Diagnostic Records and Decisions About People

Kirill Shebetov · beforeword · 24 September 2026

Six published cases and a study of how medical records are used

“Cancer” is a written word. “This person has cancer” adds a claim about a particular person. “Surgery is indicated for this person” adds a claim that an intervention is appropriate. These records make different claims, even when they concern the same diagnosis. These lines were constructed to introduce the question examined in this study.

The study follows how medical records are linked to people and used in later decisions. It asks where qualifications are retained, where new evidence is added, and where a claim quietly expands: from a specimen to the whole person, from suspicion to an established diagnosis, or from an order to completed treatment.

A description does not become what it describes. This applies to every word, number, image, and explanation in this study, including “distinction”, “boundary”, “lines”, “description”, “person”, “body”, and “text”. This statement itself is also a record and does not become what it names.

The written form does not itself present what it names. Claiming that a description refers to a particular person, specimen or event adds a claim about a connection. The analysis must show how that connection is introduced and what supports its later use. The word “refers” does not establish the connection merely by appearing in the text. An explanation of the connection also remains a record; it does not become what it describes.

The study begins with six published cases concerning cancer, HIV, SARS-CoV-2, autism, asthma, and ADHD. It then examines diagnostic terms, device data, written reports, orders, records of treatment provided, and later uses of a diagnosis. The analysis connects with ten beforeword studies and a separate study of Horizon. Constructed examples are identified as such; a published case account is not presented as the complete medical record.

1 Cancer: the result was read, but whose result was it?

A report issued on 11 March 2015 by New Zealand's Health and Disability Commissioner (HDC) describes an error involving two biopsies. The patient is designated Patient Y. [29]

One record may describe the material examined; another may identify whose material it is. Repeating the diagnostic term does not add a person’s identity, a specimen label, or a record of the specimen’s origin. If a report is used for a particular person, the next record needs to show the link between the person, the material and the document, together with the grounds on which that link is used.

That link matters to the next decision. Consider this constructed statement: “the result was read correctly, so it can be used for this patient”. This is not a quotation from the report. It skips the question of whose result it is. Correct interpretation of the material has acquired an additional meaning: correct application to a particular person.

The analysis draws on the published report. The original surgical order is not reproduced here, so its wording and all the grounds for the decision cannot be reconstructed from the table. The investigation describes an error in attribution; a summary of it does not supply the complete sequence of clinical documents.

Such links are how a record's role in a decision can be examined. To compress the account into “the letters performed the operation” would omit the people who used the result and the way the decision was reached. The task is to retain those steps and identify where separate support is needed.

2 HIV: a positive test and a diagnostic conclusion

Kaewpoowat et al. (2025) describe a pregnant patient whose screening test “was reactive” and whose supplemental antibody test “was positive for HIV-1”. HIV RNA was not detected in the testing described. They report an HIV-1 diagnosis, treatment recommended during clarification, and two days of medication. Following other tests and observation, the initial results were judged false positive. This analysis uses the published account, not the full patient record. [30, Case Presentation]

Both quoted phrases concern particular test results. “A diagnosis has been established” makes a further claim by linking results to a diagnostic conclusion. That requires rules for considering the results together and information available at the time of the decision. The words “positive” and “established” do not do that work themselves.

A subsequent prescription introduces another question: what should be done while uncertainty remains? A decision may take account of both the available results and the risks of waiting. A later revision cannot automatically establish that the earlier prescription was unjustified. It must be assessed against the grounds available then.

Two constructed statements illustrate opposite expansions. “Medication has been prescribed, so the diagnosis is final” turns a decision to act into a guarantee of the diagnosis. “The diagnosis was revised, so all treatment was mistaken” judges the earlier decision solely by the later result. Neither statement is attributed to anyone in the case. Each needs support that the wording alone cannot supply.

A revision may be well supported. The next document must still distinguish the earlier conclusion, the new findings and the revised conclusion. “Corrected” alone does not show what changed.

HIV and AIDS are also not interchangeable: a positive HIV test does not establish an AIDS diagnosis. [36] This case provides no grounds for declaring all tests erroneous or applying its outcome to another person.

3 COVID-19: isolation before and after a result

COVID-19 names the disease; SARS-CoV-2 names the virus that causes it. [35] The timing of the result and the grounds for each decision matter in this case.

In the PSNet account (2021), fever and an oxygen requirement developed during a planned caesarean section. Mother and baby were isolated before results, “due to suspicion”. After a positive test, she was “assumed to be positive”: 14-day home isolation was recommended and contact tracing began. Later results were negative; an investigation linked the initial result to contamination from an adjacent positive specimen. Retesting the remaining material was also negative. [31, The Case]

The quoted phrases refer to different moments. A summary saying “the positive result caused all the isolation” would erase the decision taken before that result. This is a constructed example of an inaccurate restatement, not a quotation from the case.

The sequence changes what can be claimed about the causes. The first task is to establish what was known at each decision. Only then can the result's role be discussed. It cannot explain an earlier action merely because both appear in the same account.

A test result, a conclusion about infection, an assessment of transmission risk and advice to isolate are not four names for one claim. Moving to a recommendation introduces the conditions under which it applies. A positive entry does not contain all those conditions.

Revision creates a separate task: tracing subsequent uses of the earlier result. The constructed statements “the result was corrected in the laboratory” and “every document using it was corrected” require different support. The first does not guarantee the second. This is a question about the history of correction, not a claim about what happened to every record in this case.

The available material is a published account, not the complete set of original documents. It allows the sequence of decisions to be examined, but does not justify adding every imaginable harm or attributing an experience to someone. A report of changes to care, a person's words and the researcher's explanation must remain distinct.

4 Autism: a judgement about a person takes the place of a response

An HDC decision dated 26 May 2025 concerns a complaint by Mr A. The organisation's response linked autism to a claim that he could never accept being wrong on most matters: “will not ever be able to accept”. The Commissioner criticised character judgements used to avoid addressing the complaint and found a breach of the right to respect. This concerns a complaint response, not a diagnostic reassessment. [32, paragraphs 20, 35, 74–77]

The addition can be identified precisely. A diagnostic designation is joined to a claim about something the person will never be able to do. The claim extends beyond the complaint at hand: it places no limit on the future, although its scope is qualified by “on most matters”.

