Examining the Interplay of NIH-Funded Research on Alcohol Use Prevention: a Two-mode Network Analysis

Francisco Cardozo

Department of Public Health Sciences
Miller School of Medicine
University of Miami.

June 29, 2023

Introduction

Introduction

  • Adolescent alcohol use is a significant public health concern.
  • The National Institute of Health (NIH) has funded numerous initiatives.
  • R grants for research and K grants for career development.
  • The interrelationships between these projects have received limited investigation.

Objectives

Objectives

  • Examine the connections between NIH-funded projects on prevention of alcohol use.
  • Uncover common themes among the projects.
  • Compare multiple grants mechanisms based on their structural similarities.

Methods

Methods

  • We searched on the NIH-Reporter platform for grants funded in 2019.
  • Keywords used: “Alcohol” and “Prevention”.
  • The project descriptions were used to construct a two-mode network.

Methods (continued)

Steps to create the network of projects:

  1. Extracted keywords from each project description.
Project keyword 1 keyword 2 keyword 3 keyword n
id 1 1 1 0 0
id 2 0 1 1 1
id 3 1 0 0 0
id n 1 1 1 1
  1. Calculated the total number of shared keywords.
id 1 id 2 id 3 id n
id 1 2 1 1 2
id 2 1 3 0 3
id 3 1 0 1 1
id n 2 3 1 4

Methods (continued)

  1. Identified projects that shared at least 90% of the keywords (40).

Methods (continued)

  1. Create a network of projects.

Methods (continued)

Network measures:

  1. Computed network measures of centrality, betweenness and constrain.

Degree Centrality: This is a measure of the number of connections a node has in a network. In a social network, for example, a person with many friends would have a high degree centrality. The more direct connections a node has, the higher its degree centrality. (Borgatti 2005)

Betweenness Centrality: This measure quantifies the number of times a node acts as a bridge along the shortest path between two other nodes. It represents the potential for a node to control information flow in the network. Nodes with high betweenness centrality can often play the role of a connector or broker in the network.(Brandes 2004)

Constraint: This is a measure of a node’s embeddedness in a network. It represents the extent to which a node is connected to other nodes that are themselves connected to each other. High constraint implies that the node is tied to a cohesive, often redundant, group of nodes. In other words, if a node is connected to many others that are also interconnected, it has a high constraint. (Burt 2004)

Methods (continued)

  1. Conducted statistical hypothesis testing to evaluate differences in network measures.

  2. Latent Dirichlet Allocation (LDA) analysis (Blei, Ng, and Jordan 2013).

LDA is a technique used for topic modeling, which is a way to categorize or summarize large collections of textual information. It helps in discovering the main topics that occur in a collection of documents. We used topicmodels package (Grün and Hornik 2011) in the R software (R Core Team 2023).

Results

Results - Projects

  • We found 147 grants related to alcohol use and prevention.

Projects

  • Direct cost $62,020,186
  • Mean cost $413,468
  • Max cost $16,697,566
  • Min cost $5,386
Organization Name Number of Projects
UNIVERSITY OF WASHINGTON 8
BROWN UNIVERSITY 7
PENNSYLVANIA STATE UNIVERSITY-UNIV PARK 5
RAND CORPORATION 5
UNIVERSITY OF CALIFORNIA, SAN FRANCISCO 5
COLUMBIA UNIVERSITY HEALTH SCIENCES 4
UNIVERSITY OF MICHIGAN AT ANN ARBOR 4
YALE UNIVERSITY 4
MASSACHUSETTS GENERAL HOSPITAL 3
MICHIGAN STATE UNIVERSITY 3
OREGON HEALTH & SCIENCE UNIVERSITY 3
UNIVERSITY OF CALIFORNIA, SAN DIEGO 3
VIRGINIA COMMONWEALTH UNIVERSITY 3

Mechanisms

NIH Research Project Grant Program (R01): - The R01 grant is the most commonly used NIH grant program. - It supports specific research projects for 3-5 years, with no set dollar limit.

“R” Series Grants: - The “R” series grants include R03, R21, R34, and R56. - R03 grants support small research projects for two years with a budget of up to $50,000 per year. - R21 grants fund exploratory projects for up to two years, with a budget usually under $275,000. - R34 grants support early peer review of clinical trial proposals.

F Series (Fellowships): - F series grants support predoctoral and postdoctoral research training. - F31 grants train predoctoral individuals in specified health areas. - F32 grants support postdoctoral applicants in health-related research.

K Series (Career Development Awards): - K series grants promote career development for individuals with doctoral degrees. - K01 grants support early-stage research careers or career redirection. - K99/R00 grants provide support for transitioning to independent research careers.

Other: - Cooperative Agreements, Investigator-Initiated Research Project Grants, etc.

Mechanisms

Network Measures

Projects

  • Mean degree of 11.7 (SD= 4.63)
  • Bootstrap test showed differences in Degree, Beteweenss and Constraint between the two mechanisms K and R01.
    • Degree: K: 15.9 (SD= 2.84), R01: 7.32 (SD= 0.736)
    • Betweenness: K: 103. (SD= 30.0), R01: 28.5 (SD= 6.59)
    • Constraint: K: 0.212 (SD= 0.0592), R01: 0.337 (SD= 0.0320)

Net of Projects

Projects

  • Four major project groups: Cancer-Biology, Genetics, Alcohol Disorders, and Testing Interventions.

Conclusion

  • Four major project groups were identified based on project descriptions: Cancer-Biology, Genetics, Alcohol Disorders, and Testing Interventions. This reveals the wide range of approaches in alcohol use and prevention research, from biological and genetic studies to the development and testing of interventions.

  • The network analysis showed differences between K and R01 grants in terms of degree, betweenness, and constraint. This suggests that these grant types are supporting different kinds of projects within the network of alcohol use and prevention research.

  • Projects who bridge the gaps or ‘structural holes’ between different topics provide diversity of information and interpretations, which gives them a competitive advantage in generating good ideas.

References

Sunbelt 2023. Portland, Oregon

Blei, David, Andrew Ng, and Michael Jordan. 2013. “Latent Dirichlet Allocation.” Journal of Machine Learning Research 3 (January): 993.
Borgatti, Stephen. 2005. “Centrality and Network Flow.” Social Networks 27 (January): 55–71. https://doi.org/10.1016/j.socnet.2004.11.008.
Brandes, Ulrik. 2004. “A Faster Algorithm for Betweenness Centrality.” The Journal of Mathematical Sociology 25 (March). https://doi.org/10.1080/0022250X.2001.9990249.
Burt, Ronald. 2004. “Structural Holes and Good Ideas.” American Journal of Sociology 110 (September): 349–99. https://doi.org/10.1086/421787.
Grün, Bettina, and Kurt Hornik. 2011. “topicmodels: An R Package for Fitting Topic Models.” Journal of Statistical Software 40 (13): 1–30. https://doi.org/10.18637/jss.v040.i13.
R Core Team. 2023. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. https://www.R-project.org/.