On the value of portfolio diversity in heart, lung, and blood research

In his widely acclaimed book The Difference,¹  Scott Page, a Professor at the University of Michigan, described a computer modeling experiment designed to test the “Diversity Trumps Ability Theorem.” The Theorem postulates that “collections of diverse individuals outperform collections of more individually capable individuals.”^{1,page10–11} The computer model showed that diversity enhanced the ability to solve problems or make accurate predictions,²  but only when 4 conditions were met: (1) the problems were difficult, (2) all problem solvers were “smart” (but not the smartest), (3) diversity was sufficient to assure that different problem solvers could exploit the solutions of others, and (4) the populations of problem solvers and collections of problem solvers were large.¹  As all 4 of these conditions are clearly met in heart, lung, and blood (HLB) research, we were stimulated to examine the diversity of topics and mechanisms in the National Heart, Lung, and Blood Institute (NHLBI) portfolio.

To further support his argument for the benefits of diversity, Page cited a number of empirical examples, including cities,³  policy-making agencies, management teams, and groups of scientists.¹  Other authors have cited examples supporting the importance of diversity in science: multidisciplinary interactions have repeatedly been shown to generate greater degrees of rigor, creativity, evolution of ideas, academic productivity,⁴  and innovation.^5,6  Page argues that, when faced with difficult problems, different people (or more generally different agents) can bring different “toolboxes.” Diverse toolboxes offer various perspectives, interpretations, heuristics, and prediction models. Diversity works, both theoretically and empirically, because application of many different toolboxes reframes confusing data into eminently solvable problems and because diverse agents naturally build on each others' work.^1,2 

Science is one of society's most valuable diverse toolboxes for improving health and for serving as a sound economic investment.⁷  Over the past 60 years, diverse groups of government-funded researchers in basic, translational, clinical, and epidemiologic sciences played pivotal roles in enabling dramatic reductions in cardiovascular mortality and morbidity.⁸  Research Amer!ca reports that every 1 million dollars invested by the National Institutes of Health (NIH) generates 2 million dollars of new state business activity.⁹  Yet, the value of a diverse biomedical science toolbox, funded through an investment that is relatively small given the nation's total health expenditures (about one penny for each dollar)^9-11  is questioned by some, with concerns being raised about what constitutes “worthy” science or wasteful projects.¹²  Others criticize funding agencies for being too conservative or spending too much in specific areas.¹³  Some scientific thought leaders have even called into question the value of funding research projects outside their own spheres or expertise.^13-16  Unfortunately, there is no simple and universally accepted approach for allocating finite research resources.¹⁷  Indeed, many of the arguments being heard today were also raised 20 years ago in another period of budgetary constraint.¹⁸ 

Not surprisingly, whenever the economic landscape forces limitations on research funding, organizations that support science engage in a cyclic soul searching exercise. They face choices regarding types of science to fund (basic, clinical/applied), levels of risk to assume (sure bet vs high risk/high reward), sizes of awards, and distributions among different types of applicants (individuals and/or large teams). The NIH recently invited the scientific community to provide input on “How do you think we should manage science in fiscally challenging times?”¹⁹ 

The NIH has long supported a diverse, balanced mix of basic and applied research. The basic-to-applied funding ratio has remained remarkably constant: in 1994, 57% of NIH research funding supported basic research, whereas 43% supported applied and development research. In 2004, the corresponding values were 55% and 45%, respectively,²⁰  whereas more recently they were 56% and 41%, respectively.²¹  Even with the formation of the new National Center for Advancing Translational Sciences, the NIH remains publicly committed to maintaining that traditional balance.^22,23 

Just as in other fields, scientific diversity has been and continues to be critical for the success of HLB research. What do we mean by scientific diversity? Stirling cites 3 parameters: variety, which refers to the number of categories; balance, which indicates “how many of each”; and disparity, which describes how well categories can be distinguished.⁵  As with other natural or man-made environments, the survival, evolution, and eventual success of scientific ecosystems depend on their ability to capitalize on diversity in variety, balance, and disparity, especially under challenging conditions. For instance, it has been argued that the driving forces in the growth and development of cities and regions can be found in the productivity gains associated with the clustering of a diversity of talented people (human capital).³ 

