FDA MDR Component Codes: What an Ultrasound Service Record Has to Carry

If the record you hand over has to survive someone else's filing, the question worth designing for is not which code to write. It is which observations, once the machine leaves your bench, nobody else can recover. This note is for the service, engineering and parts side of an ultrasound department or an independent service organisation: the people who open the console, and then hand the paperwork to whoever submits the report. It is not a repair procedure and not clinical guidance, and nothing in it changes how a probe or a console is used on a patient.

The component annex defines involvement, not cause

FDA's own one-line description of the component annex is short: those codes describe "the parts and components which were involved in, or affected by, the medical device adverse event/incident." Two verbs, and neither of them is causation. The annex records which part was in the story; it does not record what caused the story. Causation has its own code set, and FDA defines that one as the conclusion "regarding the root cause of the reported event."

The difference is not classification housekeeping. "Intermittent image dropout in the cardiac preset; panel control board replaced" is a complete answer to the component annex, because it names a part that was involved. It is not an answer to the cause annex. A service report that files it as one has written a claim that the observations underneath it do not carry — and that claim belongs to a different party, in a different box.

There is a second boundary in the same sentence. Involvement is a statement about the machine; it says nothing about whether the event was reportable. Reportability is the reporting entity's call, made under its own procedures on its own facts. A service record is evidence in front of that decision, never a substitute for it, and a record that quietly pre-decides it has moved outside the job it was written for.

One event, two code sets

The seven code types are organised by annex letter, and the split is not academic: FDA's instruction page describes two distinct coding scenarios, and they do not use the same set. Importers and user facilities complete section F of Form 3500A, including the adverse event problem codes in F10. Manufacturers complete section H, including H6. F10 asks for four code types; H6 asks for all seven.

Annex	What its codes describe (FDA wording)	F10 - importer or user facility	H6 - manufacturer
A	Medical Device Problem: problems, malfunction, deterioration of function, failure of devices	required	required
G	Medical Device Component: the parts and components involved in, or affected by, the event	required	required
B	Cause Investigation - Type of Investigation: what was investigated to specify the root cause	not in F10	required
C	Cause Investigation - Investigation Findings: the findings that identify the root cause	not in F10	required
D	Cause Investigation - Investigation Conclusion: the conclusion regarding the root cause	not in F10	required
E	Health Effects - Clinical Signs and Symptoms or Conditions: what appeared in the affected person	required	required
F	Health Effects - Health Impact: the consequences for the person affected	required	required

Three of the seven are not in the hospital's box

The cause-investigation annexes are the manufacturer's to complete. That is why a hospital engineering record and a manufacturer's report on the same event can read as if they describe two different machines: one is answering the problem-and-component question, the other is also answering the method, findings and conclusion questions, and neither vocabulary is a subset of the other.

A service organisation cannot fill those three in for the manufacturer. What it can do is leave them answerable. The material those codes are chosen from is ordinary service evidence: what was inspected, what was measured and with which instrument, what changed when a known-good part was substituted, and what was never tested at all. A record that omits the instrument or the substitution step has removed the evidence without saying so.

Two fallbacks exist because "no patient" is a real answer

For a malfunction report that involves no patient, FDA's instruction is explicit and unglamorous: select code 4582 for the clinical-signs type and 2645 (No Patient Involvement) for the health-impact type. The person completing the form is not expected to invent a patient effect in order to fill two boxes, and a service record that describes an equipment-only fault is not an incomplete record for that reason.

Specificity is a decision the record feeds

FDA tells reporters to code to the lowest, most detailed level available, and notes that a parent code is a set containing its children, so a child code alone is sufficient and no hierarchy level is skipped by using one. More than one code may be chosen from the same set. The number itself carries no meaning beyond identity: codes are two to four digits, assigned sequentially, and FDA states plainly that codes with similar numbers are not necessarily related to each other. There is no rank in the number and no severity in the digit.

That combination — most specific term, possibly several, no meaning in the digits — makes the physical identification the hard part. Nothing in the code set can substitute for knowing which part was actually in the machine.

