Estimated Publication Outputs as a Function of Number of R01 Grants per PI

Jan 30 2015 Published by under Uncategorized

In a recent post, I presented the distribution of the number of investigators who were PIs of from 1 to 6 R01 grants in fiscal year 2014.

The significance of this distribution is hard to access without some knowledge of the outputs from these grants. This is a challenging problem. Here, I attempt to estimate some of these outputs. The approach I used in very much an approximation, but some important trends can be discerned.

To estimate outputs, I collected the results in PubMed for publications from 2012 to the present for the 7 PIs with 6 R01s, the 31 PIs with 5 R01s, and randomly selected samples of 30 PIs with 1-4 R01s (screened only for name ambiguity issues or the present of large amounts of non-R01 funding).

The distributions of the number of publications over this period as a function of the number of R01s is shown below:

Pub Number Plot-250

The median number of publications for the sample of PIs with 1 R01 is 11.5. This doubles to 23.5 for the sample of PIs with 2 R01s. Essentially the same median was found for the sample of PIs with 3 R01s. The median number then increases monotonically to 44 for the PIs with 6 R01s.

In addition to the trends, it is important to note the substantial variations that occur at each level of funding. This reflects difference between publication patterns in different fields as well as the performance of individual PIs.

These data are re-plotted after normalizing the number of publications by the number of R01 awards below:

Pub Ct Normalized plot

This graph shows that the median number of publications per R01 grant is essentially constant at 11.5 for PIs with one or two grants and then drops to approximately 7.5 for PIs with 3-6 R01s.

A criticism of my earlier analyses of this type has been the lack of separation of publications in high impact journals (for this purpose Science, Nature, and Cell) because of the presumed large cost of generating such publications. The numbers of such publications as a function of PI from the assembled publication database were determined and added to the plot below.

Pub Count with CNS

PIs have between 0 and 16 Science, Nature, and Cell publications over this period. Overall, 40% of the PIs have at least 1 such paper including 17% of the PIs with a single R01 and between 37-57% of the other groups. The number of Science, Nature, and Cell publication is modestly correlated with the total number of publications with a correlation coefficient of 0.26.

The average number of Science, Nature, and Cell papers versus average total number of publications for the six groups of PIs is plotted below:

Fraction of CNS paper plot

A nearly linear relationship is observed with a slope of approximately 3.5 Science, Nature, and Cell papers per 100 total publications. From these samples, this proportion does not vary dramatically as a function of the number of R01 grants.

Again, this is a rough estimation to much many caveats apply. First, of course, publication numbers without further analysis, is a limited measure of true productivity. Second, the publication records are assembled over a period of time over which funding likely varies to some (and differing) extents for each PI. Third, relatively small samples were used for the groups with 1-4 R01s. Nonetheless, I hope these data will provide a framework for discussions about the implications of the current distribution of R01 resources across PIs and those of potential alternatives.


In response to the comment from Drugmonkey, I have plotted the number of publications normalized to the number of R01 grants with each Science, Nature, and Cell paper counted as 5 publications below:

Pub Count-CNS-5 graph


This graph looks very similar to the previous one while such papers were counted as single publications. This similarity is to be expected given that the fraction of Science, Nature, and Cell papers out of total publications does not depend substantially on the number of R01 grants as shown by the approximately linear relationship above.  Indeed, this graph looks similar even when Science, Nature, and Cell papers are weighted as 10 publications.

6 responses so far

  • Dave says:

    We are talking the number of pubs during the roughly 2 year window only (2012 - 14), right? Not the total number of pubs when you collected the data.

    Any correction for author position, or is this just all and any pub?

    • datahound says:

      Dave: It is a 3 year period (2012, 2013, 2014). No corrections for author position or anything else at this point.

  • Drugmonkey says:

    The point is not to separate out the CNS pubs, it is to multiply them by the relevant factor of cost and effort. I'd say at least a factor of 5. Maybe 10.

  • Spiny Norman says:

    You're kidding, right DM? Because my own pubs in CNS were not >1.5x the effort or expense of papers I've nad in EMBO, JCB, PNAS.

  • TheWackademic says:

    this is fantastic. Question - are you only counting PI's when they're the last/corresponding author?

    • datahound says:

      No, I am counting all papers but from a limited period (2012- ). All PIs in the analysis had publications before this so that I would estimate that almost all of these papers are from when they were PIs as opposed to being from earlier stages in their careers.

Leave a Reply