NDNQI: What It Collects and How Benchmarking Works

NDNQI — the National Database of Nursing Quality Indicators — is a subscription benchmarking database of unit-level nursing measures. It was developed by the American Nurses Association, through its National Center for Nursing Quality, and launched in 1998. It is no longer an ANA programme: NDNQI is now owned and operated by Press Ganey, which markets it as a commercial platform. That ownership change matters more than it sounds, and this page is careful about it throughout — NDNQI is a vendor product with a proprietary methodology, not a public standard, and its numbers do not mean the same thing as the public, mandatory rates a hospital reports to CMS or CDC.

The single thing most often missed about NDNQI: an NDNQI rate is a comparison-group-relative number. Your fall rate is not reported to you as a bare rate to be judged against the world; it is reported against the distribution of rates from units of the same type in other participating hospitals. Unit type is doing the work that a patient-level statistical risk model does in a CMS programme. Understanding that one design decision is the difference between reading your NDNQI report correctly and quoting a number that cannot support the claim you are making with it.

What NDNQI actually collects

NDNQI collects at the nursing unit, not the hospital. This is its defining structural choice and the reason it exists alongside, rather than inside, the federal measurement programmes. Press Ganey states the platform now carries “750+ measures, including 250+ nursing-sensitive indicators across all inpatient and ambulatory care settings”, drawn from more than 55,000 nursing care units and 16,000 ambulatory care units. The authoritative current indicator list and its specifications are Press Ganey’s to define and are issued to participating organisations — see what is not public below. What follows is the structure of the collection, and the indicator families that have long anchored the database, rather than a claim to be a current specification.

The nursing-sensitive concept

A nursing-sensitive indicator is a measure whose value is materially moved by nursing practice, nurse staffing and the nursing work environment — as opposed to one driven mainly by surgical technique, physician decision-making or case mix. Press Ganey’s own definition is that these are “measures that capture aspects of patient care most directly influenced by nursing practice, staffing, and the work environment”. The concept is the selection rule for what belongs in the database at all: a measure earns a place because nursing can plausibly change it, not because it is important.

Practically, NDNQI’s content divides three ways — the familiar structure / process / outcome split:

1. Outcome indicators — what happened to patients

Patient falls and falls with injury, reported per 1,000 patient days, with injury level classified separately from the fall count. The injury-stratified rate, not the total fall rate, is usually the one that carries weight — a unit can drive total falls down by immobilising patients, which is not an improvement.
Hospital-acquired pressure injuries, captured through a point-prevalence survey rather than continuous incident reporting, and staged.
Device-associated infections — central line-associated bloodstream infection, catheter-associated urinary tract infection, and ventilator events. These are specified to align with CDC surveillance definitions, so the same case adjudication can serve both submissions; see our pages on the NHSN CLABSI surveillance definition and LCBI criteria and on CAUTI for the definitions themselves.
Physical restraint prevalence, also survey-captured.
Paediatric-specific outcomes, historically including peripheral IV infiltration and the pain assessment / intervention / reassessment cycle.
Assault on nursing personnel and workplace-injury measures — the staff side of the outcome set.

2. Structure indicators — who was on the unit

Nursing hours per patient day (NHPPD), broken out by RN, LPN/LVN and unlicensed assistive personnel.
Skill mix — the share of total nursing care hours delivered by each of those categories.
RN education (proportion holding a baccalaureate or higher) and specialty certification rates.
Nurse turnover.

The structure indicators are what make the outcome indicators interpretable. A fall rate on its own is a number; a fall rate next to that unit’s NHPPD, skill mix and turnover for the same quarter is an argument. This pairing is most of the analytic value of the database and it is why NDNQI submissions are quarterly and unit-keyed rather than annual and hospital-keyed.

3. The RN Survey — the work environment, measured directly

The NDNQI RN Survey is a separate instrument administered to registered nurses at the unit level, and it is the component most often left out of summaries of NDNQI. It is not a generic staff-engagement questionnaire. It is built from named, published scales, the anchor being the Practice Environment Scale of the Nursing Work Index (PES-NWI) — a validated instrument measuring nurse participation in hospital affairs, the foundations of quality of care, nurse manager ability and support, staffing and resource adequacy, and nurse–physician relations. Press Ganey describes the RN Survey as comprising the Practice Environment Scale alongside scales for missed nursing care, RN-assessed quality of care, professional development access and opportunity, intent to stay, and RN job satisfaction.

