What Actually Goes Wrong

Part of Principal Investigator (PI) Training: Inspection, Findings and the Record.

About 4 minutes. Free, in your language, no account needed.

Why this matters

Findings written against investigators personally cluster into a small number of categories, and they repeat. They are not exotic. Nearly all of them are versions of the same thing: a site that was busy, that meant well, and where nobody was systematically looking.

Knowing the categories lets you check yourself against them once a quarter, which is the whole point of this lesson.

A note on evidence: the categories below are recurring patterns described qualitatively. This lesson deliberately asserts no counts, rankings or percentages, because doing so would require a primary source that is not cited here.

The recurring categories

Failure to personally conduct or supervise. The one written against you rather than the site. It shows up as unqualified or undelegated people performing trial procedures, an investigator who cannot describe what happened at their own site, or delegated tasks with no evidence anyone reviewed them. The remedy is Module 2: unequal supervision, four routines, evidence as a by-product.

Protocol non-adherence. Assessments not performed, visits outside windows, ineligible participants enrolled, prohibited medications not caught. Almost always systemic — one flawed scheduling template producing thirty deviations — which is why the root cause matters more than the individual entry.

Inadequate or inaccurate records. Source that does not support the reported data. Missing entries. Undated corrections. Notes written long after the event with no indication they were. Rarely dishonest, frequently damaging, because unverifiable data is treated as data that did not happen.

Consent failures. Superseded form versions, consent after a trial-specific procedure, missing signatures or dates, no re-consent when required, no evidence of a discussion. Module 5's monthly ten-minute review is designed against precisely this list.

Safety reporting failures. Serious events reported late or not at all, events not recognized as serious — usually via the "important medical event" criterion — and assessments recorded without reasoning.

Investigational product accountability. Missing reconciliation, storage excursions not documented or not escalated, dispensing records that do not add up, product used outside the protocol.

Delegation and oversight documentation. Work performed before the delegation date, "all study procedures" as a task description, departed staff still listed, and training files that do not support the tasks delegated.

The pattern underneath

Look at those seven and one shape appears. In almost every case something was happening correctly and nobody was checking it systematically — until it stopped happening correctly, and still nobody was checking.

That is why the response to a finding is never only "we retrained the staff". Retraining addresses knowledge. Most findings are not knowledge failures; they are failures of the control that should have caught the lapse.

Your own quarterly self-audit

Thirty to sixty minutes, four times a year, against the same seven categories. Take three participants at random and follow each end to end.

• Consent: version current on the date, signed and dated correctly, before any trial-specific procedure, discussion documented.
• Eligibility: every criterion evidenced in source, your decision dated and signed.
• Visits: within window, all assessments performed, deviations logged for anything that was not.
• Events: everything in the notes appears on the event log, with severity, seriousness, causality and outcome.
• Product: dispensing and return records reconcile for that participant.
• Delegation: every set of initials appearing in that participant's records is on the log, for that task, on that date.
• Data: reported data matches source for eligibility, primary endpoint and safety.

You will find something. That is the exercise working. Log what you find as a self-identified deviation, fix the mechanism, and check the fix later.

When a finding lands

Do not argue the finding; understand it. Investigate the actual cause rather than the convenient one. Write a corrective action for what happened and a preventive action for the mechanism. Give both an owner and a date. Then verify, weeks later, that the change held — an unverified preventive action is the most common reason the same finding appears twice.

And answer within the deadline. A late response to a finding becomes a second finding, about the site's ability to respond.

Key takeaways

• Investigator findings cluster into seven repeating categories, led by failure to personally conduct or supervise.
• The common mechanism is an activity nobody was checking systematically — not a knowledge gap.
• Run a quarterly self-audit on three random participants across all seven categories; finding something means it is working.
• Respond to findings with a mechanism-level preventive action, an owner, a date, and a later verification that it held.