How MEDHOST is migrating electronic health record data to AWS for compliance and gaining valuable insights



Original Source Here

Healthcare technology companies often turn to AWS to help them accelerate their clinical and business objectives. MEDHOST has provided enterprise information technology and electronic health record (EHR) solutions to full-service community hospitals for more than 35 years. Today, more than 1,000 healthcare facilities are partnering with MEDHOST and enhancing their patient care and operational excellence with its clinical and financial solutions, which include a fully integrated EHR solution. MEDHOST recently announced that they’re migrating all their EHR data to AWS in order to meet their compliance and analytics needs. This post discusses how MEDHOST is using AWS to give patients easier access to their health data while also bringing advanced analytics and insights to clinicians.

Meeting compliance and analytics needs

The 21st Century Cures Act, designed to give patients and healthcare providers secure access to health information, has timelines requiring healthcare organizations to provide access to patient data through FHIR (Fast Healthcare Interoperability Resources). MEDHOST needed to help their over 1,000 healthcare facility customers be compliant, but standing up and managing a FHIR server that can scale on demand is challenging for individual healthcare facilities. During the planning stages for enabling Cures Act compliance, MEDHOST realized that by enabling FHIR functionality on AWS, they could also take advantage of the ever-growing AWS AI and machine learning (ML) services in order to provide advanced analytics and unlock insights from their structured and unstructured clinical data (such as clinical notes). MEDHOST had the goal of enabling FHIR and analytics for customers by the end of 2021, and so began working closely with their AWS team in 2020 to realize that vision.

Partnering with AWS

The first thing MEDHOST did was begin contributing to the open-source project FHIR Works on AWS. FHIR Works on AWS is a new AWS Solutions Implementation with an open-source software toolkit that can be used to create a FHIR interface over existing healthcare applications and data. MEDHOST began meeting regularly with AWS Interoperability and FHIR experts to carve out a path for enabling FHIR R4 APIs. In parallel to this, MEDHOST also began working with AWS Data Lab Specialists, as well as the newest HIPAA-eligible service for storing and analyzing clinical data, Amazon HealthLake, to explore the built-in medical comprehension Amazon HealthLake offers.

Ingesting and structuring health data

The AWS Data Lab program is an accelerated joint engineering engagement between a team of customer builders and AWS technical resources to create tangible deliverables. The Data Lab Program has two offerings: a Design Lab and a Build Lab. The Build Lab is a 2-to-5-day intensive build by a team of customer builders with guidance from an AWS Data Lab Solutions Architect. The Design Lab is a new offering for customers who need a real-world architecture recommendation based on AWS expertise, but aren’t yet ready to build.

MEDHOST engaged with the AWS Data Lab team to build a working prototype of their end-to-end data pipeline. In pre-lab calls leading up to the lab, the scope of the lab was discussed, several architectures were proposed, and the target state architecture was built. For the Build Lab prototype, the team used select FHIR resource types with both structured and unstructured data.

The following diagram illustrates the overall architecture.

During the Data Lab, a sample dataset of structured EHR resource types were ingested in real time to MEDHOST’s platform in AWS through Amazon Managed Streaming for Apache Kafka (Amazon MSK). The incoming Kafka payload in FHIR R4 format was loaded into Amazon DynamoDB (via FHIR Works on AWS) and Amazon HealthLake using AWS Lambda. The MEDHOST team then made test API calls to ensure the integrity of the data that was loaded. Next, unstructured clinical notes were ingested as DocumentReference resource types into Amazon HealthLake. Because Amazon HealthLake has built-in medical comprehension for DocumentReference types, MEDHOST was able to automatically derive ICD-10 codes, diagnoses, medications, RxNorm, and other medical attributes from the unstructured data within the patients’ health records.

After all resource types were indexed, structured, and appended to Amazon HealthLake, MEDHOST exported the augmented FHIR resources to Amazon S3 in ndjson format, where an AWS Glue crawler job was created to crawl the data and build an AWS Glue Data Catalog. With the table cataloged, Amazon Athena can query all the clinical data using standard SQL. This allowed the clinical data to be easily searched, with multiple resource types joined together for full insights of patients within their data store. MEDHOST plans to build ML-based visualizations on this enhanced medical data for more accurate insights and provide clinicians with tools for predicting patient outcomes.

Getting value from AWS

One of the key aspects immediately valuable to MEDHOST is the serverless nature of their architecture. FHIR Works on AWS, using Amazon API Gateway, Lambda, and DynamoDB, allows you to scale as needed while only paying for the compute used. As an open-source project, FHIR Works on AWS also gives MEDHOST the option to customize the actual workings of the project to meet their needs. The most critical functionality that MEDHOST needed in place for their requirements was multi-tenancy, allowing their API layer to support multiple customers from a central deployment.

Instead of building their entire FHIR server from scratch, the MEDHOST team was able to use FHIR Works on AWS and only build the extra functionality needed. Similarly, Amazon HealthLake, being fully managed and built to scale on demand, also automates running medical comprehension on unstructured data without MEDHOST having to worry about building and maintaining complex services, servers, and data transformation pipelines. The service makes all insights immediately searchable, so MEDHOST can query their clinical data, as well as the derived insights, without having to worry about complex underlying infrastructure. The architecture, and the AWS services they use with their EHR data, allows MEDHOST to focus resources on their core business, provide additional capabilities for their healthcare customers, and allows them to meet their 21st Century Cures Act requirements.

Plans for the future

MEDHOST used its experience with AWS to kick-start and integrate new AWS features with its existing platform. This has given MEDHOST a huge head start toward meeting their compliance requirements and opened other options that might not have been possible before. With Amazon HealthLake and FHIR Works on AWS, MEDHOST can serve customers by creating a compliant FHIR data store in days with integrated natural language processing and analytics to improve hospital operational efficiency and provide better patient care. For healthcare customers looking to provide better access to clinical data, with integrated AI and ML, and a need to innovate faster, AWS encourages you to start exploring Amazon HealthLake and FHIR Works on AWS today.


About the Authors

Daniel Ness is a Solutions Architect supporting several AWS healthcare customers in the Southeast US. He specializes in serverless and blockchain areas, and loves to help customers build on AWS.

Arun Regunathan is a Data Lab Architect at AWS. He works with customers to understand their use case, architect a solution and build a working prototype in four days in the AWS Data Lab. He is passionate about big data and enjoys working with customers on their data and analytics use cases.

Brian Warwick is a Solutions Architect supporting global AWS Partners who build healthcare solutions on AWS. Brian is passionate about helping customers leverage the latest in technology in order to transform the healthcare industry.

Pandian Velayutham is the Senior Director of Engineering at MEDHOST.

AI/ML

Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: