Aggregate surveillance data only. No patient information and no treatment recommendations.

Methods and definitions

How NovAMR turns reports into explorable evidence

The Explorer separates what a source reported from what NovAMR normalized, calculated, and mapped.

Source before summaryEvery public value should be traceable, reviewable, and correctable.

Pipeline

From original document to public row

  1. Discover and identifyA hospital, laboratory, regional, or national aggregate antibiogram is linked to its public source or privately submitted for review.
  2. Extract with provenanceEach candidate value retains its report ID, original labels, year, location, and table or page locator when available.
  3. Normalize without erasingCanonical organism and antibiotic names support search while the source's raw wording remains in the release.
  4. Verify before releaseCandidate rows stay outside the public Explorer until they pass source review and enter a dated release.
  5. Aggregate transparentlyFilters, metric availability, denominator class, and time policy determine which rows support a displayed result.

Reported metrics

Susceptibility and resistance remain distinct

Susceptibility uses the source-reported percent susceptible. Resistance uses only a source-reported resistant value explicitly identified as resistance. NovAMR never creates resistance by subtracting susceptibility from 100 because intermediate or other categories may exist.

Denominators

Compatible counts are not silently mixed

Drug-tested n

When a count is explicitly tied to the organism-antibiotic result, it can support the primary count-weighted estimate.

Organism-level n

A source-wide or organism-row count is an approximation class and remains separate from drug-tested weighting.

No compatible n

Rows without a usable denominator are summarized as unweighted evidence with their median and range.

Time and geography

Latest available is the default

By default, NovAMR selects the latest eligible year independently for each source. A single-year filter includes source periods that intersect that year. All years is explicitly marked as pooled and is never the default.

Research-map areas use real administrative boundaries. Local comparison polygons are clipped within those boundaries and should not be interpreted as official service areas, catchments, or clinical regions.

Inspect the implementation inputs

The public data dictionary documents every exported field, null handling, count scope, and interpretation rule.