The same routine head CT is being done at up to 3.4 times the dose in different departments

Design

retrospective k-means cluster analysis of an international CT dose registry

Population

864,182 adult and 40,027 paediatric head CT scans, January 2015 to March 2021

Primary outcome

size-adjusted dose-length product and CTDIvol across acquisition clusters

Effect

more than two-fold and up to 3.4-fold variation in every stratum; adult helical scans 372 to 998 mGy·cm (relative dose 2.7)

Acquisition parameters from 864,182 adult and 40,027 paediatric head CT scans in an international dose registry, collected between 2015 and 2021, were sorted by unsupervised clustering on tube current, voltage, collimation, scan length and pitch. Scans were analysed separately by age group and by helical or axial approach, giving four strata and between four and seven clusters in each.

In every stratum, mean size-adjusted dose-length product varied more than two-fold across clusters, and up to 3.4-fold. Adult helical scanning - the largest group - ranged from 372 to 998 mGy·cm, a relative dose of 2.7. These are all routine, clinically established protocols in current use; none is exotic. The low-dose clusters generally used lower tube current and voltage, but no single parameter explained the difference, and several combinations arrived at a low dose by different routes.

The point for a working department is that this variation is invisible from inside it. A protocol is set once, often at installation by an applications specialist, and then runs for years without anyone comparing its output against anything. Head CT is among the highest-volume examinations in most Indian hospitals and is heavily used in children after minor head injury, where the dose question is sharpest. The action is to pull your own department's median size-adjusted dose-length product for routine head CT and compare it against published reference levels - and if it sits at the high end, to ask the physicist to rebuild the protocol rather than to assume the scanner is doing the right thing.

• Pull the median CTDIvol and dose-length product for your routine adult and paediatric head CT protocols from the scanner's dose report
• Compare against published diagnostic reference levels rather than against your own history
• Check that a separate paediatric head protocol exists and is actually being selected, not an adult one with reduced mA
• Involve a medical physicist before changing tube current or voltage - image quality has to be checked alongside dose
• Repeat the audit after any scanner software upgrade; protocols do not always survive one

Why it matters

It means a department can be delivering nearly three times the radiation of another for the same examination without anything looking wrong.

Don't overread it

Diagnostic image quality was not assessed - the lowest-dose cluster is not automatically the right protocol.

The statistics, in plain English

Size-adjusted dose-length product corrects for patient size, so a 2.7-fold spread between clusters is a real difference in radiation delivered, not a difference in who was scanned. Clustering is descriptive: it groups protocols that resemble each other and shows what dose each group produces, but it cannot tell you that the low-dose cluster produced adequate images - image quality was not an outcome here. That is why the action is an audit and a physicist's review rather than simply copying the lowest-dose parameters.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.