GreyBrain Pulse · Published 22 Sept 2026
A year of messy logbooks into an audit report
The academician · What’s advancing · F3
An audit report that opens with the honest state of your data instead of hiding it — including what you must start recording.
Why this matters for your role
The evidence user’s skill: follow the citation, check its context and decide whether it applies.
Your worked example
A fictional journal-club abstract describes 48 residents whose mean quiz score rose from 6/10 to 8/10 immediately after a workshop. There is no control group, follow-up or patient-outcome measure.
Put it into practice
Make a claim/evidence/limitation table. Identify why this cannot establish a lasting improvement in clinical care. Draft three questions for journal club.
Clinician prompt
A year of messy logbooks into an audit report
An audit report that opens with the honest state of your data instead of hiding it — including what you must start recording.
You are helping me turn raw departmental records into an audit report. This is a practice exercise using de-identified records. Here is the data, exactly as it was recorded: [PASTE YOUR DE-IDENTIFIED LOG — inconsistent dates, abbreviations, blanks and all] Do this in order: 1. Before analysing anything, tell me what is wrong with this data: inconsistent coding, missing fields, duplicate-looking entries, dates that cannot be right. List it as a data-quality section. 2. Propose a single consistent coding scheme and show me the mapping you used. 3. Produce the summary tables: volume by month, distribution by category, and any pattern in the missingness itself. 4. State plainly which questions this data cannot answer, and what we would have to record differently from next month to answer them. Do not fill any gap with an estimate. Missing stays missing. End with a short section titled "Your check" listing what I must verify myself before using any of this.
Synthetic practice example. Run it as-is and check the output against the supplied facts.
Your check
Re-count one month by hand against the raw log. The coding mapping is the model’s interpretation of your abbreviations and is the most likely place it is wrong.
De-identified records only. Remove names, registration numbers, dates of birth and anything that identifies a patient before you paste.
Sources inform this fieldnote; their publishers do not endorse GreyBrain. This exercise does not confer clinical competence or a certificate.