Two things follow. First, because the survey is unit-keyed like everything else in NDNQI, a hospital can put a unit’s practice-environment score beside that same unit’s fall rate, turnover and skill mix for the same period — which is the whole point of holding all of it in one database. Second, the RN Survey is what satisfies the nurse satisfaction limb of a Magnet application, distinct from the clinical indicators; see the Magnet section.

The unit-type comparison logic — and why it is the point

This is the part that changes how you read the report.

Unit type is the risk adjustment

NDNQI does not, in the main, adjust a unit’s rate using a patient-level statistical model of comorbidity and acuity. It stratifies: it places your unit into a cohort of like unit types and shows you where you sit in that cohort’s distribution. Adult critical care is compared with adult critical care. Adult step-down with adult step-down. Adult medical, adult surgical and combined medical–surgical are separate cohorts, not one. Rehabilitation, psychiatric/behavioural health, paediatric acute, paediatric critical care, neonatal intensive care, and obstetric/postpartum units each sit in their own.

The reason is that the confounding is categorical, not continuous. A medical ICU’s fall rate is low substantially because a large share of its patients are sedated, ventilated and not ambulating — not because its fall-prevention programme is excellent. An inpatient rehabilitation unit’s fall rate is high substantially because deliberately mobilising unsteady patients is the clinical purpose of the admission — not because it is unsafe. No plausible patient-level covariate adjustment reconciles those two units into one comparison; the honest move is to refuse to compare them. Categorical stratification by unit type is that refusal, formalised.

So the correct reading of an NDNQI result is always conditional: “among adult step-down units in the database, ours sits here.” Strip the conditional off and the number stops meaning anything.

The other comparison groups

Unit type is the primary stratification, but it is not the only one. Press Ganey states that NDNQI lets a subscriber view “more than 10 comparison groups, including teaching status, staffed bed size, census division, state, metropolitan area, ANCC Magnet® status, and more.” These are filters layered on top of the unit-type cohort, which is what makes the tool useful for an improvement argument and hazardous for a rhetorical one: a unit that looks unremarkable against all like units nationally may look poor against like units in academic medical centres of similar bed size, or vice versa. Both statements can be true simultaneously, and a report that does not name its comparison group has not said anything. When an NDNQI figure appears in a board paper, a QAPI report or a performance-improvement project write-up, the comparison group belongs in the sentence, not in a footnote.

What this means for your own data, practically

Unit-type assignment is yours to make, and it is load-bearing. Participating hospitals classify their own units against the database’s unit-type definitions. Mis-assign a unit — a step-down classified as critical care, a mixed unit forced into a pure medical or pure surgical category — and every comparison you draw for it is against the wrong cohort. This is the most common way an NDNQI report becomes quietly invalid, and nothing in the report will flag it.
Reclassifying a unit breaks its trend. If a unit changes type — or you correct a long-standing mis-assignment — the historical series is no longer a like-for-like time series, because the cohort it is being measured against has changed underneath it. Note the reclassification date on any longitudinal chart.
Denominators differ between indicators, so the numbers are not interchangeable. Falls are a rate per 1,000 patient days. Pressure injuries come from a point-prevalence survey and are a proportion of patients surveyed on the survey day. Staffing is hours per patient day. Adding, averaging or trading these off against one another produces a figure that corresponds to nothing.
The comparison group is a moving reference, not a fixed target. It is a distribution of other participating hospitals’ units, recomputed as the database changes. A unit’s position can move because the cohort moved. That is a feature for benchmarking and a trap for goal-setting: a target expressed purely as a position in a distribution can be missed in a period where the unit’s own performance improved.

Why an NDNQI rate is not a CMS rate

Hospitals routinely hold two numbers for what sounds like the same event, and staff reasonably assume one is a check on the other. They are not. NDNQI and the federal quality programmes are different instruments built for different purposes, and the contrast is worth stating explicitly:

	NDNQI (Press Ganey)	CMS quality programmes
Participation	Voluntary, by paid subscription	Mandatory for participating hospitals, tied to payment
Ownership	Proprietary; commercial vendor	Public programme; federal agency
Unit of measurement	The nursing unit	The hospital
Adjustment method	Categorical stratification by unit type and comparison group	Patient-level statistical risk-adjustment models
Measure selection	Nursing-sensitive by design	Whole-episode and whole-hospital outcomes
Benchmark visibility	Visible only to subscribers	Published; hospital-identifiable
Consequence	Internal improvement; supports Magnet	Payment adjustment and public reporting

Concretely: the Total Performance Score under Hospital VBP is a hospital-level composite that converts four weighted domains into a payment adjustment; the Hospital Readmissions Reduction Program applies a risk-adjusted, hospital-level excess-readmission ratio against a penalty; SEP-1 is an all-or-nothing bundle scored case by case against a published specification. Every one of those is mandatory, hospital-level, publicly reported and specified in a document anyone can read. NDNQI is none of those four things. It is a voluntary, proprietary, unit-level, nursing-sensitive instrument whose comparison values are visible only to the organisations paying for access.

