Tracking the street level location of biomedical publications, we reveal clusters of innovative activity within cities. Using the keywords describing the content of the publications we can further track the geographic flow of new ideas, and in turn, devise a measure of innovation capabilities in the biomedical sciences and in bio-pharmaceuticals.
The geography of innovative output is established in two steps. First, clusters are aggregated by connecting all loci of production less than 1km apart. Once these connections have been made a cluster consists of a set of loci, which can all be reached using those connections. Figure 2A, for instance, shows the Boston cluster of biomedical research (in red). Most cities are covered entirely by one cluster (see the examples below for Paris, San Francisco, and Basel). In the second step we establish the cores of each cluster.
The Boston cluster has eight core areas, as depicted in Figure 2B. Similarly, we find ten core areas in Paris (Figure 3B), two in San Francisco (Figure 4B) and one in Basel (Figure 5B). A core is a small region of high output density that corresponds to at least one, but frequently more than one, leading institution.
Innovative output and capabilities of clusters can be estimated using the Medical Subject Headings (MeSH) found in the MEDLINE bibliographic database. Each publication in the database has assigned to it a set of terms, drawn from a controlled vocabulary that characterizes the content of the publication. MeSH terms are related to each other through a hierarchical tree, shown in Figure 1.
Using the MeSH terms it is possible to understand the innovation output of regions in two different ways. First, the general areas of strength of each cluster can be assessed by shading each MeSH term in the tree according to the regions relative specialization in that term (the Balassa’s revealed comparative advantage). Second the innovation output of each cluster can be measured by how frequently it is the source of new terms and how quickly it adopts new terms. The profile of specialization of selected clusters is shown in Figure 2C (Boston), 3C (Paris), 4C (San Francisco) and 5C for Basel.
Boston
Figure 2A. The Boston Biomedical Cluster
Figure 2B. Boston Main Areas of Scientific Production in Biomedical Sciences
Figure 2C. Revealed Comparative Advantage (RCA) of the Boston cluster (red, RCA>1; blue RCA<1). See Figure 1 to map the areas of biomedical investigation
Paris
Figure 3A. The Paris Biomedical Cluster
Figure 3B. Paris main areas of scientific production in biomedical sciences
Figure 3C. Revealed Comparative Advantage (RCA) of the Paris cluster (red, RCA>1; blue RCA<1). See Figure 1 to map the areas of biomedical investigation
San Francisco
Figure 4A. The San Francisco biomedical cluster
Figure 4B. San Francisco main areas of scientific production in biomedical sciences
Figure 4C. Revealed Comparative Advantage (RCA) of the San Francisco cluster (red, RCA>1; blue RCA<1). See Figure 1 to map the areas of biomedical investigation
Basel
Figure 5A. The Basel biomedical cluster
Figure 5B. Basel main areas of scientific production in biomedical sciences
Figure 5C. Revealed Comparative Advantage (RCA) of the Basel cluster (red, RCA>1; blue RCA<1). See Figure 1 to map the areas of biomedical investigation