What is a preliminary radiology read, and what is it not?

A preliminary read is a documented early interpretation issued before the study is finalized, usually so a treating team can decide something now. It carries a named reader, a timestamp, and a label marking it preliminary. It is not an informal phone opinion, and it is not the version that governs the record.

Preliminary describes status, not effort or quality. The reader usually has the images, whatever history the order carried, and often no prior studies yet. The preliminary answers the question in front of the team right now, and it answers less than the final will. That is deliberate. Facilities get into trouble when they treat the preliminary as a shorter version of the final rather than a different document with a narrower job, because the two failure modes are opposite: one leaves the treating team waiting, the other leaves everyone downstream assuming the study has already been read in full. If your policy never says what a preliminary is allowed to leave out, the reader and the clinician end up working from two different assumptions about the same paragraph.

Local names vary. Wet read is the old term and still gets said out loud. Some systems label it an initial report, others a provisional interpretation, others simply preliminary. The label matters less than whether the document is stored in the record, attributed to a named reader, timestamped, and capable of being marked superseded once the final is signed. A preliminary that lives only in a phone call and a line in a nurse's note cannot be audited, and nobody can reconstruct from it what was known at the time.

Who is allowed to act on a preliminary read?

The clinician managing the patient at that moment. Acting on a preliminary is normal and often the reason it was issued at all. What has to be explicit is that the read is preliminary, that a final is coming, and who is responsible for looking at that final when it lands.

An emergency physician who waits for a signed final before making a disposition decision is not practicing more safely, only more slowly. The risk sits downstream. The preliminary text gets quoted into a discharge summary, a consult note, or a transfer packet, and from that point it reads like settled fact to everyone who was not in the room. A hospitalist picking up the patient sees the quoted impression, not the status flag. The primary care office receiving the discharge paperwork sees a paragraph with no label at all. Each of those readers is one copy further from the document that will actually be final. If your discharge summary template pulls the impression automatically, check whether it pulls the status label with it. Most do not.

Two teams usually disagree about who owns the follow-up. The emergency department's position is that the admitting or receiving clinician owns everything after handoff. The admitting team's position is that radiology will call if the final changes. Neither is written down, so the gap only becomes visible when something falls through it. Write it down. Whoever your organization names, the name has to survive a shift change, a weekend, and a substitute physician covering a service line. Coding should work from the final rather than the preliminary, and an outside institution that received a preliminary by fax needs a defined route to the final.

How does the final report differ from the preliminary read?

The final is signed, complete for the scope of the study, and written once the radiologist has whatever priors, comparisons, and clinical context the workflow could supply. It picks up incidental findings the preliminary set aside. It is the report that reaches coding, the referring office, the patient portal, and any later review of the encounter.

Several things change between the two documents. Priors arrive. A comparison study that was still transferring overnight is available by morning, and a finding that looked new turns out to have been sitting there for years. The radiologist has time to work through the parts of the study the urgent question never touched, so incidental observations appear that the preliminary set aside on purpose. Language tightens. A phrase like cannot exclude may narrow in the final, or it may stay exactly as it was because the uncertainty is real. In some workflows a second reader is involved, and the final reflects a conversation the treating team never sees. The preliminary was not sloppy. It was early.

The final also has consumers who never speak to radiology. Coders work from it. Quality registries pull from it. Referring offices file it. Patient portals release it, sometimes on a timer that starts at signature and pays no attention to whether the ordering clinician has opened the result. Every one of those consumers assumes the text it holds is current, and most have no mechanism for noticing that it changed. Ask your portal vendor what the release clock actually keys off, because that answer decides who reads a corrected report first.

What is the difference between an addendum and an amendment?

An addendum adds. An amendment corrects. An addendum documents something that was not in the signed report and does not contradict it, such as a comparison that arrived later. An amendment changes content that was wrong: a laterality error, history belonging to a different study, a finding stated in error. Both stay visible in the record.

The distinction is about what the record says happened. An addendum tells a later reader that the signed report was accurate and that more was added to it. An amendment tells a later reader that for some window of time the record contained something wrong, and that a clinician may have relied on it during that window. Those are different facts about the same encounter, and they carry different weight in a quality review, in peer review, and in any dispute where someone has to establish what was documented on a given day. Both versions have to remain retrievable. Overwriting the original text and signing again destroys the only evidence of what the clinician actually read.

Here is the awkward part. Plenty of reporting systems let a radiologist reopen a signed report, edit the text, and sign again, with the interface calling the result an addendum no matter what changed. The radiologist takes whatever the dropdown defaults to, at speed, at the end of a shift. Nobody is being careless; the tool is not asking the right question. A quality team can test this in an afternoon: pull a sample of addenda, read the before and after, and count how many corrected something rather than added something. The count usually runs higher than the department expects.

