01Two permissions, and they are not the same one
Your manager saying yes is not site permission, and site permission is not ethics approval. The organisation grants access to its people, its space and its records, in a signed letter naming exactly what you may use. Your university decides whether the study may run at all. Boards often want the letter before issuing a determination, while organisations often want the determination before signing a letter, which is the standoff nearly every candidate walks into.
The way through is a conditional letter: the site writes that permission is granted subject to ethics approval, and the board accepts it as evidence of access. Ask for it in those words and most people will simply say yes. Find out early, too, whether your employer runs its own review body, because a great many hospitals and school districts do, and theirs moves on a separate timetable from your university's.
02The data question answered too late
You already have access to the records, and that is the difficulty rather than the convenience. Access granted for your job is not access granted for a study, and treating the first as though it were the second belongs on the very short list of mistakes that can invalidate an otherwise finished project.
- Name the fields you need, one by one, then ask whether an aggregate version answers the question, since aggregate figures sidestep most of what follows
- Find out who owns each dataset. It is seldom your manager and usually an information team with a request form and a queue of its own
- Ask whether somebody else can pull the extract and strip identifiers before it ever reaches you
- Agree where it will live. An encrypted institutional drive, not a personal laptop, and under no circumstances a personal cloud account
- Write down what becomes of it afterwards, including a disposal date, because both reviewers will ask and a blank there returns the packet
03Recruiting people who report to you
Where you supervise, teach or care for the people you want to study, an invitation from you is not a free choice however warmly it is worded. Reviewers look for this and they look carefully. The remedy is not to abandon a good project; it is to hand the approach to somebody neutral, keep out of the room while it happens, and remain genuinely unaware of who said no.
Then say all of it yourself, in the submission. A conflict you disclose and manage reads as a careful investigator at work. The identical conflict spotted by a reviewer reads as something you were hoping nobody would notice. Declare the employment, the reporting lines, anything your role gains if the project succeeds, and the safeguard you propose against each. It takes ten minutes and saves a revision round, and the consent document carries a version of it too, because a participant needs to know the person treating them is also the person running the study.
04The politics that never make it into a proposal
A practice change lands on people who did not choose it: the staff who now have an extra step mid-shift, the manager whose numbers move, the department that quietly tried this two years ago and stopped. Any of them can finish your project without a single form being involved, so treat the following as part of the design rather than as diplomacy.
None of this is the science, and all of it is why projects stall in month four rather than in the proposal. Where the writing has to keep moving while access is still being negotiated, that is chapter work, and the ethics packet with its letters and its conditional wording is the IRB line. Getting a signed site letter into your inbox is a genuinely good day, and worth marking.
- Find out what was attempted before and why it ended, then account for it in your plan rather than repeating it
- Bring the people who will do the work into the design early, so the protocol survives contact with a real shift
- Agree what happens if leadership changes partway through, which occurs far more often than proposal templates suggest
- Keep a named sponsor senior enough to unstick things, and meet them to a schedule instead of only in emergencies
- Write an exit: what leaves with you, what keeps running, and who owns it once you have graduated
FAQQuestions to control
Is quality improvement work research?
Sometimes, and the distinction is made by your board rather than by you. Work intended to improve care inside one organisation, using its own established methods, is often determined not to be research. The moment you intend to generalise the findings or publish them as knowledge, the answer usually changes. Ask for the determination in writing before you begin, since it cannot be applied backwards afterwards.
My manager is happy for me to do it. Is that enough?
Not on its own. A manager can authorise your time and your access to a unit, but permission for a study usually has to come from whoever can commit the organisation, which may be a director, a research office or a designated officer. Ask your manager who signs this kind of letter, and expect the answer to involve a desk further up than you were expecting.
Can I use data I already see every day at work?
Not by default. The access you hold as an employee is granted for care or operations, and a study needs its own authorisation naming the fields, the period and the purpose. In practice this means a formal data request, often fulfilled by somebody else who de-identifies the extract first. Allow several weeks for that step, since the waiting list it joins belongs to another team entirely.
What if I change jobs partway through the project?
Tell your chair and the board immediately, because your access, your site permission and possibly your whole design depend on the employment that has just ended. Some projects can continue under a new agreement with the old site as an external investigator. Others have to be reframed at the new organisation. Both are survivable, and both are far worse if discovered late by somebody other than you.
