Antibiotic resistance rates are local, and national averages mislead

A national resistance percentage is an average of populations that behave nothing alike. Here is what actually varies, why it varies, and how to measure the version that applies to your own patients.

14 July 2026 8 min read

Antibiotic resistance rates are local. A national figure is an average across hospitals with different case mixes, different antibiotic exposure and different populations, so it describes none of them well. If you want a number that can guide an empiric decision, it has to come from the isolates your own laboratory reported.

What actually varies, and by how much

Three things move a susceptibility percentage, and none of them are the country you are in.

The first is the population the specimen came from. Outpatient urine cultures come from people who have mostly not been in hospital and have had less antibiotic exposure. Intensive care respiratory cultures come from people who have had both, often within the last fortnight. Grouping them into one figure produces a number that overstates the risk for one prescriber and understates it for the other.

The second is the organism mix. A percentage is per organism and drug pair, but the drug you would reach for empirically depends on which organisms are present at all. Urine is dominated by Escherichia coli. Respiratory specimens from ventilated patients bring Pseudomonas aeruginosa and Staphylococcus aureus, and the useful agents are different.

The third is method. Two facilities can report different numbers for the same organisms because one de-duplicated repeat isolates and the other did not, or because one is still on a breakpoint table that has since been revised.

The de-duplication effect is larger than people expect

Cumulative antibiogram guidance, specifically CLSI document M39, says to count the first isolate per patient per organism in the analysis period. The reason is arithmetic rather than ideology. Patients who are cultured repeatedly are, on average, sicker and more heavily exposed to antibiotics, so their isolates are more likely to be resistant. Counting every isolate lets a small number of patients contribute a large number of rows.

In the synthetic dataset behind the builder on this site, about one isolate in five belongs to a patient already in the set. Turning the first isolate rule off changes the numbers visibly, and it changes them in one direction. That is the whole argument for the rule, made in one click.

Why the intensive care cut is the one everyone asks for

Every stewardship pharmacist has had the same conversation with an intensivist: the antibiogram says 82 percent, and the unit says that has not been their experience. Usually both are right. The facility-wide figure includes a large outpatient population that pulls the average up, and the unit-level figure is several points lower.

The fix is not to argue about the number. It is to produce the cut, so the empiric guideline for that unit rests on that unit's isolates. Once the cut is a filter rather than a project, the conversation moves on to what to do about it.

Where national data is genuinely useful

None of this means public surveillance is worthless. National and international programmes, including the World Health Organization GLASS programme and the antibiotic resistance reporting published by the United States Centers for Disease Control and Prevention, exist to answer policy questions: which resistance mechanisms are spreading, where investment should go, which organisms deserve new drug development.

Those are real questions and they need aggregate data. They are simply not the question a prescriber has at two in the morning, which is what to start now, in this patient, in this unit.

What to measure instead

Your own isolates, de-duplicated to the first per patient per organism
Cut by specimen source first, because it changes the organism mix
Then cut by unit, because it changes the exposure history
With the isolate count next to every percentage, so thin cells are obvious
And with the rules stated on the output, so the report can explain its own trend

The practical objection, and the honest answer

The reason most facilities publish one facility-wide antibiogram a year is not that anybody thinks it is the best possible report. It is that each additional cut used to cost another week of somebody's time in a spreadsheet, and there is no week available.

That constraint is what changes when the cut is a filter. It does not make the underlying epidemiology simpler, and it does not remove the need for a microbiologist to look at the output. It removes the reason to not ask the question.