By: Randall C. Iliff

During an economic depression it is vital for market innovators to prepare for the certainty of eventual recovery. Of course, it is difficult to make the necessary investments when companies are slashing budgets and reducing their staff. However, doing nothing to prepare for the recovery isn’t a viable option. The escalating medical expectations of an aging population and the push to increase healthcare coverage ensure that there will be an elevated demand for medical devices. This demand is accompanied by the political imperative to control medical expenditures, which means that manufacturers need a medical device design process that operates at an unprecedented efficiency level. This requires attention to cost control but—more importantly—it requires solid device designs that minimize avoidable problems.

Product design employs an iterative
design-and-test approach to deliver safe and effective medical devices. However, effort spent correcting self-inflicted problems, such as incomplete and inconsistent user requirements, electromagnetic interference issues, etc., is the equivalent of manufacturing rework and carries no inherent market value. Removing even small amounts of avoidable effort greatly improves the overall efficiency of the entire design process.

Applying Design Logic to Design Processes

At its most fundamental level, the design process can be thought of as a machine. It starts with an objective, refines that objective sufficiently to begin associating possible solution elements, and iteratively solves for an optimized balance of objective and solution. The process then refines the solution to the point where it can be efficiently reproduced in whatever quantity the market desires. Medical device companies are skilled in the art of designing medical products that operate as intended, so it seems reasonable that some of this skill can also be used to create better medical product design processes. Simply focusing existing skills on the task of designing process instead of product can make a profound difference in overall market effectiveness, yet costs virtually nothing to accomplish.

When developing design processes, common sense is key. For example, if single-point failure opportunities (such as those created by eliminating hardware interlocks and depending on software alone to control x-ray emissions) are unacceptable in critical devices, they should be equally unacceptable in critical processes. If a simple user interface is essential for reliably communicating device status, then it seems reasonable to insist that process metrics also be kept simple. Just like a production run would not be authorized without an approved product baseline, significant design expenditures should not be permitted before market needs are understood.

The good news is that it’s easy to start using product design knowledge to optimize the design process. Simply watch for logically equivalent situations, examine why the underlying principle works well in product design, and then implement a process element that performs the same function.

First, Do No Harm

Before exploring how to develop a more-efficient design process, it’s important to acknowledge what makes design such a different thing to optimize than manufacturing. If medical device manufacturers don’t distinguish between these two fundamentally different tasks, they may apply principles that can’t help—and may very well harm—the essence of design.

The most fundamental distinction is that design is referenced against something that will exist in the future. In contrast, manufacturing is always referenced against something that was agreed upon in the past. It’s this past-referenced characteristic that gives variance reduction methods so much power in the manufacturing environment. If the OEM knows exactly what “done” looks like, anything that doesn’t take it on a straight line toward that goal is clearly wasted effort, and eliminating it represents an opportunity for efficiency gain.

Conversely, in medical device design, both the goal and the means to achieve it are variables. One way to address this
problem is to pretend that the goal is fixed and then apply manufacturing-derived processes. For example, the whole family of statistical process control methods can help companies identify and eliminate variance from the stated design goals. The problem with this example is that statistical process control, like virtually all manufacturing-derived methods, depends on a predefined target value against which to determine variance and a statistically significant number of samples to evaluate. Both are readily available during the manufacturing period, but often completely absent at the start of the design process.

The secret to selecting an effective design process is to understand how much “new” is involved. At one extreme are product designs that are only minor deviations from current art; at the other extreme are radically new product designs that have very little in common with the past. Another way to describe these extremes is to say that the design maturity in the first case is very high; whereas, in the second case, design maturity is initially very low.

Variance reduction methods can work very well in the special case in which design objectives are clearly defined at the start, consistently visible to all participants, and stable throughout the design period. A practical example might be moving a company’s existing product to an offshore location.

However, the design process is more frequently used to produce something that is different in some way from currently available products. If the difference is small, or confined to only one aspect, then manufacturing-derived methods can be a very useful element of the overall design process. This works well for progressive application of new technologies to constant functions, such as when a respirator user interface gradually evolves from CRT, to LCD, to touch panel.

Pure Design

The most challenging situation occurs when the design process has to create a fundamentally new solution. In this case, both the starting point and the objective are initially undefined, and a key early project goal must be to create working definitions of both. There is no logical extension of manufacturing-derived processes that can help at the very start of such projects. This is the domain of what is sometimes called pure design—the ability to create a radically new future based on possibility rather than projections from the past.