The operational consequence: do not present an NDNQI position as though it were a public rate, and do not reconcile the two. An NDNQI CLABSI figure and an NHSN-reported CLABSI standardised infection ratio can diverge legitimately, because one is a unit-cohort position and the other is a hospital-level ratio against a national referent — a difference in method, not an error in either. Where a governance paper needs a defensible external number, the public programme is the citable source. Where it needs to know which unit to work on and why, NDNQI is the instrument that can answer, and the public programme cannot. Use each for the question it was built for.

How NDNQI data supports a Magnet application

The ANCC Magnet Recognition Program is the reason most hospitals subscribe to NDNQI, and the relationship is widely misdescribed. What ANCC actually requires, in its own published eligibility criteria, is this:

“Applicants for Magnet designation must collect nurse-sensitive quality indicators at the unit level and benchmark that data against a database at the highest/broadest level possible (i.e., national, state, specialty organization, regional, or system).”

“Organizations must contribute their own data (patient and nurse satisfaction, clinical nurse sensitive indicators) to a national database that compares the organization’s data against cohort groups at the national level.”

Read those two sentences carefully, because three things follow that are routinely got wrong.

1. ANCC does not require NDNQI

The eligibility criteria name no vendor. They describe a capability — unit-level collection, contribution to a national database, comparison against national cohort groups — and NDNQI is one product that satisfies it. Other national and specialty databases can also satisfy it. NDNQI is the dominant choice, and Press Ganey markets that dominance directly, stating that “100% of the Magnet® with Distinction organizations that partner with us use NDNQI” and offering “Magnet®-approved graphs, crosswalks, and reports” built to the application’s expected format. That is a real and substantial convenience. It is not a mandate, and a hospital should not be told it is one.

2. The requirement is structural, and NDNQI’s design maps onto it exactly

ANCC asks for unit-level data compared against national cohort groups. That is a description of stratified benchmarking — the same design described in the unit-type section above. This is why the two fit together so neatly: a nursing-unit-keyed database with national like-unit cohorts is precisely the artefact the application asks an organisation to produce. It also means the pathology is shared. If your unit-type assignments are wrong, your Magnet evidence is built on the wrong cohorts, and the defect is invisible in the output.

3. Both limbs are required — clinical and nurse satisfaction

The criteria name patient and nurse satisfaction alongside clinical nurse sensitive indicators. The RN Survey is what carries the nurse-satisfaction limb, which is why a hospital cannot treat the survey as optional colour alongside the clinical indicators — it is a separate evidentiary requirement in its own right.

What we are not going to tell you: the thresholds

Magnet applications turn on how many quarters of benchmarked data are required and how many units must outperform their benchmark for how much of that period. Those specifics are not on ANCC’s public eligibility page. They live in the Magnet Application Manual, a separately purchased ANCC publication — the 2023 edition is the one currently sold through ANA’s publishing catalogue — and they have changed between editions. A commonly-cited pattern is eight quarters of unit-level nationally benchmarked data with outperformance required for a majority of units for a majority of the period, but you should not run a two-year data programme on a figure quoted second-hand from a page like this one. Buy the manual edition your organisation is applying under and read the requirement in it. That is not evasion; the appraisal is scored against the manual, not against consensus.

