Introduction to Cybernetics and the Fundamentals of Service Design with Dr. Paul Pangaro Final Project

Challenge: As the final project in introduction to cybernetics and the fundamentals of service design we were asked to create a critique of some pre-existing system (i.e. a software interface, a service design situation, or a design process, such as user-centered design, personas/use- cases, user research, “design thinking”). We were asked to show how the models taught in the class explicate the strengths and weaknesses of the system, as well as how the system may be improved, and by what process.

Solution: I chose the administration of the stroke medicine Warfarin as my service design situation, and demonstrated how a behavioral economic intervention effectively increased adherence.

Stroke is the fourth leading cause of death in the United States. Only recently did the death rate from stroke decrease for the first time in 50 years, by 4%, according to a report from the Center for Disease Control and Prevention (CDC). Nevertheless, the 2010 estimated cost of stroke in the United States is $73.7 billion, which represents about 3% of total healthcare spending. Fortunately there are several effective medications that radically reduce the probability of stroke. For example, Warfarin (Coumadin) lowers the chance of a second stroke from 26% to 3% with very few side affects. Despite this proven efficacy, doctors struggle to make heart patients adhere to their prescriptions. A cybernetic approach to this problem reveals some of the underlying causes as well as innovative solutions offered by the emerging field of behavioral economics.

This scenario represents one in which the interests (or goals) of the individual are shared by those of another participant, in this case the health system at large, a system that represents doctors, the CDC, the NIH, as well as health insurance providers. The individual has the goal of preventing stroke in himself which serves as a means of preserving quality of life. Likewise, the health system shares the goal of preventing stroke in that individual for the different but compatible purpose of reducing the rate of stroke in the United States as a whole. The two participants have a conversation to agree on a system by which to achieve their goals, which is manifested in the exchange of the drug Warfarin. The doctor provides a prescription for Warfarin each month and the patient is expected to ingest one pill daily.

In theory this system should work perfectly. It is in the best interests of the individual to adhere to the plan as decided. In addition, doctors take pains to educate patients on the risks of not adhering to their medication. Furthermore, Warfarin is prescribed to patients who have already experienced a stroke in the past; therefore patients are fully aware of the detrimental affects of a stroke on quality of life.

Unfortunately, the behavior of patients is not always logical or predictable. It turns out that the rate of adherence to Warfarin is only 66%. Consequently, many patients are at higher risk of both under and over-coagulation of their blood, since Warfarin works as a blood thinner. In fact, for every 10% increase in missed doses, patients experience a 14% increase in the odds of under-coagulation. The question emerges, what is at the heart of this seemingly self-destructive behavior?

There are two major system problems. The first and the most important is in the nature of the drug. Since it is preventative the patient has no significant feedback that it is in fact working, except for the absence of a stroke. It is safe to say, then, that there is no “active” feedback that the patient can sense, and therefore the feedback loop, to a certain extent is severed. The second problem is that while the frequency of expected behavior on the part of the patient is daily, the frequency of the doctor’s sensing capacity is monthly, and even then, the doctor is limited in her ability to sense whether the patient has adhered to the treatment over the course of the month. These issues are unfortunately not uncommon for preventative medicine.

A pilot study conducted by a behavioral economist, George Lowenstein at Carnegie Mellon, reveals how this system can be improved so as to increase adherence. In the study the researchers provided a group of patients with a device, The Med-eMonitor™ that could recognize when the patient took their medication. The patients were also provided a two-digit number. Each day the researchers would draw a random two-digit number. If either the first or second digits of the patient’s number corresponded with the first or second digits of the picked number, respectively, that patient would receive $10. In the rare case that the patient had the exact same number as the one picked, the patient would receive $100. This reward was only conferred if the patient had taken the medication that day. What they found was that rates of adherence rose from 66% to 97%.

Returning to the system model reveals some of the mechanisms at work in achieving this behavioral change. By providing an incentive for taking medication, the study created a feedback loop in parallel with the existing goal-means mechanism in which such a loop was absent. Since the means were shared by the two goals, the additional goal caused the initial goal to be achieved. The new feedback loop was especially constructive because not only was feedback provided when the patient behaved as desired, but negative feedback was also provided when the patient failed to adhere to the prescription; the patient still saw the number drawn regardless of their behavior. Therefore they were aware of any lottery they would have won if they had adhered. This interaction capitalized on a documented psychological phenomenon called “loss aversion.” Another nuance of the feedback loop is in the nature of a lottery. Even though the probability of achieving the large payout of $100 was low, the mere existence of this chance caused greater behavior change because individuals are known to overweight small probabilities. In a model sense, the comparator in this system was dynamic. Rather than generating a consistent result from the patient’s behavior, the comparator generated a more enticing result based on chance. Lastly, the model also allowed for the researchers to sense the patient’s behavior with a frequency equal to their expectation (goal). Whereas previously, the doctor was only able to check-in on the patient on a monthly basis, using the Med-eMonitor™ the researchers were able to track the patient’s behavior on a daily basis.

This model is important on two levels. Firstly, and most obviously, increasing adherence to a medication that has proven benefits is of value in itself. More importantly, by modeling this system we are able to understand the underlying architecture of the goal structure. This enables us as designers to understand the problems with the system and the ways it may be improved which can then be applied to corollaries in other realms of human behavior. Such corollaries can be found in addiction control, weight loss, and spending versus saving.

back to work