This is the category that applies to all true innovation and through which all medical devices must initially pass. As an example, today’s common pulse oximeter was once a radical innovation unlike any device or protocol that had preceded it. The design cycle began at the moment when indirect measurement of blood oxygen was first considered. However, the oximeter’s high level of design maturity today is the end result of many subsequent cycles of iterative refinement, each of which depended to some degree on the cycle before it.

The ideal design process is always the one that most efficiently raises the design maturity from the starting point to one where production can take place. This may occur in a single development stage but, more commonly, an iterative series of design stages are imposed, thus allowing the design process within each stage to be tuned to match the associated starting and ending design maturity characteristics.

Figure 1 conceptually illustrates how design maturity and design process choices are related. At very low design maturity, pure design methods are essential. At high design maturity, manufacturing-derived methods become usable. An evolving mix of both methods is needed when the maturity level is somewhere between high and low.

It is certainly more expensive, both in time and resources, to design a revolutionary product than to modify an existing one. On the other hand, the potential return of being first to market with a terrific product can be enormous. Just as balancing cost and performance is central to medical product design, getting the maximum return is crucial for each design process cycle. This depends on selecting the least-expensive process that is capable of reliably performing the required tasks.

The essential key to selecting the right approach is to understand the degree to which the design goal itself is a variable. The following questions can help quickly determine which class of process solution is really needed:

 

• Does the manufacturer have experience designing a product similar to this?
• Are all of the requirements known, and can a working range of values be assigned to each of these requirements?
• Are there dependent relationships between requirements? If so, are those relationships understood well enough to predict the composite behavior of the full system?

 

If the answer to all of these questions is yes, then by all means take advantage of the efficiency that manufacturing-derived methods can offer. If not, consider whether the potential value of the opportunity justifies use of more powerful—but also more costly—pure design methods. If pure design methods are already available within the company as a process option, the cost is simply one of execution, otherwise some allowance will need to be made for start-up time and resources.

A very common problem is pretending that the design needs are simpler than they really are in order to justify the cost and schedule estimates that accompany either a proposed sale or an internal project review decision. What often happens is that the revenue remains illusory, while the costs accumulate with each passing day. Although self-delusion may be an innate human characteristic, early attention to identifying the true design needs can greatly reduce overall project risk. A particularly effective way to discover design process needs, and compellingly communicate those needs to decision makers, is to look for situations that are analogous to those routinely encountered in actual product design.

Example 1: Interface Definition and Control

For any medical device, the most obvious external interface is with the patient and caregivers. An extensive regulatory system exists not only to control the device, but also to mandate the developmental processes and approvals needed before a device can enter the marketplace. An acceptable process is a necessary precondition to market entry—but an efficient and profitable implementation of that process is required for market survival. Failure to properly define and control system, subsystem, and device interfaces is a classic route to disaster in product design, and, not surprisingly, a key source of risk in design processes.

Interfaces are often defined simply as the boundary between system elements, but a more useful way to think of them is that interfaces are actually the set of things two or more system elements must have in common in order to function. This distinction may seem subtle, but all interacting devices, individuals, and organizations must be purposely equipped with some sort of common interface rule set.

Examine the design process flow and, instead of focusing on what takes place within the boxes, explore what happens at each of the interfaces. Imagine looking at a block diagram for a medical device that collects patient data, processes them in some form, and then presents the data to caregivers. Such a device would logically need a reliable communications channel that was available whenever needed. The device would also need at least basic acknowledgment and checksum features to confirm that error-free transmission took place. Failure to provide a reliable communications channel or to confirm reception of transmissions would introduce many problems and obviously be unacceptable in medical device design. In the best case, the patient would be at risk, and in the worst, a life-threatening condition might result.

The design process is equally dependent on reliable communications and should never be left uncontrolled. For example, if a manufacturer operates internationally, the challenge of channel availability must be overcome when dealing with multiple time zones and holiday schedules. The basic principle of checksum can be applied by confirming that requirement changes were received and understood, rather than just sending the changes and assuming they will be incorporated. Additionally, peak workload periods that occur prior to major trade shows or at the end of the year can all be addressed in the process using the same functional solutions that were employed in medical device design. Priority rule sets, device interrupt management, parallel processing, and even alarms become logical process tools in this situation.

There are many more levels of parallel insight that can be usefully explored. Channel bandwidth needs to be adequate for the type and quantity of information that must be exchanged. The link should be able to tolerate environmental noise levels, adapt when needed to bridge different data transfer rates, self-test at power up, and automatically reestablish itself when broken. If the medical device design process lacks equivalent functionality, the likely result is unpredictable project behavior and increased overall business risk. Even the principle of root cause analysis, whereby symptoms are traced back to the originating condition, is directly applicable to process optimization. Instead of tracing back to a specific device failure, the root cause analysis in this case identifies the process stage or hand-off where the “defect” occurred.