The organisation’s response places a diagnosis beside a judgement about the person’s ability to accept being wrong. The word “autism” does not supply an answer to the complaint itself: that requires the circumstances described in the complaint and the material relevant to them. Diagnostic information may be relevant when arranging support; the addition examined here is the judgement about the person.

The transition to examine occurs when a characterisation of the person takes the place of that examination. A statement about their ability to accept being wrong cannot establish what happened in the interaction concerned. Even a well-supported diagnosis does not explain every later disagreement in advance.

This is why the case belongs alongside test errors. Identifying an expansion of a claim does not require the diagnosis to be disproved first. The question concerns what the next text adds to it and how that addition is used in responding to the person.

The HDC decision concerns the complaint it examined. It contains no account of how all organisations treat everyone with autism. This analysis keeps the same scope and does not expand one case into a general judgement.

5 Asthma: an earlier explanation meets new investigations

Huang et al. describe a 16-year-old treated with budesonide and formoterol after an asthma diagnosis. Cough and wheezing worsened over a year. At transfer she remained “with a diagnosis of severe asthma”. CT and bronchoscopy found a mass obstructing about 90% of the tracheal lumen; pathology identified a schwannoma. Cough and breathlessness resolved after endoscopic removal. No tumour recurrence was recorded over 18 months. [33, Case Presentation; Outcome and Follow-up]

The account sets an earlier diagnosis alongside new findings. What supports retaining the explanation, and what calls for its revision? Repeating the term records no new grounds. A new investigation can change the conclusion if the relevance of its findings to the earlier question is shown.

It would be premature to attribute every decision to unquestioning belief in a diagnosis. The publication does not present material that assigns the delay solely to the earlier diagnosis. The narrower question here is whether the earlier term took the place of examining new observations. A claim about a particular decision requires the material relevant to that decision.

Changing the name does not finish the analysis either. “Schwannoma” is not the removed mass. An image, a bronchoscopic view and a report may support a new explanation while remaining different materials. “90%” reports an estimate; it does not transfer a section of the trachea into the text. A better-supported description can guide a decision without becoming what it describes.

This published case describes one patient; it contains no material about other people's diagnoses. It raises a narrower question: how can later findings be considered without turning the initial term into an answer to everything that follows? The same question remains relevant when a reassessment supports the earlier diagnosis.

6 ADHD: what follow-up can support

In a 2024 report, a man was diagnosed with “attention deficit disorder” and prescribed a stimulant. A later sleep study identified obstructive sleep apnoea; continuous positive airway pressure (CPAP) treatment began. A clinician subsequently recommended stopping the stimulant because of adverse effects. At the last visit, the authors reported no return of attention difficulties with continued CPAP. The original neuropsychological tests were unavailable to them; repeat testing was not performed. [34, Report of Case; Discussion]

The diagnostic term is retained as it appears in the publication. The later account of attention does not provide full access to the original assessment. It cannot reconstruct the wording of the early report or automatically establish that the original diagnosis was mistaken.

A later text may add new grounds for continuing to use a diagnosis. The earlier record remains the earlier record. If the later report does not contain a full explanation of all earlier difficulties, that explanation cannot be written into it afterwards. Any later use of the original term must also take the later information into account.

Two expansions are constructed for comparison: “ADHD always means a sleep disorder” and “once a diagnosis has been made, it already explains all future behaviour”. The first extends one history to an entire diagnostic category. The second turns a record of a particular assessment into an unlimited explanation of a person. Neither statement comes from anyone involved in the case; both require more support than is presented here.

How the later conclusion was reached also matters. The publication reports that the earlier difficulties did not return; repeat testing was not performed. That report does not become a repeat-test result. If a conclusion specifically requires repeat testing, no such result is present in the published case.

This section does not present the original patient record and does not claim an error in it. It examines only what a later text adds to the published history. It contains no claim that the disorder is imaginary and no recommendation to change another person’s treatment.

7 Six different transitions

These cases were selected for the transitions they allow the study to examine. They do not form a sample for estimating error rates. The table connects each case to its particular question.

The material also differs in what is available for comparison. In some cases, a source describes an error; elsewhere, a disputed phrase is presented, or follow-up limits the conclusions that can be drawn. Constructed examples illustrate possible expansions in a restatement. They do not become additional clinical cases.

The shared subject is what happens next with a record. A result is linked to a person; a report is used to select an action; a diagnosis enters a complaint response; a new entry is compared with an old one. Each transition allows the addition and its grounds to be identified. New content need not be mistaken. If it is justified, the analysis must account for that.

A record may be necessary for a decision. Using it can have consequences; the description still does not become the person. Tracing those consequences requires specific links between documents, decisions and reports that actions were carried out. A disease name alone does not supply them.

The following sections examine those links separately: from an image and a number to a diagnosis, an order, a device command and an account of the outcome of treatment.

8 From a report to a decision

The first case distinguished the interpretation of a specimen from the question of whose result it was. Two constructed statements take that distinction further:

“A malignant tumour has been identified in the examined material.”

“A malignant tumour has been identified in this person.”

Both statements retain the same level of certainty: “has been identified”. The second adds a claim linking the finding to a particular person. Specimen labels, records of where the material came from, and the grounds for the conclusion matter to that link. Repeating the tumour's name does not establish it. These lines are not quotations from the HDC report.

A further statement, “surgery is indicated for this person”, would add a claim that the intervention is appropriate for that person. “Surgery was performed” would report that it was carried out. Examining these claims requires different documents. Mentioning a diagnosis does not supply an operation record missing from the published account.

This study uses conflation for a transition in which one record is retold as another claim without a stated basis for the change: suspicion is retold as an established diagnosis, a specimen finding as an account of the whole person, or an order as a completed procedure. Calling the transition “conflation” adds a label; the analysis still has to present both records and specify what changed.

A new conclusion may rest on further investigations. A change in wording then calls for examination of that new evidence. Linking a diagnosis to a person does not, by itself, identify the record with the person. Referring to someone and being that person are different. These terms remain words here; what needs examination is their particular use.

