Days of Therapy per 1,000 patient-days — the standard NHSN Antimicrobial Use metric — tracked per ward, risk-adjusted via SAAR, and benchmarked against national percentiles
Every antimicrobial consumption metric ultimately traces back to two raw data streams: pharmacy dispensing records and the electronic medication administration record (eMAR). The CDC's National Healthcare Safety Network (NHSN) Antimicrobial Use (AU) Option ingests these feeds — typically via an automated electronic bridge from the hospital's pharmacy information system — to build a continuously updated antimicrobial consumption ledger for every patient care location in the facility.
The AU module requires each reporting facility to submit patient-location-level antimicrobial administration data in a standardized format defined by the NHSN AU protocol:
• Source data — most hospitals bridge data automatically from their pharmacy dispensing system (billing-based) or, increasingly preferred by CDC, from eMAR-verified administration data, which more accurately reflects doses actually given rather than doses dispensed to the unit • Patient care location mapping — every administration event is tagged to a specific NHSN-defined location type (e.g., Medical Critical Care, Surgical Critical Care, Medical Ward, Oncology, Step-Down) using a standardized CDC location crosswalk, since national benchmarking is meaningless without consistent unit-type categorization across hospitals • Antimicrobial agent mapping — each dispensed/administered drug is mapped to an NHSN-standard antimicrobial category (e.g., "Broad Spectrum Agents predominantly used for hospital-onset/multidrug-resistant infections," "Agents predominantly for community-acquired infections") using CDC's published crosswalk file, updated periodically as new agents enter the market • Denominator data — patient-days and, for some analyses, admission counts are pulled from the hospital's ADT (admission-discharge-transfer) feed for the same location and time period
Data is submitted monthly, and NHSN performs automated validation checks (implausible rate flags, missing-location detection, month-over-month variance checks) before the data is accepted into the national analytic dataset used for benchmarking.
Since fiscal year 2024, CMS requires all IPPS-subject acute-care hospitals to report AU Option data through NHSN as part of the Hospital Inpatient Quality Reporting (IQR) Program — turning what was originally a voluntary CDC surveillance tool into a mandatory, payment-linked national reporting requirement.
Because DOT, SAAR, and percentile benchmarking are all downstream calculations, data quality problems at the capture stage propagate through the entire pipeline. Common quality issues stewardship pharmacists and infection preventionists screen for include miscoded patient locations (inflating or deflating a specific unit's denominator), missing days in the ADT feed, incorrect antimicrobial category crosswalk mapping for newly added formulary agents, and dispensing-based (rather than administration-based) data that overstates true consumption when doses are returned unused. Most mature stewardship programs run a monthly reconciliation between the NHSN-submitted data and their internal pharmacy analytics dashboard before the numbers are used for any external or internal benchmarking decision.
Days of Therapy (DOT) is the CDC/NHSN-preferred metric for measuring antimicrobial consumption because, unlike defined daily dose (DDD), it is dose- and weight-independent — a pediatric patient on a low weight-based dose and an adult on a full dose each contribute exactly one DOT per agent per day, making DOT far more accurate for comparing antimicrobial use across the diverse patient populations found in a single hospital or across hospitals nationally.
DOT/1000 patient-days = (Sum of days of therapy for all antimicrobial agents in the numerator / Number of patient-days in the denominator) × 1000
Worked example for a 20-bed medical ward over a 30-day month:
• Patient-days for the month: 20 beds × ~85% average occupancy × 30 days ≈ 510 patient-days • DOT tally: if patients on that ward received, in total, 400 days of therapy across all antimicrobial agents (a patient on piperacillin-tazobactam AND metronidazole for 3 days contributes 3+3 = 6 DOT) • DOT/1000 patient-days = (400 / 510) × 1000 ≈ 784
This rate can be calculated at multiple levels of granularity simultaneously: hospital-wide, by patient care location, by individual antimicrobial agent, or by NHSN antimicrobial category (e.g., "all anti-MRSA agents," "all broad-spectrum agents for hospital-onset infection") — each view answers a different stewardship question.
