Past Research
The MDIAS team conducted two studies during the Proof of Concept (POC) Phase 1 using FDA public data sources in 2021. The purpose of the POC was to explore how the MDIAS partnership model could operate and whether it could advance device quality and patient safety. The two studies conducted, laid the groundwork for establishing trust among participating organizations and showed MDIAS’ ability to enable collaborative decision-making. The final outputs provided valuable insights and a new view into these areas.
These studies have also served as building blocks for MDIAS capabilities. They have provided the environments for participating organizations to deepen their trust in each other and demonstrated the community’s ability to engage in collaborative decision-making. The lessons learned from these studies were leveraged as MDIAS embarks upon Phase 2 studies that incorporates non-public data from MDIAS Partners.
Study 1: How are Primary Root Causes of Recalls Distributed?
Description
The Root Causes of Recalls study helped MDIAS to understand certain trends and outliers in recalls. MDIAS used data visualizations to explore, categorize, and describe trends in root causes of recalls from 2010 to 2019.
This time period was selected for analysis to holistically assess recall trends since FDA’s 2011 report on Understanding Barriers to Medical Devices, and to account for limitations of the public datasets. 
The Study Working Group executed the study from November 2020 to May 2021 in collaboration with the Trusted Third Party. The outputs were shown using the tool, Tableau, which enabled the MDIAS Partners to see the dimensions of analysis in a dashboard format. The dashboard also included additional filters to further explore the data. 
Goal of Study
The goal of this study was to take a deeper dive into recalls and understand the distribution of root causes of recalls.Â
Output of Study
- Visualizations that display the distribution of recalls in terms of seven dimensions (medical specialty, recall class, etc.).
- Benchmarks that allow individual early adopters to view the recall distribution for their devices compared to the industry as a whole.
Description
MDIAS’ second study was evaluating if MDRs are predictive of recalls, and to what extent. MDIAS applied data science to 1) identify factors most associated with future recalls and 2) predict devices with a high probability for recall.
The factors associated with higher probability of future device recall include past recall metrics, past MDR metrics, and device characteristics. Using the 2010-2020 time period, data was aggregated into quarters, and then the model used an eight-quarter window to predict recalls in the following year.
The Use Case Working Group applied data modeling capabilities and visualizations to assess early indicators and features to correlate with historical recalls, and with models that could work using public data.Â
Goal of Study
The goal of this study was to build a capability that identified groups of medical devices that may need special attention based on their adverse event history.Â
Output of Study
Model that can predict recalls with significant accuracy for devices with recent MDRs. The visual dashboard showing device level probabilities of recall and important factors identified by the model that indicate a potential recall identified by the model. The model results can predict 75% of recalls and 58% of non-recalls for devices. The findings show that MDRs can help predict recalls; past recalls and device characteristics are also predictive of recalls. The model could be matured with the incorporation of additional datasets.
Any stakeholder that is interested in participating in studies can share their expertise, time, or any other information that could contribute to a specific study or general MDIAS operations. It is not required to be a MDIAS Data Partner to participate in a MDIAS study. The level of participation across MDIAS study participant varies.
All participants receive access to insights from the results of the studies. Participants that do provide nonpublic data additionally receive access to more detailed findings.
To find out more about how you can get involved with MDIAS research, fill out the Contact Us form below!
Contact Us