Description #

The complete specification of the feature space behind the provider level healthcare claims anomaly detection work: 29 entries covering identifiers, demographics, coverage, clinical indicators, financial fields, and engineered measurements, of which 19 are implemented in the published reference model and 10 specify a not yet implemented provider level aggregation layer. Documents the exact, order dependent computation for every engineered feature, an unresolved chronic condition coding ambiguity that can silently invert the health burden index, and the fairness restrictions that exclude race from any deployed scoring model.

Files #

PathSizeSHA-256
files/feature-dictionary-healthcare-provider-fraud-detection.pdf 159,167 bytes 3cff96f873397e067cd1989bf181879298ed471756237d49f1836785a51f15a7

Each file's SHA-256 is listed above. To confirm a download is unmodified: shasum -a 256 filename

Licence: CC BY 4.0

← Back to Initiative 1