How to Prepare an Ethics Application for a Retrospective Surgical Study

Retrospective surgical studies often begin with a familiar sentence: “We will just review the charts.” The work feels modest—no new intervention, no extra blood draw, no patient visit. Ethics committees, however, do not read “retrospective” as “outside ethics.” They read it as a study that uses identifiable patient information, often from vulnerable people, and they want to know exactly how that information will be handled. The application is where your study design becomes visible. Prepare it early, write it plainly, and treat it as part of the research rather than a form to clear before the real work starts.

A strong application does more than satisfy a committee. It forces you to clarify the question, justify the cohort, define the variables, and decide who will have access to what. That preparation pays off later, when the dataset is being built and the analysis plan is under pressure. Reviewers can usually tell the difference between a study that has been thought through and one that has been retrofitted into an ethics form.

The protocol is the application

Before you open the online form, write a short protocol that a busy surgeon could understand in ten minutes. State the question, the design, the setting, the period, and the patient group. Be concrete. “All patients who underwent gallbladder surgery” is not a cohort; “adults who had a primary laparoscopic cholecystectomy at our hospital between January 2019 and December 2022” is. Define inclusion and exclusion criteria, the primary and secondary endpoints, and the variables you will extract. If you plan to compare two operative techniques, say how patients were selected for each and acknowledge selection bias.

Add a brief background and rationale. This is not a grant application, but a few sentences help reviewers understand why access to records is justified. What uncertainty exists in current practice? Why can a retrospective chart review answer it? What might change if the findings are positive or negative? For example, a study of 30-day readmission after emergency hernia repair may be justified because local practice has changed, readmission rates are unknown, and the results will inform discharge planning. That context makes the request for identifiable records feel proportionate rather than speculative.

Ethics committees are not statistical review boards, but they need enough detail to judge risk and burden. A one-page synopsis attached to the protocol helps. It should answer: what data, from whom, for how long, stored where, accessed by whom, and destroyed when. Keep the protocol and the form consistent. If the form asks for a summary and your protocol says something different, reviewers will ask why. That delay is avoidable. Make the synopsis a living document and update it if the design changes before submission.

Name the team and their roles. Who will screen charts, who will extract data, who will have access to identifiers? Committees want to see that access is limited and supervised, not simply that the team is well intentioned. Identify the supervising consultant, the data custodian, and any statistician involved. Clarify whether trainees will access records and how they will be trained. If the study involves multiple sites, name the local lead at each site and describe how data will be transferred and combined.

Build the dataset backwards

Data minimization is the principle that you collect only what the question requires. It is also a practical way to make your application credible. Start with the analysis plan and work backwards to variables. For each variable, ask: will this appear in a result, a table, a confounder adjustment, or a validation step? If not, do not collect it. For a study of postoperative complications after hernia repair, you may need age, ASA grade, BMI, smoking status, operative time, mesh type, urgency, and 30-day complications. You probably do not need full home address, occupation, or free-text clinic notes unless they are essential to define an outcome.

Design a data extraction form and a codebook before submission. The form should use pre-specified categories and clear definitions, not free text wherever possible. The codebook should define every variable, allowed values, and handling of missing data. For example, “conversion to open” needs a definition: unplanned conversion after laparoscopic port placement, not planned open surgery. “Readmission” needs a time window and source. Reviewers may not check every definition, but a clear form shows that identifiers will not be copied indiscriminately into a spreadsheet.

Plan de-identification and linkage. Assign study IDs at extraction. Keep the master linkage file—the list connecting study IDs to medical record numbers—in a separate, access-controlled location. Collect dates only as needed; sometimes intervals such as days from surgery to complication are enough, while full dates may be necessary for seasonality or long-term follow-up. Store the final analysis dataset separately from the raw extraction file. Describe who holds each file and when identifiers will be destroyed. A simple data flow described in words is acceptable: source records to extraction form, then to study ID dataset, then to analysis dataset.

Pilot the extraction on ten charts and report that in the application. It shows feasibility and catches ambiguous variables early. If two extractors will work, describe training and how disagreements will be resolved. This is not only a methods detail; it reassures the committee that access to records will be structured and limited. It also protects you from discovering halfway through the study that three different people have interpreted “major complication” in three different ways.

Consent, waivers, and the meaning of “retrospective”

Many retrospective studies request a waiver of consent. A waiver is not automatic. You need to explain why the research could not practicably be carried out without access to identifiable information and without consent. For chart review, the typical argument is that contacting hundreds of former patients would be disproportionate, could cause alarm, and many patients may have moved or died. You should also state that the study involves no more than minimal risk, that the waiver will not adversely affect patients’ rights or welfare, and that the data will not be used to re-contact patients unless they have agreed. Check your local rules: some committees require a specific waiver form, while others require the justification inside the protocol.

