In the early 2000s, the Air Force struggled with a problem: Pilots and civilians were dying because of unusual soil and dirt conditions in Afghanistan. The soil was getting into the rotors of the Sikorsky UH-60 helicopters and obscuring the view of its pilots—what the military calls a “brownout.” According to the Air Force’s senior design scientist, the manager tasked with solving the problem didn’t know where to turn quickly to get help. As it turns out, the man practically sitting across from him had nine years of experience flying these Black Hawk helicopters in the field, but the manager had no way of knowing that. Civil service titles such as director and assistant director reveal little about skills or experience.
In the fall of 2008, the Air Force sought to fill in these kinds of knowledge gaps. The Air Force Research Laboratory unveiled Aristotle, a searchable internal directory that integrated people’s credentials and experience from existing personnel systems, public databases, and users themselves, thus making it easy to discover quickly who knew and had done what. Near-term budgetary constraints killed Aristotle in 2013, but the project underscored a glaring need in the bureaucracy.
Aristotle was an attempt to solve a challenge faced by every agency and organization: quickly locating expertise to solve a problem. Prior to Aristotle, the DOD had no coordinated mechanism for identifying expertise across 200,000 of its employees. Dr. Alok Das, the senior scientist for design innovation tasked with implementing the system, explained, “We don’t know what we know.”
This is a common situation. The government currently has no systematic way of getting help from all those with relevant expertise, experience, and passion. For every success on Challenge.gov—the federal government’s platform where agencies post open calls to solve problems for a prize—there are a dozen open-call projects that never get seen by those who might have the insight or experience to help. This kind of crowdsourcing is still too ad hoc, infrequent, and unpredictable—in short, too unreliable—for the purposes of policy-making.
Which is why technologies like Aristotle are so exciting. Smart, searchable expert networks offer the potential to lower the costs and speed up the process of finding relevant expertise. Aristotle never reached this stage, but an ideal expert network is a directory capable of including not just experts within the government, but also outside citizens with specialized knowledge. This leads to a dual benefit: accelerating the path to innovative and effective solutions to hard problems while at the same time fostering greater citizen engagement.
Could such an expert-network platform revitalize the regulatory-review process? We might find out soon enough, thanks to the Food and Drug Administration. The FDA is working on a pilot project called FDA Profiles that aims to revolutionize the way regulators perform their duties by enabling them to find and target people with the right expertise to advise them. And a survey of how the FDA does its work now suggests that reform is sorely needed.
Among its responsibilities, the FDA is tasked with reviewing the efficacy and safety of new medical devices, which include a range of products, from pregnancy tests to heart stents, that reduce suffering, extend lives, treat diseases, and generate enormous economic gain for successful inventors. The Center for Devices and Radiological Health (CDRH), a division of the FDA, manages this complex process of pre-market approval and post-market review of all medical devices. Within CDRH, still another agency, called the Office of Science and Engineering Laboratories (OSEL), employs its own scientists and consults with outside experts whose job it is to understand the safety and effectiveness of medical devices from invention to eventual use.
The pathway to regulatory review and compliance for low-risk items like tongue depressors is straightforward. But life-sustaining or high-risk devices such as pacemakers and breast implants require judicious and timely pre-market approval by those with the right know-how. If these medical devices reach patients prior to being properly tested, there may well be very real human costs. Dr. Steven Nissen of the Cleveland Clinic estimates that faulty devices contributed to more than 2,800 deaths in the year 2006 alone. In a CBS News report on medical implants, Dr. Nissen affirmed, “People make the assumption that when their doctor implants a device, whether it be an artificial joint or a pacemaker, that it’s undergone very rigorous testing. That assumption isn’t always true.”
In the current model of pre-market review, several challenges exist. For one thing, the process can take too long. At present, it can take nine months just to find and convene a qualified review panel. There’s a reason for this. Only 100 scientists work at OSEL—800 total at CDRH—and there’s always a need for more expertise. Staff at CDRH also have to oversee the safety of devices throughout their life cycle, creating further burdens. This partly explains why, from 2000 to 2010, the average time to decision—the time from receipt of an application to a determination—on high-risk but potentially life-saving devices increased by nearly 60 percent, from 96 to 153 days. Over approximately the same time period, the number of submissions for pre-market approval remained the same, hovering just below 10,000 per year. Simply put, the FDA needs to get faster at its job, but doesn’t have the budget to bolster its ranks. Nor will it in the near future.
