Every academic year, a wave of medical students starts their first research project with enthusiasm and ends it with half a manuscript that never gets submitted. The failure modes are remarkably consistent. If you can recognize the patterns before you fall into them, you save yourself the wasted months. Here are the ones I see most.

1. Picking a question because the supervisor suggested it, not because you can answer it

A senior clinician says "you should look into XYZ in our patient population." It sounds prestigious. You commit. Three months later you realize there are 12 charts that fit your inclusion criteria, none of them have the outcome documented, and the supervisor is too busy to clarify. You cannot back out without it looking bad.

The fix: before you commit, write the question as PICO, scope the literature for 1 hour, and ask the supervisor exactly which data source you will pull from. If they cannot answer, the project is not as well-formed as it sounds.

2. Skipping the novelty check

You assume a clinician suggesting the project has already checked whether it is novel. They almost never have. The most common bad outcome of a student project is discovering, two months in, that a Cochrane review answered the exact question last year. Spend 20 minutes on the pre-flight check. It is the highest-leverage time in the entire project.

3. Trying to do too much

"I want to look at outcomes in elderly diabetic patients with CKD across our three hospital sites, comparing SGLT2 inhibitors, GLP-1 agonists, and metformin." That is three separate studies. A first project should answer one specific question, narrowly defined. Scope creep is invisible while it is happening and obvious in retrospect when you have 400 variables in your spreadsheet and no analysis plan.

The fix: write the primary outcome and the analysis plan before you collect any data. If the analysis plan does not fit on one page, you are doing too much.

4. Not pre-registering or pre-specifying anything

You start collecting data without a written protocol. Three months in, you decide to add a secondary outcome because something interesting showed up in the data. Now your manuscript has post-hoc analyses that look like fishing. Reviewers spot this immediately.

The fix: write the protocol before data collection. Register on PROSPERO if it is a systematic review. Pre-specify your primary outcome, statistical tests, and subgroups. Deviate only with explicit justification.

5. Underestimating the IRB timeline

You expected the IRB to approve in 2 weeks. It comes back with revisions. You revise. It comes back again. Three months later you finally have approval and the deadline you had in mind is now impossible. See our feasibility piece for realistic IRB timelines — full-board review can take 3 months alone.

6. Single-handing what should be a team effort

You try to do title/abstract screening alone because "it's faster." A systematic review with single-reviewer screening is technically not a systematic review — PRISMA expects two independent reviewers. Reviewers will catch this and reject the manuscript. The same applies to risk-of-bias assessment and data extraction. Two reviewers is non-negotiable. If you cannot recruit a second reviewer, you cannot do a systematic review.

7. Not using a reference manager from day one

You start collecting PDFs in a folder called "papers" with filenames like "really_important_paper.pdf." Two months in, you have 80 PDFs, no idea which ones are duplicates, and no way to generate a bibliography. Reformatting citations into Vancouver style by hand at the end of the project consumes a weekend.

The fix: install Zotero or EndNote on day one. Use it from the first paper you read. Tag and annotate inside the manager. Use the Word plugin to insert citations as you write — bibliographies generate themselves.

8. Skipping the statistician consult

"I'll figure out the stats at the end." Then at the end, you discover your data structure cannot answer the question you posed, or your variables are coded in ways that prevent the analysis you planned, or your sample size never had a chance of detecting the effect. A 30-minute statistician consult before you collect data prevents almost all of this. Most institutions offer free statistical consultation for students. Use it.

9. Underestimating writing time

You have data. You think writing the manuscript will take two weekends. It takes 6 weeks of evenings. Every draft round with your supervisor takes 1 to 3 weeks. By the time you have a submission-ready manuscript, the project has been running for a year and you have lost steam. Plan for 8 to 12 weeks of writing and revision, not 2.

10. Choosing a journal at the end instead of the beginning

You finish the manuscript and then start journal shopping. The format requirements differ. The word count differs. The conclusion that worked for one journal looks misplaced in another. Two weeks of reformatting later you submit, get rejected, and start over.

The fix: pick a target journal before writing. Format the manuscript to their guidelines from the start. Read the last 3 issues to confirm your topic fits.

11. Ignoring negative or null findings

Your study found no significant effect. You consider not publishing. Don't. Null findings are valuable, particularly for systematic reviews and trials. The literature is already biased toward positive results. Publishing null findings is how you build a reputation as someone who reports honestly. Aim for a journal that explicitly welcomes null results, or describe the finding clearly and contextualize against existing literature.

12. Not asking for help early

Stuck on a statistical method? Confused about a screening decision? Unsure how to interpret a result? Ask within 48 hours. The students who finish projects ask for help frequently and quickly. The students who stall sit on a confusion for 3 weeks until momentum is gone. Your supervisor, your med-school librarian, your hospital biostatistician, and your senior residents are all resources. Use them.

The meta-pattern

Almost every mistake on this list reduces to one principle: front-load the thinking. The students who finish projects spend the first month planning, scoping, registering, and confirming feasibility. The students who stall start "doing the work" immediately and run into problems they could have predicted. Spend the first two weeks asking hard questions about what you are doing and why — that is the difference between a first publication and a year of wasted time.

If you want a structured starting point, run the idea through ResearchChecker first. The 15-second screen will tell you whether the idea is novel, what gaps exist, and whether the design you have in mind is feasible based on the actual evidence landscape. It is not a replacement for the thinking, but it gets you to the right thinking faster.