see Hospital Readmissions Reduction Program and Total Performance Score: How Hospital VBP Turns Four Domains Into a Payment Adjustment. For event-review process once a flagged case is confirmed, see Sentinel Event: What It Means, and What Happens Next. For the wider cluster, visit the Patient Safety & Infection Prevention pillar.