If you need identifiers for linkage, say so explicitly and explain the plan to remove them as early as possible. If you plan to contact patients later for a follow-up study, that is a different project and should not be hidden inside a retrospective chart review. Be equally clear about vulnerable groups. Emergency surgery, pediatric surgery, and patients with cognitive impairment may appear in your cohort. That does not make the study unethical, but it requires a sentence or two explaining why their records are necessary and how their privacy will be protected. If the study includes deceased patients, state how you will handle records and whether local law requires anything additional.

Consent is not the only ethical issue. Transparency matters. Some institutions ask whether patients can opt out of research use of their data. If your hospital has a public notification or opt-out policy, mention it. If not, explain how the study will be reported in aggregate so that individual patients or surgeons are not identifiable. The goal is to show that the waiver is not a convenience but a justified part of a low-risk study with clear safeguards.

Privacy, security, and governance

Describe exactly where the data will live. “On a secure hospital computer” is too vague. Name the system: a password-protected institutional network drive with role-based access, for example. State that data will not be stored on personal laptops, USB sticks, personal email, or unapproved cloud services. If you need a shared folder, say who approves membership and how access is removed when team members leave. Include encryption for any transfer and use the institution’s approved transfer method. If paper extraction forms are used, describe locked storage and later destruction.

Address retention. How long will identifiable data be kept? What happens after the retention period? Who is responsible for destruction? If you plan to keep a de-identified dataset for future research, say so and check whether that requires separate approval. If you will share data with collaborators outside the hospital, describe what will be shared, whether it is identifiable, and what agreement governs it. Reviewers are more reassured by a specific, modest plan than by a broad promise to “maintain confidentiality.”

Small numbers can be identifying. In surgical studies, a rare complication in a small hospital may point to a particular patient or surgeon. Plan how you will report results: aggregate counts, minimum cell sizes, or grouping categories. If you are studying surgeon-level performance, explain how you will avoid publishing identifiable performance data without appropriate governance. This is especially important when the study could affect clinicians’ reputations or patients’ decisions. A useful habit is to write the results section of the future paper in outline before you submit the ethics application. It quickly reveals where identification risk may arise.

Risk, benefit, and the reviewer’s perspective

Most retrospective surgical chart reviews are minimal risk, but “minimal” is not “none.” The main risk is a privacy breach: names, dates, rare diagnoses, or free-text details could identify patients. There may also be risks to the institution or clinicians if the data reveal poor outcomes. Benefits may include improved local care, quality assurance, trainee education, or generalizable knowledge. Be honest about the balance. Do not claim direct benefit to the patients whose records are used. Instead, explain how the findings could improve future care and how results will be fed back responsibly.

Reviewers read many applications. Make their job easy. Use the same terms throughout: if your protocol says “30-day complication rate,” do not call it “postoperative morbidity” in one section and “adverse events” in another unless you define the difference. Answer the question that was asked. If a field does not apply, write “not applicable” and say why rather than leaving it blank. A concise justification can be more persuasive than a long one:

We request a waiver of consent because the study uses existing records from a three-year period, no patient contact is planned, identifiers will be removed after linkage, and the analysis will be reported only in aggregate. The study involves no intervention and no more than minimal risk to patients.

Writing the application so reviewers can say yes

Draft the application in plain language. Ask a colleague outside the project to read it and mark any sentence they do not understand. Avoid abbreviations without definitions. If the form allows attachments, include the protocol, data extraction form, and any required confidentiality or data-sharing agreement. Use headings that mirror the form so reviewers can find information quickly. Keep the title accurate: “Retrospective review of 30-day outcomes after laparoscopic cholecystectomy” is clearer than “An audit of surgical quality.”

Before submission, check consistency across every document. The number of patients in the protocol should match the feasibility estimate. The variables in the extraction form should match the protocol. The data storage description should match the security section. If the study changes after approval, submit an amendment rather than quietly altering the method. Committees value transparency, and an amendment is usually simpler than an audit finding later. If you receive reviewer comments, respond to each point directly and explain what changed. Do not simply resubmit the same text with a covering note.

A practical pre-submission checklist

1. Question and design: Is the research question specific, and is the cohort clearly defined by setting, period, and inclusion criteria?
2. Protocol alignment: Do the protocol, synopsis, and online form describe the same study in the same words?
3. Data minimization: Is every variable justified, and are identifiers limited to those needed for linkage?
4. Extraction plan: Is there a codebook, a pilot, and a plan for resolving disagreements?
5. Consent and waiver: Is the waiver justification complete, including minimal risk, impracticability, and no re-contact?
6. Security: Are storage, access, transfer, retention, and destruction described specifically?
7. Reporting: Will small cells, surgeon-level data, and free-text details be handled to avoid identification?
8. Team roles: Are screening, extraction, supervision, and data custody assigned to named roles?
9. Reviewer readability: Would a busy clinician understand the application in ten minutes?

The application is not a bureaucratic detour from the research. It is the first place where you show that the study can be done ethically and that the data will be handled with care. Prepare it early, keep it consistent, and write it as if a patient whose record you are using will read it. That standard will make the review smoother and the study stronger.

Photo: AymaneJed / Pixabay