Reject analysis

Reject analysis is a periodic audit that counts how many radiographs were taken but never used for diagnosis, and records why. In general-radiography guidance a reject is an exposure never presented for diagnosis. UK dental guidance instead grades every image acceptable or not acceptable, and sets a target for the proportion.

Synonyms: reject rate; repeat rate; retake rate; repeat film analysis

What is reject analysis?

Somebody exposed a patient, looked at the result, and decided it was no use. Reject analysis is the audit that counts how often that happened over a period, and what went wrong each time.

The careful definitions come from general radiography, not dentistry. A 2023 report from an American Association of Physicists in Medicine task group defines a rejected image as a patient radiograph that was not presented to a radiologist for diagnosis and that contributes unnecessary radiation dose to the patient. Its own scope says it concerns general radiography, and that it does not apply to screen-film radiography or any type of mammography.

Dental practice runs the same idea, built differently. UK guidance from Public Health England and the Faculty of General Dental Practice recommends a two-point scale for all forms of dental radiography and dental cone-beam CT: an image is rated either diagnostically acceptable ("A") or not acceptable ("N"). "Not acceptable" means errors in patient preparation, exposure, positioning, image receptor processing or image reconstruction that render the image diagnostically unacceptable. The target is written on the acceptable side — not less than 95% for digital imaging, not less than 90% for film imaging — capping the not-acceptable share at 5% and 10% respectively.

Same idea, different name, different denominator. Know that before you compare two numbers.

What counts as a reject, and what counts as a repeat?

Two different measurements, and the task group report is explicit about the split. A reject was never presented for diagnosis. A repeat is an additional exposure beyond the number of images expected for that exam or view, but it is sent to long-term storage and it is reviewed. Repeats then sit outside the scope of reject analysis altogether — they belong to protocol and anatomy, not to image quality.

Dentistry keeps the two apart without defining either word. The UK guidance says that whenever an image is rated "N", the record should carry the date, the nature of the deficiency, the known or suspected cause, and the number of repeat radiographs taken, if any. The repeat is a separate field on the reject record, not part of the reject count.

They can also be reported as two separate numbers. An audit of one hospital dental radiology department over six months, published in 1995 and from the film era, reported an overall reject rate of 3.06% and a repeat rate of 0.93%. Same department, same six months, two figures.

"Retake" is the word most people actually say, and the one with no technical definition anywhere. The literature review that pooled the dental rates searched its three databases with a string that opened (reject* OR retake* OR reexpos*), treating all three as interchangeable. That is a fair way to find papers and a poor basis for comparing them. When a dental paper reports a "retake rate", the label alone does not tell you which denominator was used.

What it's confused with

What rates are reported, and why do published rates vary so widely?

A 2021 literature review pooled 26 dental studies, searched to April 2021, across film and digital and across students and qualified staff. The average reject rate was 11.25% for bitewings, 16.38% for periapicals, 4.10% for panoramics, 6.08% for lateral cephalography and 2.77% for cone-beam CT.

Those averages sit on top of enormous spread. Bitewings ranged from 1.95% to 27.73%, periapicals from 2.58% to 34.42%, panoramics from 2.89% to 11.65%, and cone-beam CT from 1.64% to 20.25%. A practice quoting 3% and a practice quoting 30% can both be inside the published range for the same view.

The same review reports a third figure alongside the first two. Studies that reported bitewings and periapicals together — a full mouth series, say — gave a mean of 6.69% (3910/58,407), where the separately reported figures were 11.25% (1663/14,779) for bitewings and 16.38% (2827/17,263) for periapicals. Different studies, different image pools. The review's own recommendation is that future studies report statistics separately for each type of image.

Three more reasons the numbers move:

The counting rule. The task group recommends the reject rate be calculated per press of the exposure button rather than per image, and that practitioner-directed and non-patient exposures be excluded entirely. A rate computed a different way is a different rate.

The era. In screen-film radiography the definition of a rejected radiograph also included wasted film and non-patient imaging such as quality-control exposures. The current definition counts only patient images not used for diagnosis.