Another problem is that the people who are already in the agency might not be the experts the agency needs most. In January 2011, the FDA committed to a plan known as the Innovation Pathway to transform its regulatory process and improve how it works with entrepreneurs, both to bring new devices to the market and to protect the public. The Innovation Pathway pinpointed this expertise deficit as a key problem, citing the need for greater collaboration with outside experts. There are many reasons for the lack of expertise: high reviewer and manager turnover at CDRH (almost double that of FDA’s drug and biologics centers), insufficient training for staff, extremely high ratios of frontline supervisors to employees, and a rapidly growing workload. The Innovation Pathway undertook a pilot project to create a vetted list of experts across two dozen membership organizations, such as the American Academy of Neurology and the Society of Thoracic Surgeons. In the absence of any means to “match” people to problems with any specificity, it relies on membership in these professional associations to serve as a proxy for expertise. But while membership in the Society of Thoracic Surgeons might require that one practice as a thoracic surgeon, it doesn’t tell the agency much beyond that. Consequently, the list is fairly limited in its ability to pinpoint expertise on the basis of experience or interest. Moreover, the vetted list has done a poor job of reaching those who possess a great deal of know-how but don’t belong to one these professional associations: A doctoral candidate at a university whose dissertation is on a specific medical device, for example, would not be captured by its net.
A third issue in regulatory review is diversity. The nature of the process demands access to a diversity of expertise. There is a wide variety of devices up for review (a 3-D printed exoskeleton is very different from a heart-rate monitoring app), and they are often cutting-edge, necessitating new kinds of expertise from many different fields, such as materials science, electrical engineering, and fluid mechanics. Indeed, they are sometimes hard to describe and are often called by different names in different fields, which makes it even more challenging to get help. In addition, there is always the need for insight from people in fields where one wouldn’t assume any prior knowledge. For example, it may be an entomologist who studies the architecture of anthills in the African savanna who might know the most about green building design. It might be the poet or the philosopher who has something to share.
Finally, the review process faces a sophistication problem. A study completed by the Boston Consulting Group points to a significant rise in the number of drugs and complex devices approved in the EU long before the United States. As devices get more complex, the United States is likely to fall further behind. Although the FDA does not issue clear data on backlogs or processing times, research points to shortcomings in our current system when it comes to researching and approving complex devices. This has broad implications for the quality of health care that Americans receive. In 2010, for example, the medical device company Biosensors International shut down its operations in California due to the time and expense associated with getting FDA approval for a cardiac stent. That device is available globally, including in Mexico and Canada.
New expert-network platforms offer the promise of overcoming these problems and decreasing the cost of finding expertise. This is especially true in highly technical fields with a critical mass of academics and regulators, such as biotechnology and software technology. Academics tend to maintain accurate and up-to-date data about their credentials and publications. Regulators are far less transparent about their credentials but more so about their experience, at least internally—there is a record of which regulator worked on which regulatory action. In short, these two groups—academics and regulators—already collect data that can help in creating expert networks and automating requests to participate in the review process.
There is no shortage of knowledgeable people in the biomedical sciences. In addition to the 800 scientists at CDRH, approximately 14,000 people work at the wider FDA, while another 6,000 scientists are part of a broader population of 17,000 at NIH, and more than 83,000 at their parent agency, Health and Human Services—not to mention the hundreds of thousands of biomedical professionals working in industry and academia. But the FDA has no way to quickly discover which of these individuals possess relevant expertise to review device applications. In order to enhance its research and regulatory activities and reduce the time needed for review, the FDA needs to be able to find experts to participate in review panels, as well as to educate the agency about new trends in medical devices.
