Catching potential quality and supply chain issues early in the medical device development process can save organizations significant amounts of money. However, current protocols and systems can inhibit sharing of information between departments at the earliest stages of development. We spoke with Tim Brown, Vice President of Product Life Management (PLM) at ComplianceQuest, to discuss how new design development systems that incorporate AI and machine learning can enhance information sharing, improve product design and support quality and product lifecycle management.
How does the design development process support a company’s quality management system?
Brown: When you start looking at design and development activities and processes, they go hand in hand with quality system activities. They may be two separate departments within an organization, but they are trying to affect the same outcome, which is to consistently develop the best products. So, there is overlap between the inputs and outputs associated with development activities and quality system activities.
For example, on the quality side you have complaint codes. If you have a broken tail light, and that tail light is consistently breaking, you should have a complaint code for that. On the design side, they use that information to inform design elements that require additional testing and may even change design based on the complaints. They are using inputs and outputs from both the design and quality processes to make the product better.
What are some of the pain points in improving communication between departments?
Brown: The pain points that I have seen have been around providing visibility to information as quickly as possible. Design and development may know what they’re planning on doing, but other groups of people within the organization may not know that quite yet, and there often isn’t a mechanism to be able to share that information with supply chain, for example. Design and development may think they can build a product or component in house, but they typically don’t make that decision. Having procurement and supply chain involved early in the process can help the development process.
A big piece of the design and development processes include product lifecycle management (PLM) activities. Sometimes design and development see a PLM system as another mechanism of control. It’s another form I have to fill out and it creates more work when I could just be filling out a sheet of paper putting it in my spreadsheet. Engineers are starting to push back on developers of PLM systems to say that these systems must be more conducive to the way that they work and the way they interact with the products.
There is the quality viewpoint of a product: did this peel off or break? From the design viewpoint, I’m looking at creating some cool shapes as well as the ergonomics of the device. PLM systems need to interact with that design viewpoint versus just a quality viewpoint. And I think the trend in PLM system development is starting to move in that direction.
How are AI and machine learning impacting design development and PLM systems?
Brown: Some of the use cases that we’re starting to see, especially on the design side of things, are with the computer-aided design (CAD) models. For example, if a component of the project is going to be injection molded, you enter the process and the computer knows that there are certain requirements, such as certain angles or fillets that this process can do. Over time, if a product is going to be injection molded, the machine will learn to automatically put 2 ½ degree draw angles on the product. So, fit and function starts to take place earlier in the design side of the process.
As we get into the development side and the testing pieces of that, we start looking at detectability. If our processes are connected—and we start making connections between our processes and our designs—then the machine learning or AI capabilities can see correlations between this design and other types of products that are similar, as well as defect codes that may be aligned to that.
For example, if I’m designing a new computer mouse, but I am not considering quality issues that have occurred with other mice, AI can start to look at that and say, have you considered some of these types of activities? The human will always make the final decision, but AI and machine learning can make recommendations to help ensure that standards and potential defects are identified. That’s where AI has a place in design and development.
So, it can essentially raise red flags earlier in the design and development process?
Brown: Yes, if my output is going to cause you grief, it can bring your requirements into the design process to recognize potential problems earlier. For example, design may say, “There’s a brand new material we want to use,” and supply chain says, “Wait we have thousands of this related material right here, why don’t we use that?” For some organizations, if they have thousands of pieces of a certain material on hand, it’s probably for a reason. For instance, three years ago someone created a CAPA that said don’t use this material anymore, and no one rejected or scrapped out the material. I can take that information that was highlighted through machine learning in the design process and see that we’ve had a lot of complaints with a certain type of material.
How can these advances help support better patient outcomes and regulatory compliance for medtech companies?
Brown: It goes back to creating the best products we can. If we’re putting better products in the market, we’re going to have fewer complaints and fewer nonconformances. If we have the right controls and we can better identify the risks by identifying similar products that have similar risks, then we can start to design out some of those problems.
That typically has been done by human knowledge. If I want to develop catheters, I’m going to find an engineer who has been developing catheters for a long period of time. I’m not going to hire an engineer who has been making computer cases, because they’re not going to have the same history and knowledge points. As an organization, we can start to rely a little bit more on AI to flag some of those potential problems when we start collecting and seeing all that connected data.