  • A prior arrives after signature and the conclusion does not change: addendum.
  • A prior arrives and the conclusion no longer holds: amendment, and someone has to call.
  • The report says left and the images are of the right side: amendment.
  • The radiologist adds a measurement the referring clinician requested: addendum.
  • A dictation error reverses the meaning of a sentence: amendment.

How are discrepancies between the preliminary and the final tracked and reviewed?

By comparing the two documents for every study that had a preliminary, scoring the difference on a defined scale, and reviewing the ones that would have changed management. Most programs sort discrepancies into no difference, wording difference, and clinically significant difference, then route the last group to peer review and back to the treating team.

Someone has to read both documents. Automated text comparison is good at finding the studies worth looking at, but it cannot tell you whether a difference mattered, because report wording changes constantly without the meaning changing at all. Most programs review every preliminary-to-final pair in their higher risk categories and sample the rest. Whatever the sampling design, publish the denominator next to the rate. A discrepancy rate calculated over every study the department read is a different number from the same rate calculated over studies that actually had a preliminary, and the two get compared to each other anyway.

Grade the difference by what it would have changed, not by how many words moved. A scheme that separates no meaningful difference, wording or completeness differences, and differences that would have altered management gives the review committee something to act on. Track the clinically significant group by reader, by modality, by hour of day, and by whether a prior was available when the preliminary was issued. That last variable is the one departments forget to capture, and it explains a surprising share of the cases. Capture it at the time. Reconstructing prior availability afterward is guesswork, and the guess always flatters the department.

Keep the reporting question and the communication question apart. A discrepancy caught at finalization and phoned to the treating team right away is a different event from the same discrepancy surfacing weeks later during an unrelated chart review. The first is the process working. The second is the process failing quietly, and the report itself reads the same either way. If your log stores both as discrepancy, clinically significant, you have thrown away the distinction that tells you where to spend effort. Add a field for how the discrepancy was found and keep it mandatory.

How should a facility route a changed result back to a clinician who already acted?

Treat it as a communication event with a named owner, not a records update. Whoever signs the amendment or the significant addendum should trigger direct contact with the clinician currently responsible for the patient, document who was reached and when, and keep escalating until someone acknowledges it. Filing it in the chart is not notification.

This is where it breaks, and it breaks the same way almost everywhere. The amendment gets signed. The system flags the report as amended and drops it into a results inbox. The inbox belongs to whoever populated the ordering provider field, which was the overnight emergency physician, now off shift, or a resident who has rotated, or a generic department account nobody has opened since the last upgrade. The patient has already been discharged. The primary care office never had the study in its queue because the study was never theirs. The record is accurate and completely inert. Later, someone reading the chart for an unrelated reason notices that the impression is not what the discharge summary quoted. No individual step failed, which is exactly why this survives audits.

The fix is unglamorous. Route on the clinician currently responsible for the patient rather than the field that captured who ordered the study, and accept that answering that question sometimes takes a phone call to a nursing unit. Require a human acknowledgement; a system read receipt tells you a record was opened, not that a person understood a result had changed. Log the attempt, the time, the name of the person reached, and what was said, in the record and not in an email thread. Put a number in the policy for significant changes, because a policy that says promptly gets read differently on every shift.

One more thing catches facilities off guard: the patient may see the amended report before any clinician does. Portal release rules and amendment notification rules are usually written by different committees, in different years, and almost never tested against each other. Decide in advance what a patient sees when a report changes, and who answers the phone when they call about it. That call lands at the imaging facility or at the ordering office, and both should know which one is taking it. Write the answer into the portal policy rather than the radiology policy, because the portal team is the one that will get asked.

  • Name the role that owns notification for an amended result, and name the backup for nights and weekends.
  • Route to the clinician currently responsible for the patient rather than to whoever populated the ordering provider field.
  • Require a person to acknowledge; a system read receipt is not an acknowledgement.
  • Log the attempt, the time, the person reached, and what was communicated.
  • Define what happens when the patient has been discharged and no clinician owns the result.
  • Test the pathway with a drill that uses no patient information, before you need it.

Sources and scope

  • DICOM Standards Committee, Digital Imaging and Communications in Medicine standard
  • American College of Radiology, practice parameter for communication of diagnostic imaging findings
  • Radiological Society of North America, RadReport reporting-template resources
  • HL7 International, standards for clinical document status and exchange
  • American Health Information Management Association, professional guidance on health record integrity and amendments