The fallback component code is a statement about evidence

One component code exists for the case where the failure cannot be tied to a part: 4755, "Part/Component/Sub-assembly Term not Applicable." It is worth reading as what it is. FDA published it because "not linked to a specific part" is a legitimate outcome of an investigation, not a hole in the form. A record that names a part vaguely — "the board area", "the probe end" — does not sit between the specific and the fallback; it pushes a reader toward a specific component that nothing observed supports. What a test can and cannot settle is a subject of its own, and the element dropout note works through the case where a test's printed number is the wrong currency for the decision in front of you.

A code vocabulary and a parts list are two different vocabularies

A code set is generic by design, because it has to cover every device family FDA regulates with a single tree of terms. A parts list is specific by design: part number, revision, and the assembly the part belongs to. Neither one can stand in for the other. The component term can only be chosen accurately after something has established which physical part was in the machine, and the parts list cannot tell a reader what kind of failure occurred. Records that keep one and discard the other are the ones that force a reconstruction later — the console degradation note is a worked case of a symptom leading to a named board, and of how much of that path is lost if only the part number is kept.

Where a part has to be identified before it can be named in any vocabulary, the catalogue side is the place geprobe works — geprobe is an independent third-party supplier and repairer of ultrasound probes and parts, and the probe and part listings are here.

The vocabulary moves underneath the record

The code system is maintained, and its maintenance is published as a list rather than a diff. The most recent entry on FDA's coding-resources page is dated August 17, 2026, and it is large: 43 modified terms in Annex A; two terms added and seven modified in Annex B, with B12 and B13 retired; two added in Annex C; five modified and two retired in Annex D; 28 added and 85 modified in Annex E; four added and 13 modified in Annex F. In the component annex, four terms were added and six modified. The added identifiers are G0403402, G0403403, G0413504 and G04143; the modified ones are G0405211, G04068, G04097, G07001, G07002 and G07003.

That is as much as the page supports. It gives the term identifiers that changed, not the new wording and not the nature of each change. "G07001 was modified in August 2026" is readable off the page; what it now means has to come from the current hierarchy file. The same page records retirements in earlier rounds as well — B08 retired in 2025 and E1026 in 2024 among them — so a term that was correct two years ago can be absent from the current set rather than merely reworded.

Three identifiers can describe one component

Every FDA code also carries a single NCIt concept code, and maps to one corresponding IMDRF code; FDA's instruction page notes that an electronic submission may use FDA codes, NCIt codes or IMDRF codes for the same field. One component can therefore travel through three systems under three different identifiers. A record that stores a single string is legible to exactly one of them.

The box holds one code of each type

The paper route has a limit that is easy to forget when the record is a narrative: reporters should write only the two-to-four-digit FDA code number in each box, and additional codes go into block H10 prefixed with "F10:". This is a format constraint, not a content one — but it is why a handoff framed as evidence survives a form better than a handoff framed as a list of candidate components.

Two ways to write the same line

Same event, same console, two service entries.

Entry A: "Intermittent image dropout in the cardiac preset. Panel control board replaced. Unit returned to service."

Entry B: "Cardiac preset, phased array, cold start. Image dropout observed within 20 minutes of power-on, gone after 30 minutes of continuous use; reproduced twice on this console and not on a second probe; substituting a known-good board removed the symptom for a 40-minute run; the removed board was not tested and its history is unknown; the probe passed its element check; the fault was not observed again before handover."

The difference is not length. Entry B can be coded by someone who was never at the bench: a problem code from the behaviour, a component code from the part actually substituted, and a finding for the investigation annex, each with the observation that supports it. Entry A can be coded too, and the code chosen for it will be the code that names whatever the reader assumes about the board — which is exactly the assumption the record was supposed to carry or withhold.

Entry B also withholds three things on purpose, and the withholding is the harder half of the job. It does not name a cause, because a substitution removes a symptom rather than proving a mechanism. It does not choose a conclusion code, because that annex is not the service side's to complete. And it does not say what will happen next, because a part that was involved in one event carries no information about the next one. It also carries no code number written from memory: a code that appears without the term it came from is an assertion nobody downstream can check, and it becomes unreadable the moment the term set is maintained. The exchange-versus-repair note is where that decision continues, and it starts from the same evidence this record has to preserve.

geprobe is an independent third-party supplier and repairer of probes and parts; it is not affiliated with, authorised by or endorsed by any manufacturer named in a service record, and nothing here is a repair procedure or clinical guidance. No price, turnaround time or repair outcome is quoted or implied anywhere in this note.