Research protocol

Version: 0.1, status: concept. Last modified by GtR on January 20, 2020. Version control see here.

A research- or study protocol is a formal and detailed recipe. It describes why, what and how you will be collecting and summarizing observations, whether qualitative or quantitative.

What is a research protocol and why is it important?

Perhaps you wonder why we suggest that you turn your grant proposals into a study protocol. Well, grant proposals are documents that describe your plans in some detail, but they are written, first and foremost, to seduce reviewers and program committee members to give you the grant money. Therefore writing a grant proposal requires a slightly more journalistic touch where too much detail can actually hurt.

By contrast, study protocols are manuals of operation in which details are important. It is really a recipe book, detailing exactly what you are going to measure, when and how. This manual may also include several appendices describing exact procedures and measurement instruments such as, for example, questionnaires, scripts for interviews, instructions for handgrip strength measurement or six-minute walk tests, etc. They serve you if you have forgotten particular details of the plan and allow others to take over seamlessly, should you be unavailable for some time. Together with your daily research logbook, a good protocol will also serve you later in writing your report or papers. Many fellow researchers decided to publish their protocol on the web (e.g. Open Science Framework, or another preprint server) or as a separate research paper. We encourage you to do that too.

How to write a research protocol?

A problem is that many easily found protocol templates are really modeled on protocols for (randomized) clinical trials, whereas your research may employ quite a different study design. The good news is that templates for trial protocols include many features that may inspire you even when you are dealing with quite another study design. In particular if you are trying to answer a causal question using an non-randomized design. For orientation, try to determine what kind of study design you will be using. One may roughly distinguish the following designs:

  1. descriptive (e.g. a survey)
  2. diagnostic (discriminatory performance of some test; test-retest reliability)
  3. prognostic-descriptive (cohort of participants with some characteristic (which may be an illness) followed over time)
  4. prognostic-intervention ((randomized) cohort)
  5. literature reviews
  6. qualitative studies (see the chapter on that)
  7. mixed methods designs (e.g. a combination of #6 and #3)

An internet search (e.g. “irb study protocol template”) on institutional review board (IRB and templates) may uncover helpful advice on how to write a protocol for various types of studies.

Some additional advice

  • Consult reporting guidelines for your study design. Although they seem to be designed for the writing phase, chances are that they will give you ideas for what to think of right now, in order to write about it when you start writing papers.
  • Consult risk of bias tools. They too may provide useful ideas on what to measure, when and how.
  • If you are planning a (randomized) trial, consult SPIRIT, the special guideline for trial protocols.
  • If you are planning a systematic review or meta-analysis, consult PRISMA-P, the special guideline for review protocols. Once written (and targeting a clinical question), your protocol may be uploaded on PROSPERO, the International prospective register of systematic reviews.
  • Many studies burden participants with measurements that are not used. So attempts to focus on key variables are useful, also from an ethics and a cost perspective, and DAGs may help. Check out the concept of directed acyclic graphs (DAG) to help you measure those determinants, and those only (efficiency!), that will allow you to draw valid conclusions from your data.
  • Related documents in your protocol package are: grant proposal, (statistical) analysis plan, data management plan. The latter two are important appendices (supporting documents) to your protocol.
  • Once it is ready, put your protocol in the public domain (e.g. via Open Science Framework, or another preprint server) or submit it to a journal as a paper. Advantages of publication on servers is that you may upload amendments as new plans on how to do certain procedures arise over the course of the project.
  • If you need formal ethics approval (see chapter on ethics), develop your protocol in that process. Then use (slightly edited) sections of that text for your preregistration activities on Open Science Framework and one of the many trial registries, such as clinicaltrials.gov. If you do not need ethics approval, develop your study protocol anyway and act as described above.
  • Whatever the case, make sure that you preregister your study, irrespective of its design. For randomized trials, many journals require you to have a trial registration number in order to publish. You may check with your target journal (if you know it) if they accept OSF.io preregistration, but it is far safer to preregister (also) in a certified registry. OSF has other functionalities than these registries. You may do this in e.g. Nederlands Trial Register (NTR) or ClinicalTrials.gov. Unlike the registries’ names suggest, they do accept registrations from all research designs. On 13 January 2020 ClinicalTrials.gov contained descriptions of over 67 000 observational studies, NTR almost 800.
  • Finally, whatever you do, never start any study, however small, without a research protocol. Nor participate in studies that lack a study protocol.
  • For more support, send an email to opensciencesupport@hva.nl

Further reading