Technology

The process of encrypting data as it moves between systems, services, and users over networks to prevent interception and unauthorized access to healthcare data during transmission. HIPAA technical safeguards require encryption of ePHI in transit, implemented through TLS 1.2 or higher for API communications, SFTP for file transfers, and VPN or private connectivity for inter-system data flows in healthcare data integration architectures.

A data management architecture that uses metadata, knowledge graphs, and AI-driven automation to integrate data across heterogeneous environments — on-premises, cloud, and multi-cloud — providing a unified semantic layer that simplifies data discovery, access, and governance without requiring physical data movement. In healthcare, data fabric approaches are used to create virtual integrated views across EHR systems, claims databases, and external reference data sources.

A technology stack or managed service providing capabilities for connecting disparate healthcare source systems, transforming data between formats and standards, and delivering integrated data to target warehouses, lakes, or applications. Modern healthcare data integration platforms combine CDC-based replication tools such as Fivetran or Airbyte for structured source systems with HL7 and FHIR integration capabilities for clinical systems, replacing legacy point-to-point interfaces with a managed integration layer.

A data organization pattern dividing a healthcare data lake into distinct areas with different access controls, data quality levels, and retention policies, such as a raw landing zone for unprocessed source files, a curated zone for cleaned and validated data, and a consumer zone for analytics-ready datasets. Data lake zones provide a governance structure that enables different user groups to access appropriate data quality levels while maintaining raw data immutability for audit and reprocessing purposes.

A modern data architecture pattern that combines the low-cost storage and schema flexibility of a data lake with the ACID transaction support, data quality enforcement, and performance optimization of a traditional data warehouse. Healthcare data lakehouses typically store raw clinical, claims, and operational data in open table formats such as Delta Lake or Apache Iceberg on cloud object storage, with a structured processing layer applying medallion architecture patterns to progressively refine data quality from raw ingestion through analytics-ready marts.

An open storage specification that defines how data files, metadata, and transaction logs are organized in cloud object storage to support ACID transactions, schema evolution, and time travel queries from multiple query engines. The three major competing table formats are Delta Lake, Apache Iceberg, and Apache Hudi, each offering different trade-offs in terms of ecosystem compatibility, compaction efficiency, and merge-on-read versus copy-on-write strategies for healthcare data platform architecture decisions.

A metadata capability that tracks the origin, movement, transformation, and consumption of data assets throughout their lifecycle in a data platform, enabling data engineers and analysts to understand where a field came from, what transformations were applied, and which downstream reports or models depend on it. Data lineage is critical in healthcare data engineering for HIPAA audit trails, impact analysis when source schemas change, and root cause investigation when data quality issues arise in production reports.

A decentralized data architecture paradigm that treats data as a product owned and served by domain teams rather than managed by a central data engineering team, with each domain responsible for the quality, documentation, and accessibility of its own data products. In healthcare organizations, a data mesh approach assigns data product ownership to clinical, claims, pharmacy, and operations domains, reducing the central data team bottleneck while establishing platform-level standards for interoperability and governance.

The process of structuring relational database tables to reduce data redundancy and improve integrity by organizing data into related tables with well-defined keys, following normal forms such as first, second, and third normal form. In healthcare data engineering, normalization is applied in operational source databases and staging layers, while dimensional models intentionally denormalize data into wide fact and dimension tables optimized for analytical query performance rather than write efficiency.

A practice and set of tools focused on monitoring, alerting, and understanding the health and quality of data as it flows through pipelines, warehouses, and consumption layers — analogous to software observability for applications. Data observability in healthcare data platforms covers freshness monitoring to detect stale data feeds from EHR or claims systems, volume anomaly detection to catch source system outages, schema change detection, and data distribution monitoring to identify upstream data quality issues.

The automated coordination and scheduling of data pipeline tasks, dependencies, retries, and monitoring in a defined execution order using an orchestration tool such as Apache Airflow, Prefect, Dagster, or dbt Cloud. In healthcare data platforms, pipeline orchestration ensures that dependent jobs such as member eligibility loads, claims processing, and quality measure calculations run in the correct sequence with appropriate error handling and alerting when upstream data feeds from EHR or payer systems are delayed or incomplete.

A curated, documented, and governed data asset designed for consumption by a defined set of users or downstream systems, treated with product management discipline including ownership, quality guarantees, versioning, and consumer support. In healthcare data mesh architectures, data products such as a claims mart, member eligibility API, or HEDIS measure dataset are owned by domain teams who are accountable for their accuracy, freshness, and discoverability by consuming analytics and operational teams.

A structured set of dimensions, rules, metrics, and processes used to measure and improve the quality of healthcare data across attributes such as completeness, accuracy, consistency, timeliness, validity, and uniqueness. In healthcare data platforms, data quality frameworks are implemented through dbt tests, Great Expectations profiles, or custom SQL validation rules applied at each layer of a medallion architecture to prevent data quality issues from propagating to downstream analytics and clinical decision support.

A governance policy defining how long different categories of healthcare data must be retained, who is responsible for retention enforcement, and what processes govern secure deletion or archival when retention periods expire. HIPAA requires covered entities to retain certain records for at least six years from creation or last effective date, while state laws and CMS conditions of participation may require longer retention periods for specific data types including medical records and billing documentation.

The ongoing accountability and responsibility assigned to specific individuals or teams for ensuring the quality, accuracy, consistency, and appropriate use of data within their assigned domain in a healthcare organization. Data stewards serve as the business owners of specific data domains such as member data, claims data, or provider data, making decisions about definitions, quality standards, and access policies in collaboration with technical data engineering teams and the central data governance function.

A data security technique that replaces sensitive data values such as Social Security Numbers, patient identifiers, and credit card numbers with non-sensitive placeholder tokens that retain the format and referential integrity of the original data but cannot be reversed without access to a secure tokenization vault. In healthcare data platforms, tokenization is used to de-identify PHI in analytics environments, enabling data science and analytics work on representative data without exposing actual patient identifiers.

The practice of tracking changes to data assets, pipeline code, and ML model artifacts using version control systems analogous to software version control, enabling teams to reproduce historical pipeline runs, roll back data transformations, audit who changed what and when, and collaborate on data platform development without overwriting each other work. In healthcare data engineering, dbt projects managed in Git provide version-controlled SQL transformations, while tools like DVC and Delta Lake time travel extend versioning to raw data assets.

A data integration approach that provides a unified semantic layer allowing applications and analytics tools to query data from multiple heterogeneous source systems without physically moving or copying the data into a central repository. In healthcare, data virtualization enables analysts to query across EHR systems, claims databases, and reference data sources through a single virtual layer, reducing ETL complexity and latency for use cases where real-time or near-real-time data access is required.

An allocated compute resource in a cloud data warehouse platform such as a Snowflake virtual warehouse or Databricks cluster that executes SQL queries and data transformation workloads against stored data. In healthcare data platforms, compute cluster sizing and auto-scaling configuration directly affect pipeline latency, query performance for analytics dashboards, and platform cost — with workload isolation through separate clusters for ETL, reporting, and data science a common cost optimization pattern.

A reusable Jinja-templated function defined in a dbt project that generates SQL dynamically, enabling healthcare data engineers to standardize repetitive transformation patterns across models such as HEDIS measure logic templates, PHI masking functions, date spine generation, and ISO-11179 column naming convention enforcement without duplicating code across dozens of individual model files.