Tips for Successful CIHR Applications
The Canadian Institutes of Health Research (CIHR) is Canada’s funding agency dedicated to health research. The CIHR funds research in all areas of health, from basic lab research to clinical research to public health research. Of particular relevance to the statistical community, CIHR does fund statistical methods research. Over the course of my career, I have been a panel member, reviewer, scientific officer, and most recently chair of a panel at CIHR, and have held funds for methods research on-and-off since 1998 including a current Foundation Scheme grant. As such, I have written this short piece as a companion to Paul McNicholas’ article in the November 2016 issue of Liaison, with some tips on how statisticians can achieve success applying to CIHR as principal investigators.
CIHR has two major programs, the Project Scheme and the Foundation Scheme. Project grants are of varying duration and amount, and are directed towards a single specific research aim and project. The Foundation Scheme funds individual researchers (almost always; very occasionally a team of researchers will be funded) for a program of research, for 5 years for new investigators (investigators with fewer than 5 years in a faculty position) and for 7 years for senior researchers. The Foundation scheme is oriented typically towards either highly successful researchers or researchers with substantial promise.
A Project Scheme grant is typically a 3-5 year grant designed around a specific project or projects. Statisticians have been successful with a variety of grants. A typical grant would revolve around development of a new method with a couple of real examples that demonstrate how the method provides improved inference to address a real health research problem. Alternatively, one might address a single key motivating health research question, and develop a set of new methods to address that health research problem. Project Scheme grants fund specific projects and the personnel and equipment needed for these projects. Typically, this includes both research assistants and students. Funding levels can range widely, from \$50,000 or less per year to more than \$1 million per year for randomized trials. Over the last few years there have been changes in grant programs, but the current Project Scheme review process is organized into committees; of note, the Public, Community & Population Health - (PH1) committee lists in its description that “The mandate includes the development or application of novel statistical methods.” This is a natural home for grants addressing statistical methods with health applications.
A Foundation grant is for a set time, and is for a program of research centred around the principal applicant with the goal of supporting a sustainable foundation of health research leaders. The program of research should include a series of integrated research projects, as well as knowledge translation and training programs. Much less emphasis is placed on the details of individual projects, and much more on the researcher’s profile and potential for ongoing successful research. These grants are funded based on the applicant’s prior funding levels; CIHR computes the amount based on past levels of success with operating grants. While there is no explicit mention of statistical methods in the mandate of this competition, statisticians have been successful in receiving Foundation grants. The Foundation Scheme is currently under review at CIHR, however. Some changes have been made for the current competition, but more changes (including, possibly, cancellation of the program) are forthcoming.
CIHR also has occasional targeted competitions. These are oriented towards a focused research question or set of questions. These questions may not be that focused (recent ones include analysis of existing cohorts, and personalized medicine), and may include methods projects within the range of potential topics. These targeted competitions often have higher success rates than regular competitions.
Finally, CIHR funds large networks or teams, under a variety of funding mechanisms. One such network is the Canadian Network for Observational Drug Effect Studies (CNODES), of which I am a co-principal investigator. While it is atypical for a statistician to lead such a large network, these networks almost inevitably require statistical collaborators; it is often feasible to have some of the budget from these large networks set aside for targeted methods research. It is worth being aware of the potential for such networks, and to get involved when relevant, but most importantly to discuss up front the potential for dedicated methods funding.
One key difference between the open CIHR competitions (Project and Foundation) and NSERC Discovery Grants is success rates. For the 2016-17 Discovery Grant competition, the success rate for NSERC for returning funded applicants in Statistics was over 95 % ; the success rate in the most recently reported CIHR Project competition was 16.5%, and Foundation grant success rates are lower. It is typical that an application will have been submitted, rejected, and revised more than once before eventually being funded. However, it is important to remember that the grants are much larger than a typical Discovery Grant.
While the success rate at CIHR overall is low, statisticians appear to have, in fact, an above-average success rate (survey results presented at SSC 2017). Statisticians have been successful with grants in statistical genetics, in methods development for clinical research, public health and health services research, and in pharmacoepidemiology.
So given all of this, how should a statistician apply to CIHR? What are the characteristics of a successful application?
- A grant should have a health-research goal. This might suggest that statistical grants are not accepted; however, this is far from the case. The key is to identify a way that a new statistical development will move health research forward; for example, if it is possible to show that current statistical methods are unable to answer an important research question, this leads naturally to the need for development of a new statistical tool.
- As a natural consequence of the above, the project should use real datasets that address a real health problem; a “toy” dataset typically won’t cut it. It can be a re-analysis of (not too) old data, but the data must be real and be useful for addressing real health problems. Ideally, the project should be able to provide new knowledge about health via data.
- Typical CIHR grants have multiple co-investigators (these are known as “Named Experts” in the case of Foundation Scheme Grants). It is important that a statistical methods grant have appropriate co-investigators. Some may be other statisticians or methodologists but it is important to also have clinical or epidemiologic researchers with expertise in the substantive area under study as evidence that the health component of the research will be appropriately conducted and disseminated. Having expertise to address questions of substance and clinical relevance is essential to funding.
- Theoretical development may be of interest, but it should be there to further the health research goal; theoretical development for its own sake would not be of interest to CIHR. But if this development serves to demonstrate the qualities of a method, it is pertinent. Sometimes, simulation methods that address the properties of a method in real-world settings may be more relevant than theoretical development.
- Knowledge translation is important for all CIHR grants. For a statistical methods grant, this can take the form of software (e.g. R packages), tutorial papers, workshops or other material designed to make the newly-developed methods useable by other researchers including non-statisticians.
There are a few other important considerations. First, unlike NSERC, the review committee is mostly non-statisticians; if you’re lucky, there will be one statistician reviewing your grant, but there’s no guarantee that that statistician will have expertise in your area. As a consequence, mathematical language won’t be helpful. If necessary, put it in an appendix but make sure that the main text communicates clearly the objectives and approach without resorting to complex mathematical language. In addition, your grant will be read in-depth by at least one non-statistician. As above, make sure it’s clear to a non-statistician health researcher why this is relevant to health research.
As noted above, it may take several tries to get funded. This means that reading the reviews, and responding to them, can be very informative. It’s important to read for subtle clues in the reviews; are they telling you to submit again, or that your idea is flawed and needs significant rethinking, or are they telling you to clean it up a bit and resubmit? Often, the reviews themselves are more informative in this regard than are the scores themselves. It can also be helpful to take advantage of internal reviews. Most hospital-based research institutes now have strong internal review processes; one can learn a lot from having the grant reviewed by non-statisticians with expertise about the funding system.
In summary, submitting to CIHR can be discouraging, but for the right project, it is worthwhile. Most submissions to CIHR, across all fields, will not be funded. And for statisticians, it does take an important re-calibration of how we write our proposals. While the proposal can and should address methodological development, it must be grounded in addressing real problems in health research. However, statisticians are as, or more, successful than average when applying to CIHR, and the funding amounts are substantially higher than those for NSERC, so these risks may be worth the reward. Biostatisticians should be aware of, and take advantage of, opportunities to get funded by CIHR.
Albert Boehringer Chair in Pharmacoepidemiology
Professor, Department of Epidemiology, Biostatistics, and Occupational Health
- NSERC presentation at SSC, June 11, 2017, Winnipeg