9 A name and a treatment choice

Nickel et al. (2018) studied 550 adults with no history of thyroid cancer using three hypothetical scenarios in randomised order. The labels changed; treatment options and risks stayed constant. The table reports choices of total thyroidectomy, not operations performed. [14, Methods; Table 2]

The labels above retain the original English wording. At first presentation, differences in choosing any surgery were not statistically significant; order effects were found. This was not a study of NIFTP. [14, Table 3; Treatment Choice]

The material has four distinct levels: the scenario's wording, the participant's response, the summary table, and the authors' conclusion. A later account that calls the responses “operations performed” changes what is being reported. An account that attributes all differences to a single word without considering presentation order discards a condition of the study.

The same distinction applies to accounts of experience. The name of an anxiety scale, responses to it, and their analysis do not provide a complete account of what each participant experienced. Discussion of a result must retain how it was obtained. A record of an experience does not become the experience.

10 A new name and the conditions for using it

Nikiforov et al. (2016) proposed NIFTP for a selected group of noninvasive thyroid tumours previously classified as carcinoma. All 109 patients in the surgically treated group were alive without evidence of disease at last follow-up: 67 had lobectomy alone; none received radioactive iodine. Follow-up was 10–26 years, median 13. NIFTP means noninvasive follicular thyroid neoplasm with papillary-like nuclear features. These are outcomes after treatment. [11, Abstract; Results]

Those data do not establish outcomes for untreated tumours. Nor does renaming mean that a tumour disappeared when a word was removed from its name. The question concerns the category's criteria and the grounds for revising them.

The American Thyroid Association links the new term to reducing overtreatment. Its patient page requires examination of the removed tumour against diagnostic criteria before NIFTP is established. The management recommendations concern that established category; a new name alone is no basis for stopping treatment under a different or unresolved diagnosis. [13, Why the Name Change; How Is NIFTP Diagnosed]

The earlier and later names may be associated with different decisions. The analysis therefore needs to trace why the category changed and which recommendations are linked to it. Such a change can be substantive; it cannot be reduced to an arbitrary exchange of letters. Neither name, however, becomes the tumour.

The same abbreviation with different criteria

In 2018, Nikiforov et al. proposed replacing one NIFTP criterion. The source contrasts these formulations:

less than 1% papillae

no well-formed papillae

The proposal followed a review of later publications and difficulties in applying the criterion. These two excerpts do not constitute a complete diagnostic protocol. [12, Discussion]

The ATA page used here retains the threshold below 1%. References to the two documents cannot therefore be merged into one unchanged instruction. A particular case requires identification of the criteria's version and consideration of the full criteria; this study does not establish any clinic's protocol. [12; 13, What Is NIFTP]

These documents use the same abbreviation, NIFTP, with different criteria. “According to NIFTP criteria” leaves the version unspecified. Naming the version allows the rules to be compared; repeating the abbreviation does not.

11 What remains text

cancer

c a n c e r

In the first line, six letters appear together; in the second, spaces have been added between them. These word examples are adapted for the English edition. Reading the two lines as the same medical term is not an examination of a person. Growth, damage, or spread described using that term do not become properties of its letters. Technical use grants no exception to “growth”, “cells”, or “body”.

A document without the diagnostic term does not contain the sentence “there is no disease”. A document containing the term does not contain the sentence “this disease is present in any person one might choose”. A conclusion about a particular case needs records of its grounds and of the link to that case. Whatever conclusion is later written, the word “cancer” remains the written word “cancer”: it does not become the disease, and this sentence remains another record.

A link does not turn a record into a person

“This description refers to a person” differs from “this description is the person”. The first claim requires grounds for making the connection. The second puts the description in place of what is described. Even a very detailed medical chart does not contain the person: its “completeness” depends on what information the chosen task requires.

h u m a n l i f e

Spacing draws attention to the written form. It does not present a human life. “The diagnosis now determines the person's entire life” adds “entire life” and “determines”. Those additions require examination: what exactly is said to be determined, and on what grounds? Any later statement that the sentence is warranted or unwarranted is another record and needs its own grounds.

The assessment is written too

“Useful” adds an assessment. The word does not contain a criterion of usefulness, an account of what was used and how, or a report of the result. Those details can be supplied and examined separately. They change the grounds for the assessment; the record does not become the use or result it describes.

The same applies to “true”, “confirmed”, “useless”, and “false”. Writing the word does not establish the assessment.

“Boundary” is written too

“Marks made of lines” is another description. “Letters”, “signs”, “symbols”, “lines”, and “text” enter this study as written words. None becomes a final explanation merely by naming a feature of writing.

“Text is text” is itself writing. “Distinction” and “boundary” are written words too; this explanation is also part of the text. Neither a description nor an assessment of it becomes what it names.

12 The same diagnosis read in different ways

Every row in the following table assumes the same wording: “papillary thyroid cancer”. The table was constructed for this study; it does not report the thoughts of particular patients or professionals.

Identical wording presents the same sequence of letters. “They read it the same way” and “they read it differently” are separate claims; comparing particular people requires their responses or other material relevant to that comparison.

“The stage needs clarification”, “the options need discussion”, and “my life is over” are constructed continuations. They add different content. The last phrase must not be attributed to someone merely because the subject is serious. If it has actually been recorded, its context and qualifications must be retained; the researcher's explanation does not become the speaker's experience.

The clinician's recorded explanation, the person's response, and a later statement that “the patient understood” are also distinct. The statement “the patient understood” does not itself reproduce the clinician’s explanation and the person’s response, or show what supports that conclusion. Discussion of understanding requires the relevant material.

13 Types of record across diagnosis and treatment

“All records” covers the types of record from an account of symptoms and a specimen label to treatment monitoring and a later account of events. No type is excluded in advance. It does not mean that every medical document has been read for this study.

The table maps possible forms of record. It does not prescribe a sequence for every disease: stages may recur, run in parallel or be absent. Examining a particular case requires the documents and data themselves, not just the names in the table.

New types of data extend this map without granting words an exception. “Neural network”, “genome”, “physics”, “electronic signature” and “independent review” remain written names. The corresponding files and procedures may add significant information. Mentioning them is not enough to establish that this information has been obtained.