Three metrics compete for measuring antimicrobial consumption, each with tradeoffs:
• DOT (Days of Therapy) — dose-independent; a full-dose adult regimen and a renally-adjusted lower dose both count as 1 DOT/day; NHSN's preferred metric because it is not distorted by dose-rounding, renal adjustment, or pediatric weight-based dosing, all of which are common in real hospital populations • DDD (Defined Daily Dose) — WHO's international standard, based on the assumed average maintenance dose for an adult; useful for cross-national comparison but can significantly misrepresent actual exposure in populations with frequent renal dosing or pediatric/neonatal care, since DDD assumes a fixed adult dose regardless of what was actually given • Drug expenditure / cost — easy to obtain from finance systems but conflates consumption trends with drug pricing changes, formulary contract shifts, and generic-to-brand substitution, making it a poor proxy for actual antimicrobial exposure or resistance-pressure trends
Because DOT is administration-pattern-based rather than dose-based, it remains the metric of choice for both internal trending and the SAAR risk-adjustment model described in the next stage.
A raw DOT/1000 patient-days number is nearly meaningless in isolation — an oncology ward and a general medical ward have fundamentally different expected antimicrobial needs. The SAAR solves this by comparing a unit's observed DOT to a predicted DOT generated from a national regression model, producing a single risk-adjusted ratio directly comparable across facilities and unit types.
CDC statisticians built the SAAR predictive model using negative binomial regression on a large national baseline cohort of NHSN-reporting hospitals (initially 2015 data, periodically re-baselined). The model predicts expected DOT for a given patient care location based on statistically significant facility- and location-level covariates:
• Patient care location type (e.g., Medical ICU vs. Medical Ward vs. Oncology) — the single strongest predictor, since baseline antimicrobial need varies enormously by acuity and patient population • Facility bed size and teaching status — larger academic medical centers systematically show different baseline utilization patterns than small community hospitals • Community vs. hospital-onset infection burden proxies • For some categories, additional case-mix adjusters specific to the antimicrobial category being modeled
Six standard SAAR categories are calculated: All Antibacterial Agents, Broad Spectrum Agents predominantly for Hospital-onset/MDRO infections, Broad Spectrum Agents predominantly for Community-Acquired infections, Agents predominantly for Surgical Site Infection prophylaxis in certain surgeries, Agents predominantly used for Multi-Drug Resistant Organism (MDRO) infections, and Narrow Spectrum Beta-Lactam Agents — allowing a program to see not just whether overall use is elevated, but which category is driving it.
SAAR = Observed DOT / Predicted DOT, calculated per unit, per antimicrobial category, per reporting period:
• SAAR = 1.0 — observed antimicrobial use matches exactly what the national model predicts for a unit of this type and facility profile • SAAR > 1.0 — the unit is using more antimicrobial DOT than predicted; values persistently above 1.0, and especially above the 90th percentile of the national SAAR distribution, are candidates for stewardship review • SAAR < 1.0 — use is below the national predicted baseline; while sometimes reflecting an efficient, well-targeted antimicrobial program, unusually low SAAR (particularly <0.5) can also signal under-treatment of genuine infections and is itself sometimes flagged for clinical quality review, not just celebrated
Unlike a raw DOT count, SAAR is designed to be compared meaningfully across hospitals of different sizes, teaching status, and case mix — which is precisely what makes it suitable for the national percentile benchmarking performed in the next stage.
Once a facility's SAAR and DOT values are calculated, NHSN's AU Option Analysis Reports place them on a percentile curve relative to the full national reporting cohort — typically several thousand acute-care hospitals — for the matching patient care location and antimicrobial category, giving stewardship leadership an external reference point beyond the hospital's own historical trend.
NHSN's AU Option Analysis Reports (accessible directly within the NHSN application by facility users) generate line-list and run-chart views showing a facility's own SAAR/DOT trend overlaid against the national percentile distribution for hospitals of comparable type reporting the same patient care location and antimicrobial category:
• The 50th percentile (median) line shows where a typical reporting hospital sits • The 25th–75th percentile band (interquartile range) is often shaded to give a visual "normal range" • The 90th percentile line marks the CDC-recommended threshold above which a facility's SAAR is considered a priority target for stewardship intervention
Stewardship committees typically review this benchmarking quarterly, presenting trended SAAR and percentile position by unit and by antimicrobial category to hospital leadership and the pharmacy & therapeutics committee, since a rising percentile rank — even with a stable raw DOT number — can indicate the facility is falling behind national improvement trends even if its own absolute numbers look unchanged.
Percentile benchmarking is a powerful external reference but has recognized limitations that stewardship teams are trained to keep in mind: the national comparison cohort itself is shifting over time as antimicrobial practice patterns evolve nationally (so "improvement" relative to peers does not always mean absolute improvement), small patient care locations can show volatile month-to-month percentile swings due to low denominator patient-days, and case-mix adjustment, while statistically robust, cannot capture every clinically relevant difference between two units nominally of the "same type" (for example, two Medical ICUs with very different admission thresholds or transplant/immunocompromised patient proportions). For these reasons, NHSN and CDC guidance consistently frame percentile position as a screening and prioritization tool — flagging where to look — rather than a definitive verdict on care quality.
