Outputs and Contract

Published outputs

empi schema

Table	Description
source_person	Standardized input records combining claims and clinical demographics. One row per source system + source ID.
pair_decisions	All scored pairs with final decision (match, non_match, clerical_review) after applying thresholds, hard rules, and manual overrides.
person_crosswalk	Core mapping table: person_id ← source_system + source_id. Includes active_bool, cluster metadata, and timestamps.
person_counts	Record count per resolved person_id.
person_attrs	Survivorship golden record: one row per person_id with consolidated attributes (name, DOB, SSN, address, phone, email, etc.).
person_attrs_provenance	Audit trail: tall table with person_id, attribute_name, attribute_value, and the source record that contributed it.
person_id_map	Intermediate crosswalk used by input-layer remapping models.
work_queue	Candidate pairs in the clerical-review band, ready for analyst review. Includes work_item_id, pair_key, match_probability, and the source keys for both sides.
work_item_crosswalk	Stable mapping from work_item_id to pair_key across runs.
cluster_id_crosswalk	Tracks cluster merge/split history across runs.
overrides	Manual review decisions written by the review app. Incremental with full_refresh: false.
rematch_requests	Records flagged for re-scoring (from split/carve operations). Incremental with full_refresh: false.

core schema

Table	Description
person	Person dimension: person_id, cluster version, record count, and timestamps.

input_layer schema (remapped)

For each enabled domain, the package publishes a remapped version of the upstream empi_pre__* table with person_id overwritten to the EMPI-resolved value.

Claims domain:

Table	Source
eligibility	empi_pre__eligibility
medical_claim	empi_pre__medical_claim
pharmacy_claim	empi_pre__pharmacy_claim
provider_attribution	empi_pre__provider_attribution (when enabled)

Clinical domain:

Table	Source
patient	empi_pre__patient
appointment	empi_pre__appointment
condition	empi_pre__condition
encounter	empi_pre__encounter
immunization	empi_pre__immunization
lab_result	empi_pre__lab_result
medication	empi_pre__medication
observation	empi_pre__observation
procedure	empi_pre__procedure

All original columns are preserved. Only person_id is replaced with the EMPI-resolved value.

Input contract

Required columns for linkage

The empi_pre__eligibility and empi_pre__patient models feed the identity resolution scorer. These columns are used:

Identity core:

Column	Type	Used for
source_system	string	Record provenance, pair generation
source_id	string	Unique record identifier within source
first_name	string	Fuzzy name comparison
middle_name	string	Fuzzy name comparison
last_name	string	Fuzzy name comparison
name_suffix	string	Exact comparison
birth_date	date	Date comparison, blocking
death_date	date	Date comparison
sex	string	Exact comparison
social_security_number	string	Exact comparison, blocking

Contact:

Column	Type	Used for
address	string	Fuzzy address comparison
city	string	Fuzzy comparison
state	string	Exact comparison
zip_code	string	Exact comparison
phone	string	Exact digit comparison
email	string	Exact comparison, blocking

Demographic:

Column	Type	Used for
race	string	Survivorship only
ethnicity	string	Survivorship only

Provenance:

Column	Type	Used for
data_source	string	Source tracking
ingest_datetime	timestamp	Survivorship timestamp
file_date	date	Survivorship timestamp
file_name	string	Audit trail

Domain enablement

Which empi_pre__* models are required depends on project vars:

Var	When true	Required models
claims_enabled	Claims sources included in linkage	empi_pre__eligibility, empi_pre__medical_claim, empi_pre__pharmacy_claim
clinical_enabled	Clinical sources included in linkage	empi_pre__patient + all clinical entity models
provider_attribution_enabled	Provider attribution remapped	empi_pre__provider_attribution

When a domain is disabled, its corresponding upstream refs are not required for package execution.