name: AWS HealthLake FinOps Framework
description: FOCUS-aligned FinOps reference for teams consuming AWS HealthLake. Primary cost drivers are data store hourly charges, storage beyond the included 10 GB, FHIR query volume, medical NLP character processing, data export, and FHIR Subscription event delivery. Optimization strategies center on right-sizing data stores, batching FHIR writes, managing NLP usage, using Athena for analytics over repeated API queries, and leveraging the AWS Pricing Calculator for TCO modeling.
specificationVersion: '1.0'
url: https://aws.amazon.com/healthlake/pricing/
costDimensions:
  - name: Data Store Hourly Charge
    description: Base charge for each active HealthLake data store, billed hourly. Includes the first 10 GB of storage across all data stores and 3,500 bundled queries per hour. This is the primary fixed-like cost — reducing the number of active data stores directly lowers this charge.
    billingModel: hourly
    currency: USD
    costDrivers:
      - name: Data Store Hour
        rate: '$0.27 per data store per hour'
        notes: Approximately $194/month per always-on data store. Deleting unused or dev/test data stores eliminates this charge.
  - name: Storage
    description: Storage costs for FHIR resources and NLP-extracted entities stored beyond the included 10 GB allowance, billed monthly per GB.
    billingModel: per-GB-month
    currency: USD
    costDrivers:
      - name: Advanced Tier Storage
        rate: '$0.37 per GB per month'
        tier: Advanced
      - name: Standard Tier Storage
        rate: '$0.25 per GB per month'
        tier: Standard
      - name: Included Allowance
        rate: Free
        details: First 10 GB across all data stores is included at no extra charge.
  - name: FHIR Queries
    description: Query charges for FHIR read and search operations beyond the 3,500 queries per hour included in the data store hourly fee.
    billingModel: per-10k-queries
    currency: USD
    costDrivers:
      - name: Additional Queries (Advanced)
        rate: '$0.048 per 10,000 queries'
        tier: Advanced
      - name: Additional Queries (Standard)
        rate: '$0.015 per 10,000 queries'
        tier: Standard
      - name: Bundled Queries
        rate: Free
        details: 3,500 queries per hour included in data store charge.
  - name: Medical NLP
    description: Character-based charge for processing unstructured medical text in DocumentReference resources through the integrated NLP pipeline. NLP extracts entities, relationships, traits, and PHI from free-text clinical notes.
    billingModel: per-100-characters
    currency: USD
    costDrivers:
      - name: NLP Character Processing
        rate: '$0.0010 per 100 characters analyzed'
        notes: A 10,000-character medical note (maximum allowed) costs $0.10. Costs scale linearly with ingestion volume of unstructured notes.
  - name: Data Export and Analytics Transformation
    description: Per-GB charge for exporting FHIR data from a data store to Amazon S3 or transforming data for analytics use with Amazon Athena.
    billingModel: per-GB
    currency: USD
    costDrivers:
      - name: Export / Transform
        rate: '$0.19 per GB'
        notes: Applies to both StartFHIRExportJob output and analytics transformation output.
  - name: FHIR Subscriptions
    description: Event delivery costs for FHIR R4 Subscription notifications when resources matching subscription criteria are created or updated. Two delivery mechanisms available — Amazon EventBridge and REST-Hook.
    billingModel: per-million-events
    currency: USD
    costDrivers:
      - name: EventBridge Subscription Events
        rate: '$2.00 per 1 million events'
        freeAllowance: '100,000 events per month'
      - name: REST-Hook Subscription Events
        rate: '$0.90 per 1 million events'
        freeAllowance: '100,000 events per month'
optimizationStrategies:
  - name: Minimize Active Data Store Count
    description: Each data store costs ~$194/month regardless of usage. Consolidate environments by sharing a single Advanced data store across teams where HIPAA BAA and data segregation requirements allow, or delete idle development and test data stores when not in use.
    impact: high
    effort: medium
  - name: Use Athena for Analytics Workloads
    description: Export FHIR data to S3 once using StartFHIRExportJob and query it with Amazon Athena rather than issuing repeated FHIR API GET and search requests. Athena SQL queries against S3 are far cheaper than per-query HealthLake charges at scale.
    impact: high
    effort: medium
    url: https://docs.aws.amazon.com/healthlake/latest/devguide/exporting-and-analyzing.html
  - name: Batch FHIR Writes Using Bundle Resources
    description: Use FHIR Bundle resources (batch type preferred over transaction) to group multiple resource writes into a single API call. This reduces per-request overhead and makes more efficient use of write capacity quotas. Batch bundles use half the write capacity of transaction bundles.
    impact: medium
    effort: low
    notes: Maximum 500 resources per Bundle.