14 Describing a device and its operation

A verbal account of an image, a test result or a treatment choice requires words. This applies to a paper, a clinician's explanation, a manufacturer's instructions and this study. Terms, formulae and references can make a description more precise, but cannot turn it into what it describes.

A device does not need a natural-language sentence at every processing stage. DICOM distinguishes pixel data, numbers, codes, image references and coordinates. Software can transform data; a verbal explanation gives a separate account of that transformation. [3–5]

In the Seg Pro V3 materials published by the FDA, the manufacturer describes automatic contouring of organs at risk from transferred DICOM images. Viewing and editing require a compatible planning system. The device is not intended for lesion detection or independent clinical decisions; primary interpretation of its output requires specialist review. This describes an automated stage, not evidence of autonomous treatment. [10, Device Description; Intended Use / Indications for Use]

A computation may be performed without a separate verbal comment. Establishing how a result was obtained and which decisions use it requires information about the computation, the software, its conditions and its later use.

The constructed phrase “the device saw the cancer by itself” combines several possible claims: the device recorded data, software processed them, a result was produced, and a medical interpretation was proposed. The documents must allow these steps to be distinguished. “By itself” identifies neither the software, its version nor its scope of use.

The claim “it works” needs information about the result. A written report about the device’s operation is not the operation itself.

15 An image and its verbal description

A medical image is not a sentence made of letters. DICOM describes pixel data as a stream of samples, with separate attributes defining their representation. In its two monochrome modes, the minimum value after the relevant transformations is displayed as white or black. The sample value and the display rule differ; this does not make the image arbitrary. [3, C.7.6.3.1]

A region may be marked, a measurement made or an annotation added. The words “lesion”, “contour” and “size” in an explanation do not become that region. A displayed contour adds graphical information. The label “tumour” calls for an explanation of why the marked area is interpreted in that way.

A constructed example:

Description: “A mass has been identified. Further investigation is needed to determine its nature.”

Restatement: “The image has already proved that it is cancer.”

The restatement adds “proved” and “cancer”, removing the need for further investigation. The change in the claim can be located precisely. These two lines cannot determine which diagnosis later investigations will support. A new report needs assessment on its own grounds; it cannot change what the first version said.

A constructed example in the opposite direction turns “no signs were detected in this image” into “there is no disease anywhere”. The restatement loses the image's coverage, the interpretive criterion and the conditions of detection. Repeating a negative statement does not support a broader conclusion.

When an image is added to the analysis, its display, the marked region and the commentary need to be examined. The phrase “the image is shown” cannot substitute for showing it.

16 A numerical result and its diagnostic interpretation

A number may look more definite than a verbal explanation. Without the quantity's name, unit, time and conditions of comparison, it does not give a complete diagnostic claim. FHIR Observation provides separate elements for value, time, specimen, method, device, interpretation and reference range. The format distinguishes results from clinical diagnoses while allowing links between them. Not every element is mandatory in every implementation. [6, §§ 10.1.1–10.1.4]

In the NCI's account of diagnosis, abnormal laboratory results do not establish cancer on their own: they are considered alongside other information and investigations. This does not make abnormalities insignificant or settle the meaning of an individual test without its context. [1, Lab tests used to diagnose cancer]

A missing value, zero and a negative result must be distinguished. DICOM allows an empty measured-value sequence with a reason: the value is unknown or absent, or measurement or calculation failed. An empty record cannot silently be read as “measured and equal to zero”. [5, C.18.1] This distinguishes records; it does not diagnose a particular patient.

The constructed pair “the result is within the reference range” and “the person is completely healthy” changes the scope of the claim. The second statement turns a single test result into a conclusion about the person’s overall health. It needs grounds that the first line does not supply on its own.

An accurate record of a result does not become the measurement or what was measured; a written unit does not become the measured quantity. Information about calibration and quality control can help assess the result. It also needs scrutiny: the words “accuracy” and “control” do not substantiate themselves.

17 Linking a specimen and a report to a person

The NCI describes pathology reports as containing patient and specimen information, gross and microscopic descriptions, a diagnosis and comments. Comments may include pending tests and diagnostic alternatives. The report informs subsequent planning. This describes a document rather than showing a particular specimen. [2]

A name helps link a document to a person within an established procedure. Checking whose result it is requires identifiers, labels, dates and records of the material's origin. These can support the link. Neither matching names nor an explanation of the link turns the document into the person.

“Specimen” and “cells” also remain words. A prepared specimen or an image supplies further material for examination. The phrase “the specimen was examined” reports work without performing it: a past-tense verb does not remove that distinction.

Examining material may support a diagnosis. Demanding to find the complete diagnostic sentence inside the specimen disproves nothing. Yet “it is in the specimen” cannot replace the classification criteria, the examination conditions or the limits of the conclusion. Later uses of the report need to specify which material was examined and which question it answers.

A separate step occurs when a finding about material is read as a complete description of a person. A microscopic description and a diagnostic term do not contain that person's relationships, plans or every later event. A subsequent text about these matters adds claims. They were not contained in the diagnostic word.

18 How a record contributes to consequences

Showing how letters affect a life requires tracing the decisions and actions between a record and its consequences. A diagnosis may inform the next step; an order may enter a system that carries it out; a result may support a further decision. “A word changed a life” does not show which links operated in a particular case.

FHIR ServiceRequest distinguishes a proposal, a plan and an order for a service. A request may link to later results and procedure records. The standard describes intention, requests and links to later results separately; a statement that a procedure was carried out in a particular treatment history requires the corresponding record from that history. [8, Scope and Usage; intent element]

A constructed example: a report names a category; a separate decision considers the diagnosis, additional conditions and the selected procedure; an order specifies parameters; a log records that the procedure was carried out; a new report describes the condition afterwards. Each record adds something. The record does not become the action it describes.

This allows a record's influence to be examined without assigning letters physiological powers of their own. The questions are who read the record or which system processed it, which rule informed the decision, which settings were used and what was done. The outline itself, and the words “rule”, “decision” and “action”, cannot replace documents about the event or the grounds for linking them.

Administrative use requires the same approach. A diagnostic code, a payment rule, a decision imposing a restriction and a report of its implementation are separate steps. This study does not establish the rules of any particular state or insurer. Repeating a diagnosis neither identifies the applicable rule nor grants authority to act.

