Implementing a modern health-data strategy

For implementing your modern healthcare-data strategy, we recommend following these principles:

• Create an operating model for a data-driven organization – Identify the roles, competencies, and the target operating model needed to create a data-driven organization. Cultivate data literacy in business, IT, and anyone involved in patient care, including patients. Embrace the innovative potential of the cloud to accelerate delivery of business value. Start with a hybrid data strategy so that your organization can move quickly. Harness existing on-premises tools and technologies with cloud-based solutions to create nimble and efficient data products. AWS offers a suite of products to adopt hybrid cloud models to help accelerate your transition to cloud.

• Work backwards from frontline needs – For each organizational role, identify what data is needed, when, and in what format. Next, determine the origin of the data and how to deliver it on time. Deliver the data in a format that the users can easily understand and apply. For example, use AWS HealthLake and Amazon QuickSight to build dashboards that include understandable data visualizations. Where possible, build self-service solutions that end users can access and manipulate without the need for analyst or data scientist intervention.

• Automate the data pipeline – If a frontline healthcare worker must manually transfer data from one system to another, that step delays data delivery. It introduces data gaps and errors, distracts frontline staff from patient care, erodes staff morale, and reduces staff productivity. Automation might seem expensive, but consider the total cost of manual data processing in your return-on-investment (ROI) calculations. If data sources require manual data transfer, consider whether you can keep the data in place. To acquire data from medical devices, you can use AWS integration with medical devices, and use AWS Glue to build an operationally efficient data pipe.

• Move from monolith to modular – Monolithic systems have interdependencies that prevent innovation in any component and that complicate troubleshooting when things go wrong. A modern health data strategy should be modular: comprised of independent components with well-defined interfaces so that you can innovate in each module without disrupting other modules. Use data stores that support interoperability standards. For example, consider using HealthLake, a HIPAA-eligible Fast Healthcare Interoperability Resources (FHIR)-compatible data store, along with off-the-shelf data-ingestion software, and use AWS HealthOmics to transform genomic, transcriptomic, and other omics data.

• Use managed and serverless services – Decrease the undifferentiated heavy lifting of server and operating-system configuration, patch management, and monitoring by using managed services, where the cloud service provider manages the underlying infrastructure for you. Shift your IT staff resources from system management (keeping the lights on) to data innovation. For example, use AWS Lambda or AWS Fargate for compute services, Amazon Aurora Serverless for relational databases, and Amazon Redshift Serverless for your data warehouse.

• Simplify and shorten data pipelines – Moving and transforming data is potentially expensive and time-consuming. It can also introduce errors into data solutions. To optimize cost, accelerate data delivery, and improve data quality, do the following:

  • Use data where it lives.

  • Minimize extract, transform, and load (ETL) operations.

  • Use federated data access.

  For example, use AWS managed services to implement data mesh architectures, minimize the overhead involved in data movement, and use federated query.

For additional information and details on implementing an architecture to support a modern health data strategy, see Appendix D: Additional guidance for implementing a modern health data strategy.