Example 2: Tolerance Stack/Design Margin

Despite the repeatability of modern manufacturing methods, the reality of medical device design is that component tolerances must always be understood in terms of the full assembly or system impact they produce. In fact, the trade-off between required tolerance level and part or system cost is a common design optimization activity.

Device manufacturers understand the cost of pursuing increasing levels of predictability and look to balance that cost against overall benefit. Sometimes the answer is to specify precision parts, sometimes to make the overall design tolerant of variability, and sometimes a little of both is the best choice. This logic is just as useful when the objective is optimizing process.

As a practical illustration of tolerance stack in medical design process, the impact of marginal errors in judging market needs, incomplete understanding of regulatory requirements, and testing under conditions that don’t genuinely reflect the user environment would result in devastating liability exposure. As is true with devices, individual process parts can be easy to identify but are often challenging to characterize as a system. In such cases, it often works better to examine the stack from the top down.

Any design process should be good at converging on a specific design target and should be able to reach a stable state in as few iterations as possible. If the company has difficulty stabilizing requirements, a possible reason is that the variance level present at early stages is too great for the subsequent process stages to accommodate. Early design must be deliberately tolerant of wide variance in order to avoid prematurely excluding promising ideas, but, in later stages, such variance multiplies the design workload and dilutes available resources. Decision gates between stages should therefore examine both the results to date and the uncertainty carried into subsequent process stages.

Example 3: The Boolean Algebra of Design

The familiar situation of googling is analogous to the process of medical device design. Try a Web search using the word “design” and more than a billion results will be identified. Try the search again using “design AND medical AND efficient AND approved AND profitable” and more than 99.9% of the results go away. Add a few more terms and the number of results may very well drop to zero.

The medical device design process seeks to solve a complex equation that consists of all the attributes that the final product must exhibit (or not exhibit). The equation is, of course, subject to the constraints associated with supplier and regulatory systems, entrenched company and user precedents, and the need to stay in business in a competitive market. Although solving the equation is a difficult task, setting it up correctly is even more critical.

The pressure to make progress on assigned tasks often leads the design team to jump past the setup stage and begin trying to solve the partially defined equation. It is extremely helpful to recognize that setting up the equation and solving it are both tasks that must be overtly managed. Just as a search can be useful or useless depending on the criteria, so follows the logic of the design process. Start too broadly and time is wasted sifting through items that aren’t remotely applicable; make the search too narrow and potentially valuable results will be missed.

Solving complex design problems can be frustrating at times, and it is only human to look for shortcuts. Does your process attempt to deal with the massive equation of design one variable at a time, or does it fully acknowledge the mutual dependencies involved? Strategies for iteratively reducing the number of remaining variables are an effective way to make progress toward design closure, whereas merely examining one variable after another can go on
forever.

Albert Einstein’s famous quote, “Things should be made as simple as possible, but no simpler,” contains practical guidance to help designers deal with complexity. Examine the variables involved and divide them into sets based on interactions. Any prediction of behavior, or attempt to explain symptoms, must take into account all of the interacting elements. This logic is absolutely common to both medical products and the processes used to design them.

For example, attempting to fix a nonconforming temperature condition without considering the power consumption, operating mode, environment, and packaging would be futile because all of these variables interact. Burdening the analysis with unrelated factors (such as the typeface used for panel markings) clearly offers no value. With process, the goal is the same: find the parts that contribute and exclude the rest.

It is always safest to examine the whole process to uncover process elements that interact in any given context. By doing this, one omits nothing and exposes the simplest system description that contains all of the variables—satisfying the “as simple as possible, but no simpler” test.

Conclusion

One of the best tools medical device manufacturers have available to determine what will or won’t work in the design process is comparing the underlying logic to what they know works in products. As was explored in the earlier examples, virtually every aspect of medical device design is based on a logical principle that has a functional equivalent in the design of process. Interfaces are always a source of risk, the cumulative effect of tolerances must always be understood, and interacting elements must always be addressed as a set. The richness of both product and process domains means that thousands of such parallels are waiting to be discovered.

This insight can help a medical device manufacturer determine the source of avoidable design problems, protect and nurture the process elements most critical to long-term success, and learn more effectively from past experience. In any market, these benefits would be worthwhile. In the difficult market companies face today, they may be essential to survival.