Additional documents may provide convincing evidence of the consequences of using a record. The record still does not become a disease or a person.

19 A treatment plan and a session record

The DICOM radiotherapy session record module distinguishes specified and recorded delivered meterset parameters, measured doses and calculated doses. These are three of its fields and their roles. [9, C.8.8.21, Table C.8-57]

A meterset value cannot simply be renamed the dose received by a person: the standard separately describes units and relationships under specified conditions. The listed fields need not all be present or populated in every file. [9]

Checking a particular restatement requires comparing the number with its source. If a planned value is called delivered, where was delivery recorded? If a planning calculation is called a measurement during the session, which field supports that phrase, and why has its role changed? Matching numbers may answer different questions.

Agreement between specified and recorded values does not turn the record into the exposure itself. Logs and quality-control records may give strong grounds for concluding that a procedure was performed. Examining them does not make every instance of delivery doubtful. It helps establish precisely what is recorded as delivered and what supports that record.

The constructed example “the log contains a session record, so all treatment is complete and the outcome has been achieved” adds two conclusions: completion of the entire treatment and achievement of the outcome. Each needs separate grounds. One session record does not supply them automatically; the verb “achieved” does not complete treatment on the page.

20 Prognosis and a person's future

The NCI describes prognosis as an estimate of the likely course that takes several characteristics into account. Statistics from large groups cannot predict an individual's exact future. This uncertainty limits conclusions drawn from the data without making every prediction equally well founded. [15]

The constructed statement “this outcome frequency was found for this group” and the continuation “this is exactly what will happen to me” make different claims. The second changes who the claim concerns, its certainty and its scope: a group description becomes certainty about one person. That step needs a separate justification. The number alone does not provide one.

A stage recorded in a document needs to be read with its staging system and the time to which it refers. In the NCI's account, the original stage refers to diagnosis; information about later changes is added. Retaining the designation does not mean the condition has remained unchanged. [16]

“There is no future” shows neither the future nor its absence. The continuation “this wording tells the person how to feel” would add another claim. If the first phrase is recorded, its exact wording and later additions can be examined. Someone else's explanation must not replace the person's own account.

A diagnostic term, a prognosis and a life story may be connected. The connection does not make them identical: a category does not contain every event, a probability is not an individual outcome, and a description of a person does not become that person.

21 Transferring and correcting records

One diagnosis may recur in a report, a patient record, a discharge summary and a message from another system. Several repetitions do not necessarily mean several independent investigations. Distinguishing a copy from a new conclusion requires knowing where each passage came from. A copy can preserve useful information without adding new grounds of its own.

FHIR DiagnosticReport provides for individual results, textual and coded conclusions, a rendered report and a status. Preliminary, final and corrected reports cannot be read identically without explanation. The standard does not guarantee that a particular export preserves every qualification: checking this requires the exports themselves. [7, Scope and Usage; Resource Content]

A constructed example: the original record says “suspected”, while the transferred list gives only the disease name. Both records, the purposes of the fields and the transfer rules need to be compared. The field may be intended for provisional diagnoses, with the status stored separately. That status must then be included in the check. One word removed from this structure is not enough to establish an error.

A correction requires the old and new versions to be distinguished. If both survive, they can be compared. If the earlier version is unavailable, merely stating that it was corrected does not reveal what it contained. After a diagnosis, code or date changes, a question remains: which documents have already used the old version? “Corrected” does not show whether all subsequent copies have been updated. That must be checked against the records of the particular system.

An AI response that restates a report becomes another step in this sequence. A confident tone, matching terminology and a list of references do not grant it independent authority to prescribe treatment. Checking the response requires the original document, the full answer, the claims it adds and information about its later use. The model's name cannot replace any of these steps.

22 How to examine a particular conclusion

The first question is whether a claim has changed, and on what grounds. Looking for a conflation in every use of a medical record would settle the answer before the documents had been read. A different finding must remain possible: the new conclusion is justified and needs no correction.

Consider a control example constructed for this study. The first report records a suspicion and calls for further investigation. The next document gives the date of a new investigation, its result, the grounds for classification, and an established diagnosis. More has happened than the disappearance of the word “suspected”: new material has been added. The questions are whether it concerns the same case and whether it supports the new conclusion. That conclusion cannot be rejected simply because it did not appear in the first document.

Start with one phrase and the documents relevant to it. Then record:

1. The exact wording of the original passage, its source, date and version.

2. The exact wording of the later claim and where it is used.

3. Negations, conditions, degrees of certainty and time limits retained or omitted.

4. Added material and the grounds for linking it to the same case.

5. The attributed status and related decision, if they are presented.

6. A specific correction, or an explanation of why none is needed.

7. A question that the included material cannot answer.

The examination of the selected passage can end here. Action does not require an endless chain of explanations; the point is to state what the material examined supports. The final explanation remains a record that can itself be examined.

The criterion itself, its applicability and the method of checking it can also be challenged. “Sufficient”, “meets the criteria” and “error” do not close that question. An objection must be considered on its substance, rather than dismissed with “those are words too”.

This process helps establish what material a conclusion concerns, what supports a decision, where an action is reported as having been carried out and what has changed in a new version. It does not, on its own, determine whether treatment should be chosen or stopped.

23 Eight constructed examples

These pairs were constructed for the study. They are neither quotations from medical records nor accounts of actual errors. In each pair, the second line is treated as a restatement of the first without additional support. If a later stage is recorded in another document, that document needs to be presented and examined separately: it is not present in the pair itself.

The corrections concern written claims. They are not clinical orders and may change when further material is added.

From suspicion to an established diagnosis

Record: “The findings are suspicious for malignancy. Further investigation is required.”

Restatement: “Malignancy has been established.”

Change: the requirement for further investigation has disappeared and the claim has become stronger. The pair says nothing about the completion of that investigation.

Correction: “The cited report records a suspicion and the need for further investigation. If a diagnosis was subsequently established, provide the document and date separately.” Neither “disease established” nor “disease excluded” appears in this pair.

From the material presented to the whole body

Record: “No signs were identified in the material presented.”

Restatement: “There is no disease anywhere in the body.”

Change: a negative finding about the material presented has been extended to the whole body.