With risk-adjusted SAAR values and national percentile position in hand, stewardship teams apply a structured triage process to identify which units and antimicrobial categories deserve focused intervention resources — since no program has the staffing to deep-dive every unit every month.
A typical outlier identification workflow proceeds in layers:
1. Screen — pull all patient care location × antimicrobial category combinations with SAAR > 1.0 for the most recent reporting quarter 2. Prioritize by percentile — among SAAR > 1.0 units, rank by national percentile position; units at or above the 90th percentile are typically escalated first 3. Confirm persistence — a single elevated month is often noise (especially on small units with low patient-day denominators); CDC and most internal protocols require sustained elevation across at least 2–3 consecutive reporting periods before committing significant stewardship resources 4. Drill into agent-level detail — once a unit/category combination is confirmed as a priority, the team pulls agent-specific DOT data (which specific broad-spectrum agents are driving the elevated category SAAR — is it one drug or a broad pattern?) and reviews individual patient charts, often through the PAF review process itself 5. Assign an intervention — depending on root cause, the response can range from targeted prescriber education, a new or revised clinical pathway/order set, formulary restriction, or an intensified PAF review cadence specifically for that unit
ICU and oncology/hematology units are disproportionately represented among national SAAR outliers — not necessarily because care is worse there, but because empiric broad-spectrum coverage for undifferentiated sepsis and febrile neutropenia is guideline-recommended as a starting point, making these units structurally prone to elevated broad-spectrum SAAR even in well-run stewardship programs. Outlier flagging must distinguish appropriate high empiric use from unnecessary prolonged use.
Once a unit is formally flagged, most stewardship programs establish a defined accountability loop: the finding is presented to the unit's medical director or service chief, a specific improvement target and timeline is set (e.g., "reduce broad-spectrum SAAR from 1.35 to 1.15 within two quarters"), and progress is re-measured at the next reporting cycle using the same SAAR/percentile methodology — closing the loop between measurement and the stewardship intervention impact tracked in the final stage.
The ultimate purpose of the entire measurement pipeline — dispensing capture, DOT calculation, SAAR risk-adjustment, percentile benchmarking, and outlier flagging — is to drive a measurable, sustained decline in unnecessary antimicrobial exposure. The before/after DOT trend line on a flagged unit following a targeted intervention is the clearest evidence a program can present that its work translates into real clinical impact.
A rigorous intervention-impact analysis uses an interrupted time-series approach: the unit's DOT/1000 patient-days (and/or category SAAR) is plotted continuously across a baseline period, an intervention start date is marked, and the post-intervention trend line is statistically compared to the counterfactual trajectory the pre-intervention trend would have predicted. This is more rigorous than a simple before/after mean comparison because it accounts for pre-existing trends and seasonal variation (antimicrobial use characteristically rises during winter respiratory viral season, which can otherwise be mistaken for — or mask — an intervention effect).
Common interventions tracked this way include: launching or intensifying PAF review specifically on the flagged unit, introducing a new clinical pathway or order set (e.g., a sepsis order set with a built-in 48–72 hour reassessment prompt), restricting a specific driver agent to ID-approval-only, or targeted prescriber education paired with individual peer-comparison feedback reports (academic detailing), a technique with strong evidence of effect in the prescribing-behavior literature more broadly.
Well-documented stewardship intervention case studies — published in journals such as Infection Control & Hospital Epidemiology, Clinical Infectious Diseases, and Open Forum Infectious Diseases — commonly report DOT reductions of 15–35% on specifically targeted units within 6–12 months, with the SAAR trend normalizing toward or below 1.0 and the unit's national percentile rank dropping out of the flagged range. Critically, the literature is consistent that these gains are not self-sustaining without continued measurement and intervention: units that discontinue active PAF review after an initial successful intervention frequently show partial or full drift back toward their pre-intervention baseline within 12–18 months, which is why the DOT/SAAR/percentile measurement cycle described across all six stages of this pipeline is run continuously, quarter after quarter, rather than as a one-time audit.
A 2017 CDC Vital Signs report analyzing hospitals with mature antimicrobial stewardship programs estimated that broad national adoption of this full measurement-to-intervention cycle could reduce inappropriate inpatient antibiotic use by roughly 30% — directly slowing selection pressure for resistant organisms at the population level, not just improving one hospital's internal metrics.