Antibiogram: How to Build and Read a Cumulative Antibiogram

A cumulative antibiogram is a periodic summary of how the bacteria isolated at one facility responded to antimicrobial testing — usually a table of organisms down the side, antimicrobial agents across the top, and the percentage of isolates susceptible in each cell. Its only real job is to answer one question at the bedside: when a clinician must start therapy before culture results return, which agent is most likely to cover this organism here?

That makes the antibiogram a decision tool, not a report. The governing standard is CLSI M39, Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, currently in its Fifth Edition, published 24 January 2022 (it replaced M39-A4 from 2014). This guide covers what an antibiogram must contain, the decisions you make while building one, how to read it correctly, and — importantly — the specific ways it is routinely misused.

What CLSI M39 actually governs (and what it does not)

M39’s own stated scope is the storage, analysis, and presentation of cumulative antimicrobial susceptibility test data. It gives recommendations for preparing routine and enhanced antibiograms to guide selection of empirical antimicrobial therapy, and it is written for clinical microbiologists, pharmacists, physicians, veterinarians, epidemiologists, infection prevention practitioners — and for the LIS and EHR vendors who build the software that generates these reports.

What M39 explicitly excludes is just as important, because it is where most confusion starts. Per its own scope statement, the guideline does not cover:

• procedures for selecting isolates for antimicrobial susceptibility testing (AST);
• performing AST;
• interpreting individual AST results;
• confirming the accuracy of AST results.

Those belong to the M02/M07/M100 family of CLSI documents. M39 assumes the susceptibility results reaching it are already final, accurate, and in a usable format. If your isolate-level testing or breakpoint application is wrong, M39 will not catch it — it will aggregate the error and give it the authority of a published table.

What changed in the Fifth Edition

CLSI’s published overview of changes for M39-Ed5 lists, among others:

• added definitions for “cumulative antimicrobial susceptibility test data report” and for “antibiogram” — the two had been used loosely and interchangeably;
• added considerations for extracting data from different sources (automated AST instrument, LIS, electronic health record) when preparing an antibiogram;
• combining results from rapid diagnostics and antimicrobial resistance marker testing with the antibiogram for empirical therapy selection;
• developing antibiograms for yeast and antifungal agents;
• developing antibiograms for multiple facilities, long-term care facilities, and veterinary practices;
• describing ways an antimicrobial stewardship programme may use antibiogram data;
• considerations for preparing cumulative data for peer-reviewed publication;
• statistical techniques including percentiles, interquartile ranges, MIC₅₀ and MIC₉₀;
• a general comment explaining use of the “^” symbol with intermediate breakpoints for agents known to concentrate in urine;
• deletion of the recommendation to list percent intermediate alongside percent susceptible for penicillin with viridans group streptococci.

The through-line in that list is that M39-Ed5 assumes a modern, multi-source, stewardship-driven data environment rather than a single instrument printout — which is exactly the environment most hospitals now build antibiograms in.

The decisions you make while building one

Building an antibiogram is not a report-generation task. It is a sequence of judgement calls, each of which changes the number in the cell. These are the ones that matter.

1. Which isolates count

The single most consequential convention in antibiogram construction is that the report is built from patient isolates, not specimens. A patient with a prolonged bacteraemia who is cultured daily for two weeks would otherwise contribute a dozen identical, resistant results and drag the whole organism row toward resistance. The standard approach is therefore to include only the first isolate of a given species per patient per analysis period, and to apply that rule consistently across every organism.

Deciding what counts as “the same” isolate is where teams disagree: the same species from a different body site, the same species with a materially different susceptibility profile, and repeat isolates spanning separate admissions are all handled differently by different laboratories. Write the rule down and apply it uniformly — an antibiogram whose de-duplication rule varies by organism cannot be compared year over year, which destroys its second-most-valuable use.

2. Surveillance isolates and screening cultures

Isolates recovered from surveillance screening — MRSA nares swabs, VRE rectal screens, carbapenem-resistant organism screens — represent colonisation, not infection, and they are heavily enriched for resistance by design. Mixing them into a clinical antibiogram systematically overstates resistance and pushes empirical therapy toward broader agents than the infecting-organism data supports. Exclude them, and say on the report that you have.

3. Which body sites, and whether to stratify

A single facility-wide table hides the variation clinicians most need. Resistance in urinary isolates differs from resistance in blood isolates; ICU differs from general wards; paediatric differs from adult; outpatient differs from inpatient. Stratification is what turns an antibiogram from a compliance artefact into a prescribing tool.