The equipment and the operator. Reject data often cannot be easily compared between radiography systems from different vendors, because nothing standardises what the systems record. And the dental review reported that staff generally had a lower reject rate than students, while the 1995 audit found significant differences between qualified staff and those in training.

Even the targets disagree. The Conference of Radiation Control Program Directors suggests a range of 5% to 7%; AAPM Report 151 suggests a target of 8%, with 10% as the upper and 5% as the lower threshold for investigation, and 5% with a 7% upper threshold for paediatric clinics. All of those are general radiography. No dental body retrieved here sets a numeric reject-rate target at all — the dental target is an image-quality target.

On receptors, the honest answer is directional only. The review observed that reject rates seemed lower for film systems than digital ones, and lower for photostimulable phosphor plates than for solid-state intra-oral sensors, proposing ergonomic reasons — sensors are bulkier, less tolerable, harder to position. It also said the underlying reasons were not investigated in depth. A pattern worth watching, not a finding.

How the audit is run

The UK dental route is specified plainly. Ratings are assigned either prospectively, as every radiograph is viewed, or retrospectively, from a representative sample drawn from clinical records at intervals. Either way the analysis is done at least once every six months, on a sample of not fewer than 100 images unless the radiographic workload is too low to support that number, and reported separately for each imaging modality in use.

Every image rated "N" gets a record: date, nature of the deficiency, suspected cause, number of repeats if any. Those records are analysed on the same six-monthly cycle. Where the agreed performance targets are not being met, the guidance says identifying and acting on the cause is particularly important.

One step is easy to skip. The task group recommends that rejected images themselves be reviewed periodically, because it is not uncommon to find the recorded reject reason was wrong.

Reading a reported rate

You are far more likely to be handed a rate than to run one. Four questions make a quoted number readable.

What was counted? Rejects, repeats, "retakes", or grade-N images. Four different denominators, and the label often will not tell you.

Which views? Periapicals pool about 5% higher than bitewings, and combining views moves the figure again.

Who took them, and when? Student clinics and film-era departments are not benchmarks for a digital practice.

Against what target? If the comparison is a general-radiography threshold, say so. UK dental guidance targets the acceptable share, and notes that its targets represent what is achievable in the majority of well-managed practices — practices serving particular patient groups, such as those with additional needs, will not always reach them.

A high rate is not automatically a verdict on the operator. One investigation reported in the task group found 93.1% of rejected chest radiographs still provided sufficient diagnostic information, and 77.9% of musculoskeletal ones — implying only about 7% of those chest radiographs needed rejecting. General radiography, not dentistry, and no dental equivalent was retrieved. But it names the failure mode: a rate can rise because readable images are being discarded.

Note what the audit cannot do for you. Every figure here is an aggregate over many images and months. None of it decides whether to expose again for the patient in the chair — that is a single-image judgement made at the time, on different reasoning.

Common errors

Positioning is the recurring answer, from three directions. In dental intra-oral imaging the commonest errors leading to reject were positioning error and cone cut. In general radiography, patient positioning was the primary reason for rejection across the digital studies reviewed. And the 1995 dental department audit found positioning errors the most frequent cause.

Patient movement comes second, reported for extra-oral imaging and for cone-beam CT. In that audit, movement accounted for a larger proportion of rejects in patients under 16 than in adults, though the overall rejection rate for that group was not significantly higher.

Anatomy shows up too: the mandible had a lower reject rate than the maxilla, and the anterior region a lower rate than the posterior. Periapicals reject more often than bitewings, and the review's proposed explanation is receptor placement — a periapical receptor has to sit deep enough to capture the apices, which means the palate or the floor of the mouth. The geometry behind that is why the two views are taken differently.

Two counting errors are worth naming. The first is treating an error as a reject: the review's own stated limitation is that many studies reported common errors rather than reasons for rejection, and an error does not necessarily lead to a reject. A processing fault or an artifact that leaves the image readable is a fault, not a reject. The second is comparing your number to a published one without checking that both counted the same thing.

Practise this

Rejects start as faults you can name. Work through a real set of bitewings and see how many you would have called unacceptable — and why. Try a set on Papilla.