The Finding Nobody Followed Up: Closing the Loop Between a Report and a Referrer

A radiologist reads a CT ordered for something else entirely, notices a small pulmonary nodule, and writes a sentence recommending a follow-up study in six months. The report goes to the referrer. Six months pass. The clinic has no idea whether the patient came back, whether the referrer saw the sentence, or whether the nodule is still small. There is no register, because the recommendation exists only as prose inside a document nobody can query.

This is the transformation we would run for that clinic. It is presented as a concept, built on the same method we use in client engagements, so the before, the after, and the path between them are all visible.

Why does the loop stay open?

Because the report is treated as the end of the transaction. The clinic’s system is built to produce and deliver a document, and it has no concept of a commitment made inside that document. A follow-up recommendation is a commitment with a date on it, and it is stored as a sentence.

The surrounding pressure makes this worse rather than better. The Canadian Association of Radiologists, with Deloitte, reports average Canadian waits of 84 days for MRI and 66 days for CT, and estimates the cost of imaging underinvestment at $64 billion a year in lost GDP. The CAR published national maximum wait time access targets in March 2026. In a system that congested, a follow-up study that nobody chases is not a study that happens late. It is a study that does not happen.

Before and after: the recommendation

Drag the handle. The before is a recommendation living inside a report; the after is the tracked register designed in its place.

The referring clinician still owns the follow-up decision. What changes is that the clinic can see its own open loops rather than assume they closed.

What we built

The concept above is the output of the same engagement steps we run in client work, and several are decisions a competent peer could argue with.

• We made radiologist confirmation of every extracted recommendation mandatory, one click per report, and refused to make it skippable. An extraction that quietly missed recommendations would produce a register that looks complete and is not, and a false sense of closure is worse than an acknowledged gap.
• We stopped escalation at the referrer and never let it reach the patient. This is the decision most likely to be argued with, because a patient with an untracked nodule is the person with the most at stake. A report is not the whole clinical picture, and a clinic contacting patients directly about findings, without the clinician who knows their history, would eventually cause a serious harm.
• We built the register as a first-class object with a date, an owner and a state, rather than as a reminder. Reminders get dismissed; open items get reviewed.
• We made unanswered escalations a person’s problem on a schedule, because an automated chase that never terminates is just a slower way of doing nothing.
• We wire evals from day one against the clinic’s own historical reports, scored on whether the extraction found every recommendation a radiologist would have flagged, with recall weighted far above precision.

How it works

The RIS stays the source of truth for reports. When a report contains a follow-up recommendation, it is extracted and the radiologist confirms it as part of signing. Confirmed recommendations enter a register with an interval, a referrer and a state. As the date approaches, the referrer is notified, and if the loop stays open past a threshold it goes to a named person at the clinic rather than into a longer automated sequence. A loop closes on a confirmed study, a confirmed clinical decision not to proceed, or a person closing it with a reason.

Because reports and the register stay in Canadian infrastructure and processing runs with zero data retention, the design lines up with what PHIPA expects: a privacy impact assessment before deployment, an agreement naming the processor as an agent under section 17, and an audit log on every access.

Why this pays back

The economic return is modest and the risk return is not, so a clinic should buy this for the second reason. Break-even for the first slice is about a hundred and fifty tracked follow-ups a month at the clinic’s own loaded administrative cost, which is achievable for a clinic of this size and is not the real argument.

The limit worth saying plainly, and it can lose the sale: this creates visible responsibility where there was previously comfortable ignorance. A clinic that turns this on will discover open loops it did not know about, and will then have to decide what to do about each one. Some administrators will regard that as new liability rather than as newly measurable safety. We think that reading is wrong, and it is a legitimate position, and it is much better raised now than discovered in month two.