The tension is statistical: every stratification divides the isolate count, and below a certain number of isolates a percentage is noise. This is the core trade-off in antibiogram design — granularity versus stability — and it has no universal answer. A large academic centre can stratify by unit; a 60-bed community hospital often cannot stratify at all without falling under the reporting threshold for most organisms. M39 addresses combining data across facilities precisely because small facilities face this problem.

4. Which agents to display

An antibiogram displays the agents on your formulary that are plausible empirical choices, not every agent the instrument tested. Two considerations pull against simply printing everything:

• Selective and cascade reporting. Laboratories routinely suppress results for broad-spectrum agents unless narrower agents test resistant. If suppressed results are excluded from the antibiogram, the denominator for that agent is not the same as for the others, and the resulting percentage is not comparable — it reflects a pre-selected, more resistant subset.
• Stewardship signalling. Displaying an agent on the antibiogram implicitly endorses it as an empirical option. Coordinate the agent list with your antimicrobial stewardship programme rather than treating it as a purely laboratory decision.

5. How to handle intermediate results

The conventional presentation is percent susceptible only — the percentage of tested isolates that were susceptible, with intermediate and resistant isolates both counted in the denominator and neither counted in the numerator. Reporting “percent susceptible plus intermediate” inflates apparent coverage and should not be done silently. M39-Ed5 specifically added guidance on the “^” annotation for intermediate breakpoints where an agent concentrates in urine — a case where an intermediate result genuinely does carry different clinical meaning depending on the site of infection.

Note also that CLSI has introduced the “susceptible-dose dependent” (SDD) category for certain organism–agent combinations, which is not the same as intermediate and depends on achieving a specific dosing regimen. Decide explicitly how SDD is displayed, and footnote it.

6. How often to publish

Annual publication is the near-universal practice, and the reason is statistical rather than regulatory: shorter intervals rarely accumulate enough isolates for stable percentages, while longer intervals let the report drift behind real resistance trends. Facilities with unusually high volume sometimes publish more frequently for high-count organisms only. Confirm the interval M39-Ed5 recommends before stating one as a requirement in your own policy.

How to read an antibiogram correctly

Most misuse of an antibiogram comes from reading a number as if it meant something it does not.

The percentage is conditional, not absolute

“82% susceptible to ciprofloxacin” means of the E. coli isolates we tested in this period, 82% were susceptible. It does not mean an individual patient has an 82% chance of a susceptible organism. The isolates that reach a microbiology laboratory are the ones a clinician thought worth culturing — a sicker, more treatment-exposed, more previously-hospitalised population than the average patient with that infection. Antibiograms are systematically biased toward resistance relative to the community, and this bias is strongest for organisms that are rarely cultured in mild disease.

Rows are not independent

A common error is to read down a column and pick the agent with the highest overall susceptibility. Empirical therapy has to cover the likely organisms for that syndrome, weighted by how often each causes it — not the organism with the most favourable row. An agent with 95% susceptibility against an organism that causes 3% of your cases is not a better empirical choice than an agent with 85% susceptibility against the organism causing 60% of them.

Low denominators produce meaningless precision

If a cell is derived from a small number of isolates, a single additional resistant isolate can swing the percentage by tens of points. Reporting a percentage without the isolate count invites over-interpretation. Display n for every organism, and suppress cells below your chosen threshold rather than printing a fragile number.

The antibiogram describes the past

An annual antibiogram published in March describes the previous calendar year. During an active outbreak — when the relevant question is what the epidemic curve says about when and how transmission happened rather than what resistance looked like last year — or after a change in formulary or referral pattern, it can be materially out of date. It is a baseline for empirical therapy, not a substitute for current local surveillance or for the patient’s own prior culture history — which for an individual patient is a far stronger predictor than the facility aggregate.

The antibiogram and antimicrobial stewardship

The antibiogram is the evidentiary base for empirical therapy guidelines, and CLSI added explicit stewardship-use guidance in the Fifth Edition. In practice the stewardship programme uses it to:

• set and periodically revise syndrome-specific empirical regimens (community-acquired pneumonia, urinary tract infection, intra-abdominal infection) against local rather than national resistance;
• identify agents whose local susceptibility has fallen far enough that they should be removed from empirical guidelines even if they remain on formulary;
• detect divergence between units, which frequently signals a transmission problem rather than a prescribing problem — and therefore routes to infection prevention;
• track the effect of a stewardship intervention over successive annual reports, with the caveat that resistance responds slowly and confounded by case-mix change.