Correction: “The conclusion concerns the specified material and the conditions of examination.” A broader conclusion needs separate support. The original pair contains no statement about the presence or absence of disease beyond the specified material.

From a missing value to a result

Record: “No value was obtained.”

Restatement: “The result is negative.”

Change: the absence of a result has been read as a result of a particular kind.

Correction: retain the statement that the value is missing and give the reason, if one is recorded. The reason cannot be guessed from an empty field; zero, a missing value and a negative result are not synonyms.

From an order to a completed procedure

Record: “The procedure has been ordered.”

Restatement: “The procedure has been performed.”

Change: the second line reports that the procedure was carried out, although the first records only an order.

Correction: “The available document records an order. A report that it was carried out requires the relevant material.” The absence of that material from the pair does not mean that the procedure did not take place.

From a planned calculation to a measurement during a session

Record: “The plan gives a calculated dose value.”

Restatement: “This dose value was measured during the session.”

Change: a calculation in the plan has been read as a report of measurement during delivery. The pair gives no grounds for that report.

Correction: retain the reference to the plan, model and conditions, and present the record of measurement during the session separately, if it exists. Calculation may contribute to obtaining a measurement result; the distinction here concerns the particular roles of a planned estimate and a record of a session that took place. [17]

From a questionnaire response to an operation

Record: “The participant selected a surgical option in the questionnaire.”

Restatement: “The participant underwent surgery because of that term.”

Change: the restatement reports an intervention instead of a response and adds a causal claim about the individual.

Correction: “A choice in a hypothetical scenario was recorded.” The pair gives no information about actual treatment or its causes. The correction preserves the subject of the terminology study without attributing a different result to it.

From a diagnosis to an entire life

Record: “The report states a diagnosis.”

Restatement: “Everything important about the person is now known.”

Change: a medical conclusion has been extended to every important aspect of the person, without a criterion for importance.

Correction: “The record concerns a particular diagnostic question.” This does not prescribe how important the diagnosis should be to the person or replace their own account.

From a file check to a clinical conclusion

Record: “The file conforms to the specified format.”

Restatement: “The diagnosis is correct and the treatment is appropriate.”

Change: the result of a structural check has been extended to two medical claims.

Correction: name the file properties checked and examine the grounds for the diagnosis and treatment separately. Technical correctness may be necessary for processing a file, but it does not cover every condition needed for a clinical conclusion.

24 Sources and scope

This study examines published material and concepts; it does not conduct a new clinical trial, patient examination or hospital audit. Its bibliography contains 36 sources: 25 external medical, technical, metrological and investigative materials, ten works from beforeword’s Research section, and a separate work on the Horizon materials.

The six cases were selected to examine different transitions, rather than to estimate the prevalence of errors. Three appear in clinical articles, one in PSNet and two in HDC decisions. The material available consists of published accounts and the excerpts they contain, rather than complete patient records. The sources’ labels for people have been retained; no additional personal information was sought.

The experiment using different names in hypothetical scenarios and the NIFTP studies are considered separately. A questionnaire response, an investigation, a clinical account and a revision of a classification address different questions. Their findings cannot be combined into one patient’s history or used interchangeably.

Medical publications are read as dated reports with their own methods and limitations. Standards describe formats and the distinctions they provide for: a field’s presence in a standard does not mean it is populated in every actual document. A manufacturer’s description in FDA materials does not replace an independent study of every instance of use.

The analysis identifies the excerpt or paraphrase, source and version, the author’s explanation, the scope of the claim and the limits of the available material. Translations are kept separate from original quotations, and constructed examples from published cases. A proposed correction is not presented as a change already made within a medical system. The Russian and English editions retain the same cases, arguments and reference numbers; constructed examples are adapted to the language of each edition.

25 Connections with beforeword research

The following map identifies ways of comparing records used from ten works in the Research section and the separate analysis of Horizon. It uses the main Russian pages in the website edition saved on 23 September 2026 and a Horizon manuscript dated the same day; it does not claim to have rerun every software check or read every archived appendix. These are methodological connections, not independent confirmation of the works by one another: formal results are not transferred to medicine, and beforeword’s proposed approach is not presented as an existing clinical or legal standard.

Record Boundaries

The initial record and a later statement are compared as separate versions: a correction is not inserted retrospectively into the original report. New grounds may change the conclusion while retaining their own date and content; if the initial version is unavailable, it cannot be reconstructed by guesswork. [18]

“Did You Understand Me?”

The answer “yes”, a repeated diagnosis, a signature and the phrase “the patient has been informed” are shown separately from the sentence “the patient understood”. None of the first records contains that later sentence or presents the understanding attributed to the person. A claim about a particular person’s understanding needs grounds relevant to that claim. [19]

“How You Look from the Outside” Comes Later

An image, a selected region, a measurement and a conclusion are examined separately. Grounds for reading a fragment do not automatically extend to the whole person and their future; a claim about the image’s origin also needs grounds for that connection. [20]

Reading as Enclosure

Several copies of a diagnosis are not several independent investigations. Repetition records repetition of the entry; if a new investigation supports or changes the diagnosis, its material, date and grounds are presented separately. [21]

No Authority in the Output

An order and its verbatim reproduction in an AI response need separate grounds for their use: provenance, a connection to the particular case and authority to act. Making an entry may perform an action defined by a system or change a document’s status, but the conditions for that action cannot be recovered from “signed”, “authorised” or “confirmed”. [22]

Evaluation Does Not Leave the Record

Checking a file’s format does not establish that a diagnosis is correct or a treatment appropriate: the properties checked and the scope of the check must be stated. The same applies to a quotation matching its source and to evaluations in this manuscript. “Criterion”, “useful” and “verified” remain written words and do not establish their own validity; “unjustified” does not replace a substantive objection. [23]

Mathematical Records

Units, thresholds, probabilities and calculated values are considered together with their conditions and rules of application. If a record is absent, the collection contains no corresponding record; a statement of absence everywhere would be a further claim requiring its own grounds. Successful use of a formula does not make it what it denotes; the results of phases 0–7 retain the scope of their declared constructions. [24]