  - name: Control NLP Ingestion Volume
    description: Medical NLP costs accumulate per character for every DocumentReference with unstructured text. Disable NLP for data stores where NLP-enriched search is not required, or pre-filter documents to exclude clinical notes that do not benefit from entity extraction before ingestion.
    impact: medium
    effort: medium
    url: https://docs.aws.amazon.com/healthlake/latest/devguide/using-medical-nlp.html
  - name: Schedule Import Jobs During Off-Peak Hours
    description: StartFHIRImportJob is asynchronous — queue large import jobs during off-peak periods to avoid contention. Only one concurrent import job is allowed per account, so batching strategically reduces wall-clock time and allows the data store to serve queries at full throughput during business hours.
    impact: low
    effort: low
  - name: Right-Size Export Jobs
    description: Export only the FHIR resource types and date ranges needed for analytics workloads rather than full data store exports. This reduces the per-GB export charge and S3 storage costs for downstream data.
    impact: medium
    effort: low
  - name: Use FHIR Subscription REST-Hook Over EventBridge When Cost Matters
    description: REST-Hook delivery ($0.90/M events) is less expensive than EventBridge delivery ($2.00/M events) for high-volume subscription scenarios. Evaluate architecture requirements — EventBridge provides richer routing and fanout capabilities that may justify the premium.
    impact: low
    effort: medium
  - name: Use AWS Cost Allocation Tags
    description: Apply AWS resource tags to each HealthLake data store to attribute costs by team, project, environment (prod/dev/test), or clinical domain. Up to 200 tags per data store are supported. Tags appear in AWS Cost Explorer for charge-back and show-back reporting.
    impact: medium
    effort: low
    url: https://docs.aws.amazon.com/healthlake/latest/APIReference/API_TagResource.html
  - name: Estimate TCO with AWS Pricing Calculator
    description: Use the AWS Pricing Calculator HealthLake template to model total cost of ownership before provisioning, factoring in data store hours, expected storage growth, query volume, NLP character volume, and subscription event rates.
    impact: high
    effort: low
    url: https://calculator.aws/#/addService/HealthLake
costAllocation:
  - dimension: DataStore
    description: Each HealthLake data store generates a distinct hourly charge visible in Cost Explorer. Use one data store per billing boundary (environment, product line, clinical domain) for clean attribution.
  - dimension: Environment
    values:
      - Production
      - Staging
      - Development
      - Test
    notes: Tag data stores by environment and delete non-production stores when idle to eliminate hourly charges.
  - dimension: ResourceTags
    description: AWS resource tags on data stores propagate to Cost Explorer line items. Recommended tag keys include Team, Project, CostCenter, and Environment.
  - dimension: NLPVolume
    description: Track DocumentReference ingestion volume (character counts) to forecast and attribute NLP charges by clinical data source or ingestion pipeline.
  - dimension: ExportVolume
    description: S3 export job sizes are visible in CloudWatch Logs and the export job response. Use these to attribute analytics transformation costs to downstream consumers.
monitoring:
  - name: CloudWatch Metrics
    description: Monitor HealthLake data store operation counts, latencies, and errors via Amazon CloudWatch to detect unusual query volume spikes that drive cost.
    url: https://docs.aws.amazon.com/healthlake/latest/devguide/monitoring-healthlake.html
  - name: AWS Cost Explorer
    description: Use Cost Explorer with the HealthLake service filter and resource tag dimensions to view daily and monthly charges broken down by data store and cost category.
    url: https://aws.amazon.com/aws-cost-management/aws-cost-explorer/
  - name: AWS Budgets
    description: Set monthly budgets for HealthLake spend with alert thresholds to receive notifications before costs exceed planned levels.
    url: https://aws.amazon.com/aws-cost-management/aws-budgets/
  - name: CloudTrail API Audit
    description: AWS CloudTrail logs all HealthLake management API calls (CreateFHIRDatastore, StartFHIRImportJob, etc.) with timestamps and caller identity, enabling cost audit trails and detecting unauthorized data store creation.
    url: https://docs.aws.amazon.com/healthlake/latest/devguide/logging-using-cloudtrail.html