A global medical device manufacturer specializing in patient monitoring equipment for hospitals and medical centers replaced an unsustainable manual complaint review process with a validated, AI-based decision model. The result: an eliminated backlog, a sustained "no missed reporting" standard for adverse events, and a 75% reduction in manual processing resources.
Challenge
The manufacturer needed help managing customer complaints and regulatory reporting as throughput and time sensitivity increased. The acquisition of a competitor's product portfolio, combined with a regulatory audit, exacerbated the situation by necessitating a review of service records for potential reportable events.
The existing manual process for evaluating customer complaints used a decision-tree-based methodology that was accurate but time-consuming and heavily dependent on trained resources. The increased volume of complaints and service records from new product lines and third-party sources made the manual process unsustainable. Scaling up the manual resource pool was not feasible because of the time required to recruit, onboard, and train. With a regulatory audit in play, the stakes for data integrity and complete, timely adverse-event reporting were high.
Approach
USDM developed an AI-based solution to mirror the existing manual decision-tree process and handle the volume of complaints and service records. Because the model automated a regulated quality process, sound AI governance and compliance and ongoing validation were built in from the start. Development included:
- Consolidating Data Sources: Integrating complaint records from legacy systems, service management sources, and third-party partners into a consolidated database for processing.
- Sequential Record Processing: Managed manually to adhere to regional and regulatory requirements.
- Creating the AI Complaint Evaluation Model: Using a three-tier approach for evaluating, filtering, and categorizing data source records, with subsequent processing and distribution, the model:
- Identified potential and non-potential complaints
- Assessed potential complaints, as per regulatory standards, for patient impact
- Evaluated reportability based on severity and regional regulations
- Integrating with their Quality Management System: Incorporating the AI model's outputs with the Complaint Management modules of TrackWise.
- Handling AI Model Outliers: Using a sub-process to identify and transfer records not processed by the AI model to regional centers for manual evaluation.
- Validating the AI-based Process: Performing ongoing checks for compatibility and accuracy, applying a risk-based computer software assurance (CSA) mindset to keep the automated process audit-ready.
- Self-Monitoring AI Model: Employing real-time performance monitoring and machine learning algorithms to maintain accuracy and adapt to new data sources.
Results and Benefits
The implementation of the AI model led to significant improvements, including:
- Eliminating the backlog of potential complaints and service records
- Adhering to “no missed reporting” goals for adverse event reporting
- Reducing manual resources by 75%; future goal is 40% of original manual resources count
The success of this project highlights the importance of a comprehensive approach to AI integration that involves data consolidation, process adaptation, and continuous validation and monitoring. It's a prime example of how AI can be strategically implemented to address specific operational challenges while maintaining compliance with industry regulations.
This case study demonstrates the efficacy of AI in transforming manual, resource-intensive processes into efficient, automated systems in the life sciences sector. The integration of AI in complaint processing streamlined operations and ensured compliance with regulatory standards, which demonstrates the potential for AI to enhance operational efficiency and regulatory adherence in medical device manufacturing.