FULL STOP

The initial record, a later reading, the stated grounds and a correction are presented as separate records even when the later conclusion is well supported. If the conclusion’s words are absent from the initial passage, that shows only their absence from that passage; the sentence “the conclusion is false” does not appear by itself. Code can participate in a device’s operation without becoming the procedure performed. [25]

Inscription and Reading

Identical diagnostic wording is examined in its different uses: as a clinical conclusion, an administrative code, a message to a person and part of that person’s own account. Matching wording or answers does not establish identical understanding, and a new explanation remains another passage to examine. [26]

The Written “I”

The explanatory sentence is examined for its negations, expressions of possession, verbs and attributed states. A diagnosis alone therefore does not warrant assigning an experience to a patient, nor does the author’s explanation replace that experience; a statement that letters are absent from the body is not an examination of the body. [27]

Text in Place of Text and the Horizon materials (title translated)

The analysis retains the original passage, the connection between claims, the condition and the proposed correction as it follows a claim into later use. For medical records, this means retaining negation, degree of certainty, date and the document’s role, so that a recommendation is not turned into a report that the recommended action was carried out, or a general result into an individual consequence. [28]

26 When this analysis needs to change

An inaccurate quotation, missing context, a different document version, or a misread field can change a conclusion. The specific passage must then be corrected. The word “boundary” does not protect the text from such an objection.

An apparent conflation may turn out to be unsupported. For example, another field may preserve the status of suspicion, a new diagnosis may rely on a further investigation, or an applicable rule may define the authority to use a record. Those details must be considered. This collection records only the documents included here; if the analysis relies on another document, that document needs to be presented separately.

A causal claim also needs its own support. Choosing surgery in a hypothetical scenario is not a report that surgery took place; a new classification does not describe every patient's life; a log format does not show how a particular clinic operates. The importance of the subject cannot supply missing documents.

“Those are words too” does not answer a substantive objection. Using the phrase to dismiss every addition would stop the analysis from considering the very grounds it asks to examine.

This work does not present a clinical trial of the method, a measurement of its public impact, or proof of novelty. It presents selected materials, comparisons and proposed corrections. The sentence “a record does not become what it describes” contains no order for a particular person and does not replace the medical grounds for starting, changing or stopping treatment.

27 What to check when a record is used again

The analysis centres on specific questions: whose result it is, what its status means, which decision it supports, what is recorded about the treatment provided, and which claims about the person have been added. Medical sources and technical standards already distinguish many of these questions. This work does not claim that beforeword discovered them.

The beforeword approach examines what happens when a record is used again. Repeating a diagnosis may introduce new content. A correction must identify that addition and its grounds, rather than declare every new statement an error.

Where no error is found, a diagnostic record still does not become a disease. A treatment report does not become the treatment. An account of consequences does not become a person's life. New grounds can change the assessment of a report; that later assessment is another record and does not turn the earlier report into what it describes.

In this conclusion, “diagnosis”, “treatment”, “consequences”, “life”, and “evidence” are written words too. This sentence receives no exemption.

No description becomes a person. This line also remains a record.

Sources and corpus

The links below identify the texts used in preparing this work on 24 September 2026. Descriptions of medical publications are brief; constructed examples are identified in the main text. DICOM section URLs containing current may change their content: the pages read were labeled PS3.3 2026d. The FHIR edition is R5 5.0.0. Dates and edition labels are written here as words and numerals too; they are reproduced as information supplied by the sources.

Medical and technical materials

[1] National Cancer Institute. Tests and Procedures Used to Diagnose Cancer. Sections on laboratory tests, imaging, and biopsy. https://www.cancer.gov/about-cancer/diagnosis-staging/diagnosis

[2] National Cancer Institute. Surgical Pathology Reports. Report structure, comments, and subsequent use. https://www.cancer.gov/about-cancer/diagnosis-staging/diagnosis/pathology-reports-fact-sheet

[3] DICOM PS3.3 2026d. C.7.6.3 Image Pixel Module. Pixel data and rules governing their display. https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_C.7.6.3.html

[4] DICOM PS3.3 2026d. C.17.3.3 Document Content Macro. The distinct content types in a structured report. https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_C.17.3.3.html

[5] DICOM PS3.3 2026d. C.18.1 Numeric Measurement Macro. Values, units, and reasons for a missing value. https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_C.18.html

[6] HL7 FHIR R5 5.0.0. Observation. Values, measurement context, and the resource’s scope. https://hl7.org/fhir/R5/observation.html

[7] HL7 FHIR R5 5.0.0. DiagnosticReport. Results, interpretation, presentation, and report status. https://hl7.org/fhir/R5/diagnosticreport.html

[8] HL7 FHIR R5 5.0.0. ServiceRequest. Proposal, plan, order, and links to subsequent results. https://hl7.org/fhir/R5/servicerequest.html

[9] DICOM PS3.3 2026d. C.8.8.21 RT Beams Session Record Module. Fields in a treatment session record. https://dicom.nema.org/medical/dicom/current/output/chtml/part03/sect_c.8.8.21.html

[10] FDA. 510(k) materials for K251306, Seg Pro V3. The manufacturer’s description and conditions of use were used; the summary preparation date is 28 January 2026. https://www.accessdata.fda.gov/cdrh_docs/pdf25/K251306.pdf

[11] Nikiforov YE, Seethala RR, Tallini G, et al. Study of the revised nomenclature for the encapsulated follicular variant of papillary thyroid carcinoma. JAMA Oncology. 2016;2(8):1023–1029. DOI 10.1001/jamaoncol.2016.0386. The sample description, results, and proposed NIFTP designation were used. https://pmc.ncbi.nlm.nih.gov/articles/PMC5539411/

[12] Nikiforov YE, Baloch ZW, Hodak SP, et al. Publication on the revision of the diagnostic criteria for NIFTP. JAMA Oncology. 2018;4(8):1125–1126. DOI 10.1001/jamaoncol.2018.1446. The Discussion and the table of revised criteria were used. https://pmc.ncbi.nlm.nih.gov/articles/PMC6584712/

[13] American Thyroid Association. NIFTP. Patient information on the name, conditions for diagnosis, and subsequent management. The difference between the papillary threshold given there and the 2018 publication is retained in §10. https://www.thyroid.org/niftp/

