The Road Not Traveled: We Need an On-Ramp for Innovation
We seem to be moving towards a national industrial policy. Part of that is an effort to promote innovation in health care. President Biden has created the Advanced Research Projects Agency — Health (ARPA-H), following the pattern of the Defense Advanced Research Projects Agency, to foster a renewed inventiveness in health care techniques and tools. The idea is for the federal government to fund high-risk/high-reward innovations.
It might work, but as a casualty of one such effort, I offer the following as a cautionary tale.
I was Executive Director of the National Collaborative for Bio-Preparedness (NCB-Prepared) from 2010 to 2013. We were a project that brought expertise and capabilities from the University of North Carolina at Chapel Hill, North Carolina State University, and SAS Institute. We were federally-funded to create new systems to detect intentional threats to human health in the US by using unconventional data sources and advanced analytics. Our budget was supported by a specific act of Congress in one of the last earmarks before Republicans suspended the practice while they had Congressional majorities.
I’m not going to name any individuals because that’s not the point of this essay. If they want to comment on this posting, they are free to do so. I’m more interested in the institutions and organizations involved.
We developed an approach based on the use of novel (at that time) sources of data and advanced analytics to detect an underlying health issue. It was not reportable disease epidemiology which looks at diagnosed cases. It was not syndromic surveillance which is looking for well-known diseases in patterns of symptoms recorded by clinicians. Our approach was more formative: We drew on data such as search terms, poison control reports, and emergency medical service (EMS) records that included the chief complaint in the patient’s words as well as the EMT’s notes that fell short of the formal data that becomes a clinical health record. These were messy and disorganized and rarely brought together for analysis. But that was the purpose of our project.
To oversimplify, we proved the concept in retrospect and prospect.
The first by having a team of analysts figure out that there had been a norovirus outbreak in North Carolina without knowing anything about the pathogen or understanding the resulting disease. They were essentially mathematicians looking for anomalous patterns in years of data comprising tens of millions of data points. We did not even tell them what to look for, just gave then access to the data and said to find anything interesting. They did, and not only detected the outbreak, they were able to predict it before it could show up in reportable disease registries that public health officials use.
The second was a look into the future. Again, without knowing what might appear or how severe it might be. We examined search data in the summer of 2009 and localized influenza-like terms to a few counties across two states. We then used EMS data from those areas to look for buried references to the same types of issues and discovered that there looked to be a sudden spike that our predictive analysis indicated would be a major influenza outbreak in the upcoming flu season, well above the average from previous years. We reported these findings in September to Homeland Security in a white paper. They thanked us, and we heard little more.
They got back to us in December as the flu season was fulfilling our predictions with alarming accuracy. They wanted to know more. We tried to explain that the utility of our findings was greatest in September and October when steps to prepare for a greater than usual outbreak would be effective. But they wanted to show something on their giant status screens. We responded that when such an event was visible at that level, it was no longer a prediction but the event itself. Rather than see us as the warning sign that a curve was coming up and you should slow down, they wanted us to tell them how to drive at the peak of the curve
Rather than populate the big boards in their emergency operations center, we had tried to convince Homeland that we should be empowering local health care and public health officials at the county level and below: people like school nurses, private practice clinicians, and county public health directors. We wanted to create a biosurveillance version of hyperlocal weather forecasts that capture in a sentence what a person wants to know. (E.g. Today will have a low of 63F and a high of 86F with a 30% chance of afternoon showers.) They were uninterested.
The reasons were organizational and budgetary. Homeland in this context saw themselves in a zero-sum game with CDC. Any gain by CDC would be a loss to them in annual budgets. Moreover, the overlapping jurisdictions for health threat detection did nothing to encourage their cooperation and leverage resources in common. Each agency had their congressional champions and detractors and crossed budgetary swords every year.
While we offered an interesting approach that could have become a national information system, the political and budgetary realities prevented Homeland from taking a risk by expanding our activities since it might prompt a fight in Congress.
Innovation is more than a new or better idea. It’s a new or better idea that people use. There are plenty of novel thoughts and even developments, but any current set of organizations can often be a barrier to advancement of them. The trick is to break through those barriers, even if it means upsetting the established order.
I wish the ARPA-H group well, but they have a difficult road ahead of them.
