10 Mar Root Cause Analysis for Medical Devices
Debugging intermittent failures in any instrument is difficult. However, the task becomes even more challenging when the instrument is a medical device that interfaces directly with a patient’s body. Every patient’s anatomy is different, so how do you identify what parameters have the greatest impact on device performance? When testing cannot be performed on actual patients, how do you create a representative test fixture that mimics human anatomy?
I recently managed a project where we were faced with, and overcame, these challenges. Read on to find out how we performed a Root Cause Analysis for a difficult-to-debug medical device.
Assemble a cross-disciplinary team
At Key Tech, we have a talented staff of Industrial Designers, Human Factors experts, Mechanical Engineers, Computer Engineers, Electrical Engineers, and Engineering Technicians. All of these disciplines provide critical expertise in determining the root cause of medical device failures. Device issues could be due to a range of causes, including user errors, hardware malfunctions, assembly defects, or firmware bugs. To quickly identify the root cause of an issue, multiple potential causes must be explored in parallel. For example, a Mechanical Engineer creates a test fixture to investigate potential hardware flaws while a Computer Engineer examines firmware, software, and configuration files for issues.
Leveraging our decades of technical experience, we commonly perform a thorough engineering review of our client’s designs. This is critical to developing a deep system understanding and identifying potential root causes. Often, this involves a detailed review of operating manuals, QA and assembly procedures, material and component selection, and tolerance stack-ups.
We supplement our expertise by seeking insights from external sources. Our clients’ regulatory experts, manufacturing teams, and sales departments all have valuable input, whether to help understand the issue or to identify what potential solution spaces would ultimately be feasible to implement.
Gather and analyze as much clinical data as possible
To develop a full understanding of the issue, collect and analyze as much clinical data as possible. On my recent project, we obtained log files from both successful and failing device startup tests. Using both sets of data allowed us to look for key differences between successful and failing devices, which was ultimately critical to reproducing the device issue.
When log files are large or unwieldy, we will develop custom python scripts to quickly plot and analyze the data, enabling visualization of trends and identification of attributes that should be replicated with in-house testing.
Sales teams, who often have the most direct interaction with clinicians and users, are another potential source of insight into understanding device performance. Are certain clinicians experiencing failures more frequently than others? Do particular devices have high failure rates, while other devices rarely fail? While this information is often qualitative, it is crucial in building a deeper understanding of the issue.
Test easy-to-replicate bounding conditions or extremes
Once the failure is well-defined and clinical data is reviewed, we brainstorm a wide variety of potential root causes. Ultimately, the goal is to develop confidence in which root cause(s) are contributing to failures by replicating the issue in controlled tests, using fixtures to simulate real-world operating conditions.
However, prior to developing a complex anatomical model, it is often helpful to test easy-to-replicate bounding conditions and extremes. For example, if a potential root cause is a saline delivery pathway in the device being intermittently clogged by tissue, running tests where that pathway is permanently clogged with adhesive can indicate how device performance changes with a clog without needing to build and tune a soft tissue model to induce clogging. When data collected with this exploratory testing is compared with clinical data, it provides clues as to whether that particular variable is contributing to in-field device failures.
Create anatomical models
Anatomical models allow for the simulation and tuning of real-world operating conditions and they are critical in reproducing device failures. Readily available, off-the-shelf medical simulators may be sufficient or at least useful for initial testing, but they often do not provide the level of flexibility and specificity required for a Root Cause Analysis involving complex human anatomy.
On my recent project, we tested a handful of off-the-shelf simulators but ultimately found that developing a custom anatomical model was necessary. In our case, we had the most success constructing anatomical models with medical gelatins. We molded the gels in-house, rapidly iterating on anatomical dimensions and gel durometer to represent the variability from patient to patient. To select representative model parameters, we referenced academic papers and industry standards. We embedded sensors in a subset of our models, expanding the quantitative data gathered from each test.
With each model variation that we tested, we compared the log data from the lab testing to the previously obtained clinical data. Through trial and error, we identified several key variables that influenced whether a device test would pass or fail and tuned those parameters to closely mimic real-world data.
Outcomes
With the approach outlined above, we successfully identified the root cause of the in-field failures on my recent project. Based on our newly developed, detailed understanding of the system, we also uncovered and proposed solutions for several other potential failure modes that were originally outside of the scope of our Root Cause Analysis. Finally, the anatomical model that we developed became a critical tool for our clients for any device testing needs going forward.
Do you need help determining the cause of failures in your medical device? We would love to hear from you. Reach out to us at TalkToUs@keytechinc.com!
- Root Cause Analysis for Medical Devices - March 10, 2025
- Differentiating in an Already Crowded Market - July 24, 2024
- Rapid Prototyping for Speed to Data - January 10, 2024