[14] Nickel B, Barratt A, McGeechan K, et al. Study of changes in terminology for papillary thyroid cancer, responses about treatment, and anxiety ratings. JAMA Otolaryngology–Head & Neck Surgery. 2018;144(10):867–874. DOI 10.1001/jamaoto.2018.1272. Methods, Tables 2–3, and Treatment Choice were used. https://jamanetwork.com/journals/jamaotolaryngology/fullarticle/2697097

[15] National Cancer Institute. Understanding Cancer Prognosis. Estimates of disease course and the limits of applying group statistics to an individual. https://www.cancer.gov/about-cancer/diagnosis-staging/prognosis

[16] National Cancer Institute. Cancer Staging. Staging systems and the time to which the initial stage refers. https://www.cancer.gov/about-cancer/diagnosis-staging/staging

[17] JCGM. International Vocabulary of Metrology. VIM, entry 2.49, Measurement Function. The measurement function and the calculation of an output value from input values. https://jcgm.bipm.org/vim/en/2.49.html

beforeword works

Entries 18–27 draw on the main Russian pages in the website package saved on 23 September 2026; the public URLs identify the corresponding works. Complete identity between all current website pages and the saved package is not claimed here. Entry 28 was read in a separately saved edition. English titles for entries 18–27 follow the English pages in the saved website corpus; this does not change which language edition was used for the analysis.

[18] Shebetov K. Record Boundaries: A Source-Order Method for Testing Written Attributions. beforeword, 2026. https://beforeword.xyz/research/record-boundaries/

[19] Shebetov K. “Did You Understand Me?”: What, Exactly, Does This Written Record Present? beforeword, 2026. https://beforeword.xyz/research/understanding/

[20] Shebetov K. “How You Look from the Outside” Comes Later: How a Particular Record Acquires Generalized Status. beforeword, 2026. https://beforeword.xyz/research/view-from-outside/

[21] Shebetov K. Reading as Enclosure: A Constructive Separation of Self-Return and Evidential Licensing. beforeword, 2026. https://beforeword.xyz/research/reading-as-enclosure/

[22] Shebetov K. No Authority in the Output: A Self-Applying Boundary Between AI Output and Permission to Act. beforeword, 2026. https://beforeword.xyz/research/no-authority-in-the-output/

[23] Shebetov K. Evaluation Does Not Leave the Record: Before the Word “Evaluation”: The Boundary Between Checking a Record and Establishing Its Meaning. beforeword, 2026. https://beforeword.xyz/research/evaluation/

[24] Shebetov K. Mathematical Records. Phases 0–7 series. beforeword, original release 7.2.1, 2026. The titles of the original releases and the scope of their formal results are distinguished from the public reading of their headings; no transfer of those results to medicine is claimed. https://beforeword.xyz/research/mathematics-phases-0-7/

[25] Shebetov K. FULL STOP: The Boundary of a Presented Record. beforeword, 2026. https://beforeword.xyz/research/full-stop/

[26] Shebetov K. Inscription and Reading: The Limits of Written Accounts of Reading. beforeword, 2026. https://beforeword.xyz/research/inscription-and-reading/

[27] Shebetov K. The Written “I”. beforeword, 2026. https://beforeword.xyz/research/written-i/

[28] Shebetov K. Text in Place of Text: Writing, the Connecting Relation, and the Attribution of Status. A Source-Based Analysis of Horizon (title translated). beforeword, 23 September 2026. Main manuscript in the saved package, §§1–70; file beforeword/research/beforeword_Research_RU.html. Publication page (Russian): https://beforeword.xyz/research/horizon/

Published cases and terminology notes

[29] Health and Disability Commissioner. Open disclosure failure following biopsy swap. 12HDC01574. 11 March 2015. Paragraphs 19, 23, and 28. https://www.hdc.org.nz/decisions/search-decisions/2015/12hdc01574/

[30] Kaewpoowat Q, et al. Navigating false positive HIV test results: a case report. ASM Case Reports. 25 March 2025. DOI 10.1128/asmcr.00097-24. Case Presentation. https://journals.asm.org/doi/10.1128/asmcr.00097-24

[31] Martin SA, Kanjilal S, Schiff G. A Postpartum Woman with an Erroneous SARS-CoV-2 Test. AHRQ PSNet. 28 April 2021. The Case. https://psnet.ahrq.gov/web-mm/postpartum-woman-erroneous-sars-cov-2-test

[32] Health and Disability Commissioner. Failure to provide adequate advocacy and disability support services to two consumers. 22HDC03019 and 23HDC01424. 26 May 2025. Paragraphs 20, 35, and 74–77 were used for Mr A. https://www.hdc.org.nz/decisions/search-decisions/2025/22hdc0301923hdc01424/

[33] Huang HR, Li PQ, Wan YX. Primary intratracheal schwannoma misdiagnosed as severe asthma in an adolescent: A case report. World Journal of Clinical Cases. 2021;9(17):4388–4394. DOI 10.12998/wjcc.v9.i17.4388. Case Presentation; Outcome and Follow-up. https://www.wjgnet.com/2307-8960/full/v9/i17/4388.htm

[34] Awadalla TO, Igwe O, Okeafor CU, Attarian HP. Improvement of attention deficit disorder symptoms after treatment of obstructive sleep apnea in an adult: a case report and mini review. Journal of Clinical Sleep Medicine. 2024;20(5):825–827. DOI 10.5664/jcsm.11034. Report of Case; Discussion. https://pmc.ncbi.nlm.nih.gov/articles/PMC11063700/

[35] World Health Organization. Naming the coronavirus disease (COVID-19) and the virus that causes it. https://www.who.int/emergencies/diseases/novel-coronavirus-2019/technical-guidance/naming-the-coronavirus-disease-(covid-2019)-and-the-virus-that-causes-it

[36] Centers for Disease Control and Prevention. Getting Tested for HIV. Positive HIV test results and AIDS are described separately. https://www.cdc.gov/hiv/testing/index.html

The descriptions of the articles in entries 11, 12, and 14 identify their subject matter; they are not presented as the publications’ original titles. Source names, links and bibliography entries are written here too. They identify the material to consult when checking a paraphrase; the claim is then compared with that material and the grounds cited.