Enter FDA Profiles. Funded by the NIH, Profiles is an open-source expert discovery and networking tool that aids in finding researchers with specific areas of expertise. It imports “white pages” information, publications, and other data sources and assembles a searchable library of electronic CVs. Although unique in government, Profiles is not revolutionary—it doesn’t upset the apple cart of power and change who makes the ultimate decisions at the FDA about devices. Its goal is to provide regulators with better and more relevant information more quickly. Profiles uses a platform developed by Harvard Medical School with support from the NIH that is in use at the biomedical faculties of 240 institutions, including Harvard, Penn State, Boston University, and University of California, San Francisco.
Slated to launch later this year, the FDA Profiles pilot project will create an online, searchable expert directory, initially of FDA staff in the first phase and later of outside scientists. Its expertise algorithms are designed to add publications automatically to a person’s Profiles bio from online sources and then to analyze these data to build a catalog of people’s expertise and their research network. Networks are automatically created based on current or past co-authorship history, organizational relationships, and geographic proximity. The system can also pull data from online biographies, professional society memberships, education and training records, affiliations, regulatory accomplishments, publications drawn from open databases, and grants culled from Grants.gov to auto-populate a profile. In addition, a researcher can manually add data about his or her interests, skills, and projects.
With both passively gleaned and actively contributed data, Profiles is intended to speed the process of locating the “right” experts by algorithmically matching people to opportunities to participate in the review process. For instance: Type in “stent” and FDA Profiles will list the relevant people at the FDA, with explanations for why they were suggested for inclusion in a panel. Although Profiles will initially function as a kind of “FDA LinkedIn,” the plan is to connect with other networks to find expertise outside the agency. In the future, there are ambitious plans to incorporate patents as well as data from LinkedIn, SlideShare, and other social media.
The FDA Profiles project will allow regulators to gather empirical evidence about whether targeting expertise helps the agency overcome the challenges in the regulatory-review process. The FDA needs expertise at different stages of decision-making, from identifying emerging trends to reviewing a specific device. FDA Profiles affords a chance to study whether expert networking and targeting participation by those with specific kinds of expertise work better at different stages of regulatory review. What we stand to learn from Profiles is whether it leads to faster, more comprehensive, and more effective review practices that get safe devices to the market faster and, ultimately, improve people’s lives by spotting problems sooner. FDA Profiles can also tell us the impact of expert networking on crowdsourcing more generally, and whether targeting participation helps to improve the process of working collaboratively across an agency or department and with the broader public. But drawing conclusions will require adding testing into the process, including randomized controlled trials, such as comparing what happens when the agency issues a general call to participate in a regulatory-review panel versus populating such a panel using Profiles. Doing so will advance our understanding of how to introduce approaches to policy-making that are both empirically validated yet agile in design.
The role of the regulator is fraught. Regulatory agencies are frequently under siege from industry groups that bemoan burdensome and costly requirements. At the same time, consumer groups assault them for failing to prioritize the public interest and protect citizens from dangerous and fraudulent products and services. Critics on both left and right challenge regulators’ ability to impose requirements and complain about their inefficiency and ineffectiveness. In the context of device review, grant review, and other domains where expertise is clearly called for, using expert-network platforms can only democratize what are now comparatively closed processes that typically rely on the same people to participate. Although the data from the United States are anecdotal, data from Europe show that the same repeat players serve on multiple scientific advisory committees again and again.
New technology has opened up possibilities for changing how we make decisions that could improve both the effectiveness and legitimacy of regulation. FDA Profiles could change the default on how the agency makes decisions by bringing in diverse expertise at the outset, and has the potential to do more once it begins to get used outside the FDA. The hope is that by finding the right experts and matching them to opportunities to participate, the agency can apply their knowledge toward better and faster decision-making. And the FDA’s willingness to experiment with projects like FDA Profiles potentially heralds a sea change in how government agencies use empirical research to drive institutional innovation. When social-science research is integrated with the practice of policy-making, we can gain insight into our decision-making apparatus and apply what we learn to the broader goal of improving governance.
As President Obama said in his 2009 Memorandum on Transparency and Open Government, “[K]nowledge is widely dispersed in society, and public officials benefit from having access to that dispersed knowledge” and, hence, to collective expertise and wisdom. There is no agency or topic that would not benefit from the ability to match people with relevant expertise to the opportunity to participate in the regulatory-review process. It’s long overdue to start using the tools available to us to inform how agencies make such important decisions.