Technology

A SQL SELECT statement defined in a .sql file within a dbt project that transforms source data into a table or view in the target data warehouse, with lineage automatically tracked through ref() and source() functions. In healthcare data engineering, dbt models are used to build standardized clinical and claims data marts, implement HEDIS measure logic, create DIM_MEMBER and FACT_CLAIM_HEADER tables, and apply ISO-11179 column naming conventions systematically across an entire data platform.

A reusable collection of dbt models, macros, tests, and seeds published to the dbt Package Hub or a private repository that extends dbt project capabilities with pre-built functionality. Widely used healthcare data engineering dbt packages include dbt_utils for common SQL utility macros, dbt_expectations for extended data quality tests, and specialty packages for FHIR data modeling, HEDIS measure calculation, and claims data transformation standardization.

A small reference dataset loaded from a CSV file into the data warehouse as a table through the dbt seed command, used for static lookup data that changes infrequently such as ICD-10 code reference tables, CARC and RARC code descriptions, ZIP code to county crosswalks, and HEDIS measure specifications. In healthcare data engineering, dbt seeds provide version-controlled, reproducible reference data that eliminates dependence on manual database loads for frequently referenced lookup tables.

An automated data validation assertion defined in a dbt project that runs after model execution to verify data quality expectations, such as not_null, unique, accepted_values, and relationships tests for referential integrity. In healthcare data engineering, dbt tests enforce critical data quality rules such as NPI length validation, ICD-10 code format checks, and claim amount non-negativity, providing automated regression testing that catches data quality regressions when source system schemas or business logic changes.

The process of removing or transforming protected health information from a dataset such that the remaining data cannot reasonably be used to identify an individual patient, enabling the data to be used for analytics, research, and secondary purposes outside the HIPAA Privacy Rule. HIPAA defines two de-identification methods: expert determination by a qualified statistician, and safe harbor removal of 18 specific identifiers including names, geographic data smaller than state, dates more specific than year for patients over 89, and direct identifiers.

Unique system-generated key assigned to a debit transaction representing a charge, balance due, or amount owed within healthcare billing and accounts receivable systems. Supports tracking of financial obligations across patient accounts, claims adjudication, and revenue cycle management workflows.

A boolean indicator marking a record as logically deleted in a healthcare data system without physically removing it from the database. Supports soft delete patterns required for HIPAA audit trail compliance. Records with delete flag set to true are excluded from active data queries but retained for historical reporting.

An open-source storage layer developed by Databricks that adds ACID transaction support, scalable metadata handling, and data versioning to Apache Spark and cloud object storage such as AWS S3, Azure Data Lake Storage, and Google Cloud Storage. Delta Lake is widely used in healthcare data platforms to enable reliable upserts for slowly changing dimension management, time travel queries for point-in-time analytics, and schema enforcement preventing corrupt data from reaching downstream analytics layers.

Unique system-generated key assigned to a dermatology specialty record or encounter within clinical and claims systems. Enables tracking of skin-related diagnoses, procedures, and specialist referrals across care management platforms, utilization reporting, and specialty network workflows.

A record linkage methodology that matches patient records based on exact agreement on one or more defined identifier fields, such as Social Security Number, date of birth plus last name, or medical record number plus facility code. Deterministic matching is faster and simpler than probabilistic matching but requires high-quality, complete identifier data to avoid both false matches and missed matches in master patient index implementations.

A graph structure in which nodes represent tasks and directed edges represent execution dependencies between tasks, with no circular dependencies, used in data pipeline orchestration tools such as Apache Airflow to define the execution order and dependency relationships of a multi-step data workflow. In healthcare data engineering, DAGs define the sequencing of ingestion, validation, transformation, and quality check tasks that must execute in specific order to produce correct analytics outputs.

Unique system-generated key assigned to a patient care directive, such as an advance directive or do-not-resuscitate order, within clinical documentation systems. Supports consistent retrieval and enforcement of patient care instructions across care settings and clinical decision-making workflows.

Unique system-generated key assigned to a clinical or administrative document within health information and document management systems. Enables consistent retrieval, version control, and audit tracking of records such as discharge summaries, consent forms, and clinical notes across healthcare data platforms.

Unique system-generated key assigned to a specific medication dose record within pharmacy and medication management systems. Supports tracking of prescribed and administered dose amounts across pharmacy dispensing, medication reconciliation, and clinical decision support workflows.

Unique system-generated key assigned to a treatment or therapy duration record within clinical and care management systems. Enables consistent tracking of prescribed treatment time periods across medication management, therapy planning, and longitudinal care coordination workflows.

A database security feature that automatically replaces sensitive data values with masked representations such as asterisks, partial values, or null values when queried by users without the required access privilege to see the full value, without changing the actual stored data. In healthcare data platforms, dynamic data masking protects PHI fields such as member Social Security Number, date of birth, and name from unauthorized access while allowing authorized users to see unmasked values for legitimate business purposes.

Unique system-generated key assigned to an endocrinology specialty record or encounter within clinical and claims systems. Enables tracking of hormone-related diagnoses, specialist visits, and treatment plans across care management platforms, utilization reporting, and specialty referral workflows.

An extended master patient index that resolves patient identity not just within a single healthcare organization but across multiple organizations in a health system, payer network, or health information exchange. The EMPI maintains a reference index of global patient identifiers that map to local identifiers in each participating system, enabling longitudinal patient records that follow a patient across providers, payers, and care settings within a defined network.

Unique system-generated key assigned to a defined care episode representing a patient's treatment period for a specific condition. Used in clinical, claims, and population health systems to group related encounters, procedures, and costs into a continuous care event for analysis and reporting.

A boolean indicator marking a record that failed validation or processing rules in a healthcare data system. Used in ETL pipelines, claims adjudication, and EDI processing to tag records requiring manual review or reprocessing. Error flag records are quarantined from production data until resolved.