TL;DR
This paper maps the geographical distribution of biomedical research in the USA and China using millions of geocoded author affiliations from PubMed. They found that research is concentrated around a few stable hubs in both countries, with most researchers being within a reasonable distance of these centers. This suggests a persistent, centralized structure in biomedical research despite overall growth.
What Problem Does This Paper Address?
The rapid growth and geographical spread of scientific literature over the past few decades has prompted an interest in understanding the spatial organization of research. Specifically, this paper addresses the question of how biomedical research is geographically distributed within large, diverse countries like the USA and China. Previous studies have explored city concentrations, country-level output, and the role of patents and industrial activity in innovation. A key challenge is understanding the relationship between geographic proximity and scientific collaboration, and whether research tends to cluster around specific "hubs." This paper aims to provide an updated framework for studying the centralization of biomedical research at national and regional levels by analyzing geocoded author affiliations.
Methods
The researchers employed K-means clustering to identify geographical hubs of biomedical research. They utilized geocoded author affiliations from nearly 20 million PubMed articles published over the past 30 years, focusing on the lower 48 US states and mainland China.
- Geocoding: Author affiliations were geocoded to obtain latitude and longitude coordinates for each publication.
- Centroid Calculation: For each country (or region within), a geographical centroid was calculated by averaging the latitude and longitude of the affiliated cities.
- Distance Measurement: Vincenty distance, which accounts for the Earth's oblate spheroid shape, was used to measure distances between cities and centroids.
- K-Means Clustering: This algorithm was applied to group research affiliations into a predefined number of clusters (k). The number of clusters was determined by observing the point at which the average distance to the nearest centroid decreased significantly with each additional cluster.
- Temporal Analysis: Centroid positions and regional cluster patterns were analyzed over time (1988-2016) to track any shifts or changes in the geographical distribution of research.
Datasets
The primary dataset used in this study is derived from MapAffil, a tool that provides disambiguated and geocoded place names for 37 million affiliations from nearly 20 million PubMed articles. This dataset covers publications from 1867 to 2016. The study specifically focuses on affiliations from the lower 48 US states and mainland China within the period of 1988-2016.
Key Insights and Ideas
- Persistent Hubs: Despite the massive growth in biomedical publications, the core geographical hubs of research in both the USA and China have remained remarkably stable over the past 30 years.
- Geographic Reachability: The findings suggest that the majority of researchers are located within a relatively short geographical distance of these identified research hubs. This implies that geographic proximity continues to play a significant role in fostering scientific collaboration and activity.
- Southward Drift: While longitude has remained relatively stable, both country centroids have experienced a slight southward shift over the decades. This shift is more pronounced in China (1.7 degrees) than in the USA (0.2 degrees).
- Centralization vs. Spread: The study highlights a pattern of centralization around a few key areas, rather than a uniform spread of research activity across the entire geographical expanse of these large countries.
- Regional Stability: Even when examining specific years, the regional clustering of research in the USA appears to be nearly identical, reinforcing the idea of stable research landscapes.
Key Results
- The overall centroid for the USA (1988-2016) is located in southern Illinois (-89.2 longitude, 38.7 latitude).
- The overall centroid for China (1988-2016) is located in Shandong province (116.2 longitude, 34.7 latitude).
- The US centroid has shifted south by approximately 0.2 degrees over 30 years.
- The Chinese centroid has shifted south by approximately 1.7 degrees over 30 years, with most of this shift occurring in the 1990s.
- For the USA, 4 centroids are sufficient to reduce the average distance to below 100 miles, and 6 centroids to bring it below 50 miles.
- For China, only 3 centroids are needed to get below 100 miles, and 4 for below 50 miles.
Limitations and Future Work
- "Lower 48" Focus: The analysis for the USA is limited to the "Lower 48" states, excluding Alaska and Hawaii, which are noted as geographical outliers.
- Temporal Granularity: While the study covers a 30-year period, finer-grained temporal analysis (e.g., yearly) might reveal more nuanced shifts.
- Future Work: The authors suggest that this work sets the stage for studying the centralization of biomedical research at national and regional levels across the globe and over time. Further research could explore how these patterns differ across various scientific disciplines or how they are influenced by funding structures and international collaborations.
Summary updated