Amazon HealthLake
A fully managed service that securely stores, transforms, and analyzes FHIR-compliant healthcare data, accelerating data lake construction for medical organizations
Overview
Amazon HealthLake is a fully managed service that provides a healthcare data store compliant with the HL7 FHIR R4 (Fast Healthcare Interoperability Resources) standard. It stores electronic health records (EHR), insurance claims data, clinical trial data, and more as unified FHIR resources, enabling natural language processing for structuring unstructured data, timeline-based patient data visualization, and predictive analytics through machine learning models. As a HIPAA-eligible service, it includes encryption and access control for protected health information (PHI) by default.
FHIR Data Store and Data Ingestion Pipeline
HealthLake's data store provides a FHIR R4-compliant RESTful API and handles the FHIR R4 resource types listed in the developer guide, including Patient, Observation, Condition, and MedicationRequest. Data ingestion includes import jobs from S3 buckets, as well as writes through the API with individual resource CRUD operations and FHIR Bundles (batch or transaction). C-CDA and CSV data from existing systems can be converted to FHIR R4 with the Data Transformation Agent (in preview as of October 2026), while for formats such as HL7 v2, one configuration converts the data to FHIR format with a custom transformation Lambda or a similar component before ingestion. Writes to the data store are version-controlled, and past versions are read with the FHIR history and vread interactions. Integration with the SMART on FHIR authentication framework enables fine-grained access control based on OAuth 2.0.
Integrated Medical NLP and Structuring Unstructured Data
Integrated Medical NLP can be applied to unstructured text such as clinical notes and discharge summaries stored in HealthLake. The developer guide states that this feature is turned off by default and is turned on through a support case (as of October 2026). This feature calls the Amazon Comprehend Medical DetectEntities-V2, InferICD10-CM, and InferRxNorm API actions to extract medical entities along with their relationships and traits. The results, including ICD-10-CM and RxNorm codes, are stored as an extension to the DocumentReference resource, and entities with the sign, symptom, or diagnosis traits become new Condition and Observation resources linked to the source document through a Linkage resource. According to the developer guide, the resources generated by NLP can be read with GET requests but are not covered by FHIR API search, and are searched with SQL in Athena.
Analytics Platform Integration and Compliance Design
HealthLake data can be directly analyzed with Athena or Amazon Quick Sight (formerly Amazon QuickSight) via Lake Formation. The developer guide describes each FHIR resource type as an Apache Iceberg table that Athena queries with SQL without an export first. One configuration combines SageMaker and uses HealthLake data for training patient readmission risk prediction and disease progression models. For compliance, server-side encryption with KMS, audit logging of all API operations via CloudTrail, and private network access through VPC endpoints are available by default. HealthLake is a HIPAA-eligible service, and HIPAA workloads are handled after signing a BAA (Business Associate Agreement).
References (Official AWS Resources)
The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.
- Amazon Comprehend official page
- Amazon SageMaker official page
- AWS Lake Formation official page
- AWS Documentation (official)
If this page and the official documentation disagree, treat the official documentation as authoritative.