For how the stewardship programme itself is structured, staffed and assessed against the CDC Core Elements, see the dedicated guide on antimicrobial stewardship programmes — this page deliberately does not repeat that material. The relationship runs both ways: an antibiogram with no stewardship programme to act on it changes no prescribing, and a stewardship programme without a current antibiogram is guessing at local epidemiology.

Who builds it, and who owns it

Antibiogram production sits at a junction between departments, and the failure mode is a report that is technically produced but clinically unowned:

• Clinical microbiology owns the data extraction, de-duplication and accuracy of the susceptibility results.
• Pharmacy / stewardship owns the agent selection, the translation into empirical guidelines, and dissemination to prescribers.
• Infection prevention uses it alongside surveillance data — see the role of the infection preventionist — and is the route by which an unexpected resistance signal becomes a transmission investigation.
• Information systems own the LIS/EHR extract, which M39-Ed5 singles out as a modern source of error: the same query run against the instrument, the LIS and the EHR can return different isolate sets.

Enhanced antibiograms

A conventional antibiogram gives one percentage per organism–agent pair. An enhanced antibiogram — a category M39 addresses directly — presents combination or conditional data instead, for example the proportion of isolates susceptible to at least one of two agents used together, or susceptibility conditioned on a resistance marker detected by rapid diagnostics. These are more useful for empirical decision-making and considerably harder to produce correctly, since combination percentages cannot be derived by multiplying the two individual percentages — susceptibility to two agents is usually correlated, not independent.

If you are producing an enhanced antibiogram, this is the point at which reading M39-Ed5 directly stops being optional.

Antibiograms outside the acute hospital

M39-Ed5 explicitly extended coverage to long-term care facilities, multi-facility (system-wide) antibiograms, and veterinary practice. Each brings its own problem:

• Long-term care has low culture volumes, high colonisation rates, and a substantial fraction of cultures sent for asymptomatic bacteriuria — the isolate mix is not comparable to acute care and should not be merged into an acute-care table.
• Multi-facility antibiograms solve the small-denominator problem but only if the constituent facilities genuinely share an organism population; combining a tertiary transplant centre with a rural critical access hospital produces a number that describes neither.
• Veterinary antibiograms use different breakpoints and different agent panels, and are not interchangeable with human data.

Frequently asked questions

What is a cumulative antibiogram?

A periodic summary — conventionally annual — of the percentage of bacterial isolates at a facility that tested susceptible to each antimicrobial agent, aggregated by organism. Its purpose is to guide empirical therapy before culture and susceptibility results are available for an individual patient.

What standard governs antibiograms?

CLSI M39, Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, Fifth Edition, published 24 January 2022, replacing M39-A4 (2014). It is a paywalled CLSI guideline; the specific numeric thresholds it recommends should be read from the document itself.

Why only the first isolate per patient?

Because repeat isolates from the same patient are highly correlated. Including them lets a single heavily-cultured patient dominate an organism’s row and biases the table toward resistance, which in turn pushes empirical therapy toward unnecessarily broad agents.

Should intermediate isolates be counted as susceptible?

No — the conventional presentation is percent susceptible only, with intermediate isolates in the denominator. Reporting susceptible-plus-intermediate overstates coverage. The narrow exception M39-Ed5 addresses is agents that concentrate in urine, annotated with “^”.

Can I use another hospital’s antibiogram?

Only as a rough sanity check. Resistance patterns are local, driven by that facility’s case mix, referral network, formulary and prescribing history. Using a neighbouring institution’s data as a substitute for your own is the specific error the antibiogram exists to prevent.

Is an antibiogram required?

Requirements vary by accreditor, by state, and by programme. Antimicrobial stewardship requirements applying to hospitals presuppose access to local susceptibility data, and laboratory accreditors have their own expectations. Confirm the specific requirement that applies to your facility with your accreditor and state agency rather than assuming a single national rule — this is an area where obligations genuinely differ.

How does an antibiogram differ from surveillance data?

An antibiogram summarises susceptibility of isolates from clinical cultures to guide therapy. Infection surveillance — for example the CLABSI and CAUTI definitions used for NHSN reporting — counts infections against standardised case definitions to measure infection rates. They draw on overlapping laboratory data and answer completely different questions.

Related reading

• Antimicrobial Stewardship Program: CDC Core Elements and Requirements
• Infection Preventionist: Role, Responsibilities, and CBIC Certification
• Transmission-Based Precautions: Contact, Droplet, and Airborne
• CLABSI: NHSN Surveillance Definition and Reporting
• CAUTI: NHSN Surveillance Definition and Reporting
• Patient Safety & Infection Prevention

See also: Defined Daily Dose.