What is proprietary, and where your own numbers come from

This page deliberately states no NDNQI benchmark values, means, medians or percentile thresholds. That is not caution for its own sake — those values are proprietary to Press Ganey, differ by comparison group, and are recomputed as the participant pool changes, so a number quoted from a third-party page is wrong in at least one of three ways before you use it. Specifically:

Benchmark distributions are behind a customer login. Comparison-group means, medians and percentile positions are delivered through Press Ganey’s subscriber reporting portal to participating organisations. There is no public lookup, no free tier and no open API. You cannot look up another named hospital’s NDNQI results, and neither can anyone else look up yours.
Indicator specifications are issued to subscribers. The operative definitions — what counts as a fall with injury, how the pressure-injury prevalence survey is to be conducted, how nursing hours are attributed — are set out in the guidelines Press Ganey supplies to participating organisations, not in a public standard. This is a material difference from a CMS measure, whose specification manual is downloadable by anyone.
Your own data comes from your own submission. Your organisation’s NDNQI figures are computed from what your hospital submits each quarter. The route in is your nursing quality or Magnet programme office, which holds the site credentials, the unit-type assignments and the submission calendar. Press Ganey has also announced direct integration of nursing quality data from Epic, which changes how the data is extracted but not who can see the benchmarks.

If you need externally citable nursing-quality evidence for an audience that cannot be given portal access — a regulator, a public board paper, a journal reviewer — NDNQI positions are the wrong artefact, and the public federal measures are the right one. If you are writing up an improvement project, SQUIRE 2.0 is the reporting guideline that will ask you to describe your measures and their context anyway, at which point “benchmarked against like unit types within a proprietary national database” is a perfectly respectable and honest description.

Where NDNQI sits alongside the rest of the safety programme

NDNQI is a measurement instrument, not a safety programme. It tells a unit where it stands; it does not tell anyone what happened or what to do. The surrounding machinery is separate and is covered elsewhere on this site: sentinel event review handles the individual catastrophic case that no rate will surface; the Just Culture algorithm governs how an organisation classifies behaviour after an adverse event, which is what stops a rising indicator from turning into a search for someone to blame; the PSO work-product privilege determines what analysis can be protected; the National Patient Safety Goals set accreditation-driven requirements including falls and pressure-injury expectations; and early warning score implementation and rapid response team activation criteria are the deterioration-detection layer whose effects show up in outcome indicators quarters later. The people who run this work day to day — infection preventionists for the device-associated indicators, and quality staff often holding the CPHQ credential — are usually the ones reading the NDNQI report alongside everything else. For the wider map, see the patient safety pillar.

Frequently asked questions

What does NDNQI stand for?

National Database of Nursing Quality Indicators. It is now branded by its owner as the Press Ganey National Database of Nursing Quality Indicators.

Who owns NDNQI?

Press Ganey owns and operates it. It was developed by the American Nurses Association through its National Center for Nursing Quality and launched in 1998; ANA subsequently divested it. Content describing NDNQI as an ANA programme in the present tense is out of date.

Is NDNQI mandatory?

No. Participation is voluntary and by paid subscription. This is the sharpest single contrast with CMS quality programmes, which are mandatory for participating hospitals and carry payment consequences.

Does the Magnet programme require NDNQI specifically?

No. ANCC requires unit-level nurse-sensitive indicator data benchmarked against a national database and names no vendor. NDNQI is the most widely used way to satisfy that requirement and Press Ganey supplies Magnet-formatted reporting, but other qualifying national and specialty databases exist.

Why can’t I compare my NDNQI fall rate to a published national fall rate?

Because an NDNQI figure is a position within a cohort of like unit types inside a proprietary database, not a free-standing rate against a public referent. A published national figure has a different denominator, a different population and a different adjustment method. Comparing them produces a number that describes neither.

Is NDNQI the same as NHSN?

No. NHSN is CDC’s public surveillance system and the source of the standard case definitions for device-associated infections. NDNQI is a commercial benchmarking database that carries infection indicators aligned to those definitions alongside falls, pressure injuries, staffing and the RN Survey — measures NHSN does not collect. They answer different questions and are not reconciliations of one another.

What is the NDNQI RN Survey?

A unit-level survey of registered nurses built from published scales — the Practice Environment Scale of the Nursing Work Index at its centre, plus scales covering missed nursing care, RN-assessed quality of care, professional development access and opportunity, intent to stay, and job satisfaction. It supplies the nurse-satisfaction evidence a Magnet application requires.

Where do NDNQI benchmark values come from, and can I see them?

They are computed from the submissions of participating hospitals and delivered through Press Ganey’s subscriber reporting portal. Only participating organisations can see them, and only for their own units against comparison-group distributions. There is no public lookup.

What are nursing-sensitive indicators?

Measures whose values are materially influenced by nursing practice, nurse staffing and the nursing work environment — falls and falls with injury, hospital-acquired pressure injuries, device-associated infections, restraint use, nursing hours per patient day, skill mix, RN education and certification, and nurse turnover, among others. The category is a selection rule: a measure qualifies because nursing can move it.