AI Platform for Personalized Treatment for a Major Medical Device Manufacturer

Explore how Reveal built an AI platform that could formulate personalized treatment at scale by analyzing patient data of over 20 million users.

Overview

In this case study, a global leader in medical devices needed a way to process data from over 20 million patients. They needed a platform that could process the data at scale and leverage AI to provide patient-specific­ personalized treatment.  Reveal Health Tech helped by building a platform that harnessed AI to predict patient compliance, offering real-time insights and guidance for both patients and providers.

Opportunity | Key Issues with Identifying Target Patients

The primary opportunity in this case study was to identify which patients need the most guidance and who would benefit most from interventions.  The ability to make these predictions would enable providers to make choices around which patients to prioritize.  Technical issues included:

• The need for compliant, scalable infrastructure to allow global teams to rapidly prototype and deploy models.
• Insufficient infrastructure to manage large-scale, complex patient data and support parallel ML model development.
• Developing models that could adapt to multiple approaches for improving patient outcomes, especially for device compliance.

Approach | The AWS-driven Solution Blueprint

Reveal Health Tech began by recognizing the need for an enterprise-wide AI platform capable of processing vast amounts of data. This platform integrated billions of data points from medical devices and other data sources. Engineered using a state-of-the-art, modern infrastructure, the platform ensured stringent data privacy protocols were followed while also ensuring compliance with HIPAA controls. The platform supported developing, deploying, and serving multiple ML models in parallel.

Reveal trained an ML model to predict patients who were most likely to deviate from the prescribed treatment plan. The model took the prediction one step further and even suggested which patients were more likely to respond with interventions like coaching, reminders, etc.  We then developed two critical applications that made use of the new models: one offering patients personalized feedback to enhance device usage, and the other equipping providers with actionable insights to ensure patients remain compliant with their prescribed therapies. Providers were able to trigger emails and calls to notify the patient. Finally, we introduced a cutting-edge, RAG-based virtual assistant, allowing users to ask questions and receive real-time support in managing their medical devices.

Reveal HealthTech leveraged advanced AWS-based infrastructure to build a robust, scalable, cloud-based platform. Key components of the solution included:

• Data Platform: Utilized AWS Glue for extracting, transforming, and loading (ETL) processes and Amazon S3 for secure and scalable data storage, enabling seamless data integration from various sources.
• Data Feature Engineering Pipeline: AWS Lambda and AWS Glue facilitated real-time data transformations, while S3 served as storage for intermediate datasets. Athena enabled efficient querying, and Airflow orchestrated the entire process. EMR was leveraged for distributed data processing to enhance feature extraction and engineering.
• Machine Learning Ops Pipeline: Bitbucket was used for version control, while SageMaker Notebooks handled model development. The models were stored in the SageMaker model repository and containerized using Amazon ECR and SageMaker containers. AWS CodeBuild was employed to run QA tests and deploy models into production environments.
• Operationalized Inference: CloudWatch monitored the health of the inference endpoint, Amazon API Gateway managed API traffic, and AWS PrivateLink ensured secure communication between services without exposing data to the internet.
• Model Monitoring Pipeline: Sagemaker Model Monitor tracked model performance and drift. The Sagemaker A2I workflow enabled human review for re-labeling, while EventBridge and CloudWatch coordinated event handling and alerting on model performance metrics.

Features| Adherence Made Easy

• Simplified ML Product Development : A globally-available platform enabled Data Scientists to deploy models in weeks rather than months.
• Increased Adherence : Clients in the Beta program had improved rates of compliance. Care providers were able to engage with more patients during the day due to improved workflows and targeted patient lists.
• Virtual Assistant: Client released the first patient-facing chat assistant leveraging RAG. The automated risk evaluation system significantly shortened the review cycle from weeks to days at less than $0.01 per input output pair.

Outcomes | Impact and Improvements

• With the introduction of the ML platform created by Reveal, our client’s data scientists were able to deploy machine learning models in weeks instead of months.
• The platform enabled internal functional teams to use it across multiple customer-facing products, increasing its impact across the organization.
• Due to its success, the platform was highlighted by the CEO during a quarterly earnings call and was featured in a case study published by Amazon Web Services.
• The predictive ML model contributed to a significant improvement in patient treatment adherence, resulting in better prognoses for patients.
• Improved adherence rates also helped address reimbursement challenges faced by providers.
• The platform empowered frontline providers by helping them identify patients most likely to respond to interventions, enhancing the overall effectiveness of care.
• Globally, the platform allowed data scientists to deploy personalized care to over 18.5 million patients, significantly boosting patient adherence rates.

Conclusion

Reveal Health Tech’s innovative ML platform empowered our client to enhance patient outcomes and streamline provider interventions through scalable, AI-driven solutions. As a result, the client was able to deliver personalized care to millions of patients while driving organizational success and recognition at the executive level.