Enhancing Patient Safety Intelligence with AWS AI/ML Services

In the healthcare industry, patient safety is of utmost importance, and healthcare organizations constantly strive to improve safety protocols. AWS (Amazon Web Services) provides advanced AI/ML (Artificial Intelligence/Machine Learning) services that can revolutionize patient safety intelligence and streamline workflows for healthcare organizations.

Currently, patient safety reports in healthcare organizations rely on a combination of manual and automated processes. These reports, entered into the RL Datix reporting system, contain both discrete data points and free-text narratives. However, the data within these reports remains largely inaccessible for real-time analysis and trending throughout the organization. Patterns and trends are often limited to specific units or service lines, making it challenging to identify organization-wide issues.

To address these challenges, AWS offers solutions that leverage AI and ML capabilities. A Proof of Concept (POC) collaboration with University of Utah Health focused on the automated analysis of medication-related patient safety reports. The solution aimed to reduce manual work, improve insights from reports, and uncover patterns across the organization.

Key components of the solution include:

1. Amazon Comprehend Medical: This service was used to detect high-risk medications mentioned in the reports. The results were summarized in an interactive dashboard built on Amazon QuickSight.
2. Automated Data Processing: The entire data processing pipeline was automated using an event-driven, serverless architecture powered by AWS Lambda. This ensures efficient processing and analysis of patient safety reports.
3. HIPAA Compliance: Since patient safety reports contain sensitive information, all services used in this solution are HIPAA eligible. The project was carried out in a HIPAA-compliant landing zone account to ensure data security and compliance.

To enhance the efficiency of patient safety reporting, different transformer-based LLMs (Language Models) were refined and compared using AWS partner Huggingface. These models effectively detect and classify high-risk medications based on the free-text descriptions in the reports.

The solution architecture ensures secure and compliant ML environment. It includes features such as data encryption at rest using AWS Key Management Service (AWS KMS), network isolation, and role-based access control using Amazon SageMaker and Amazon Identity and Access Management (IAM).

The outcomes of the POC project demonstrated the effectiveness of the AI approach, achieving high precision, recall, accuracy, and F1 scores. Building upon this success, the plan is to engage with AWS partners to develop other use case applications and test a production-ready system that includes complete clinical data. This will enable the integration of electronic health record (EHR) information into the analysis and further improve patient safety intelligence.

By leveraging AWS AI/ML services, healthcare organizations can enhance patient safety, reduce time-to-insight, and uncover valuable insights from patient safety reports. The combination of advanced analytics and AI-driven analysis empowers healthcare professionals to make data-driven decisions and drive continuous improvements in patient care.

Source: AWS