Schneider offers a different construct, proposing that scientists come in 4 “flavors,” all of which are essential for moving any scientific field forward.²⁴  The scientists of the first flavor excel at being able to visualize the “fuzzy front end.” Their out-of-the box ideas are then translated into doable experiments designed and executed by scientists of the second and third flavors. Their experiments allow the new ideas to be methodically tested, and then synthesized and further developed into new hypotheses by the fourth flavor of scientists who collect, categorize, interpret, and pass on large amounts of data. Schneider's categorization may be oversimplified, but it illustrates how biomedical science is a relay exercise, better a collection of relay exercises, by which scientists (better groups of scientists), interact to solve the many complex problems presented by human health and disease. As a community, we are most successful when we achieve active engagement of diverse problem solvers, including basic scientists, engineers, translational and clinical researchers, clinical practitioners, statisticians, policy experts, patients, communities, and indeed all those who can at some level understand and translate along the way.

To illustrate the intellectual diversity of NHLBI's HLB portfolio, we used the NIH funding database publicly available on the NIH RePORT Web site,²⁵  “Estimates of funding for various research, condition, and disease categories (RCDC).” The RCDC system uses sophisticated text data mining (categorizing and clustering using words and multiword phrases) in conjunction with NIH-wide definitions used to assign projects to categories. We extracted and reported here funding data only about those projects that were: (1) active in fiscal year 2010, (2) performed in an extramural location (outside the NIH), and (3) funded or primarily administered by the NHLBI. We designated projects as clinical if they were categorized as “clinical research” or “clinical trials.” We report funding levels according to research mechanism (Table 1), to clinical status, and to RCDC categories that include the great majority of HLB research (Tables 2 and 3); it should be noted that the categories are not mutually exclusive. We calculated values for numbers of projects, total funding, and quartile costs per project according to mechanism or topic using the SAS Version 9.2 “Proc Tabulate” procedure. The data indicate a generally well-balanced distribution of NHLBI funds among clinical and nonclinical projects across topics and funding mechanisms. The individual research project grant (R01) mechanism predominates funding both for the clinical and nonclinical awards, being by far the most prevalent for the latter, whereas the cooperative agreement (U01) mechanism is used by many more clinical projects. Using another NIH tool that tracks published acknowledgments to NHLBI awards, we estimated that the grant portion of the portfolio illustrated here has generated > 45 000 publications garnering approximately 2.4 million citations to date. (We are assuming that all NHLBI grantees comply with the obligation in their grant awards to include an acknowledgment of NIH funding in all manuscripts resulting from their NIH-supported research.²⁶ )

When times are tough, it is tempting to retreat into a conservative, short-sighted investment stance. However, just as with any other long-term investment portfolio expected to weather various conditions, the NHLBI needs to maintain a strong commitment to investing in a diverse science portfolio that balances risk and long versus short time pay-offs. Diversity, like any other investment strategy,⁵  has downsides, including increased transaction costs, losses of economies of scale, difficult standardization, and internecine conflicts that arise from fundamental preference differences¹  about ultimate goals. Even knowing that we oversee a diverse portfolio still begs a number of other critical issues, such as identifying hot, potentially transformative, gaps; assessing whether the balances between topics are optimal; considering trade-offs between relatively conservative (blue chip) and innovative (high risk) investments; and applying the concepts of diversity to “big science” infrastructure projects, such as NHLBI-funded population cohorts. Nonetheless, in HLB research, the 4 criteria listed by Page¹  for the “Diversity Trumps Ability Theorem” are met: heart, lung, and blood diseases are complex problems; NHLBI funding is highly competitive in all areas, meaning that only high-quality proposals are being funded; the NHLBI funds a widely diverse set of researchers and research groups who are increasingly collaborating with one another; and the research community is large. By recognizing and leveraging each others' strengths and working together¹⁸  to discover, implement, and educate, the diverse HLB research community as a whole will be better poised to continue to improve human health toward achieving the important goals set forward by health initiatives, such as the “Healthy People 2020”²⁷  and “The Million Hearts.”²⁸  A united HLB research community will also be a stronger voice, a voice that can better inform public opinion about the value of publicly funded biomedical science: it takes diversity of vision, teamwork, time, and money to back the best science.