Data in
Laboratory information system exports in, antibiogram out
The fastest integration project is the one you do not have to start. Every laboratory information system can already produce the file this needs.
Short answer
A laboratory information system export needs seven fields to build an antibiogram: a patient key, the collection date, the specimen source, the unit or location, the organism, the antibiotic and the S, I or R result. MIC values are useful and optional. Anything else in the file is ignored.
The seven columns, and what each one is for
Column names vary between systems and are detected automatically, with a manual override when the guess is wrong.
- Patient key: any stable per-patient identifier, including a hash you generate. Needed only for de-duplication.
- Collection date: drives the period filters and the quarterly trend.
- Specimen source: urine, blood, respiratory, wound and so on. The most useful filter on the report.
- Unit or location: intensive care, inpatient, emergency, outpatient. Free text is grouped during mapping.
- Organism: the reported name. Synonyms and abbreviations are normalised.
- Antibiotic: the agent tested. Codes and full names both work.
- Result: S, I or R. SDD and NS are mapped explicitly rather than guessed.
Wide files and long files
Laboratory exports come in both shapes: one row per isolate with a column per antibiotic, or one row per isolate and antibiotic pair. The long shape is what the browser demo parses today. The wide shape is handled during onboarding, because the mapping is a one-time job specific to your system.
De-identification before you send anything
You do not need to send a real medical record number. A one-way hash generated on your side is enough for the first isolate rule, because the analysis only needs to know that two rows belong to the same patient, never who that patient is.
In the browser demo the question does not arise at all: the file is parsed locally and no row is transmitted.
Related reading: exporting out of WHONET, what gets built from the export. The four steps from susceptibility testing data to an antibiogram cover the method end to end.
Questions
On this page's topic
Do you support HL7 or FHIR feeds?
A scheduled ingest of an export your system already produces is part of the Network plan. A live interface is an Enterprise conversation. Neither is required to get a report.
How large a file can the browser demo handle?
Tens of thousands of rows parse comfortably. Beyond that the practical limit is your machine, and a facility that size is usually on a scheduled export anyway.
What if our unit names are inconsistent?
They usually are. Mapping free-text locations into a small set of reporting groups is part of onboarding, and the mapping stays yours to edit.
More on antimicrobial resistance reporting
- UTI antibiotic resistance Urine is the highest volume specimen most laboratories process, which makes it the one sourc...
- carbapenem resistant organism Carbapenem resistance is the event that changes what a hospital can do next, and it is the l...
- empiric therapy Every empiric decision is a bet placed before the culture is back. The only honest way to im...
- resistance map Public resistance maps are useful for policy and useless for prescribing. The map that chang...
- antibiotic stewardship program A stewardship program lives or dies on whether it can answer questions quickly. Most of the...
- CLSI breakpoints A breakpoint revision can move a susceptibility rate by several points without one organism...
Start with your own export
See your own resistance picture, cut the way you actually work
Build the antibiogram on the synthetic sample dataset, or drop in your own CSV and watch it parse without leaving your browser. When you want an account, one email address is the whole sign-up.
Your file is parsed in your browser and never uploaded. No card required.