Gains:
- Ability to produce drafts that explain the test result to the clinician and the patient in an appropriate, accurate and measured language with artificial intelligence
- Ability to maintain the boundary between interpretation and diagnosis and avoid language that would create false reassurance or panic in the patient
- Ability to acquire the habit of transparently stating method, interference and boundary information in report notes.
The most visible thing a laboratory produces is the report. Perfect measurement on the device, accurate verification, meticulous quality control — it all adds up to a report that a human will eventually read. And the report is addressed to two very different readers: on the one hand, the clinician (physician) who will use the result in his clinical judgment, and on the other hand, the patient who, more and more often, sees his own result directly. The same number must be explained to these two readers in very different language. Technical transparency to the clinician regarding method, interference and limits is required; An understandable explanation that does not cause panic or false reassurance to the patient. Artificial intelligence is a powerful assistant in producing rapid and consistent drafts between these two languages — but it is the responsibility of the expert to ensure that the report is accurate, proportionate and maintains the boundary between interpretation and diagnosis.
In this unit you will learn how to adapt the report for two readers; how to provide method, interference and boundary information transparently; and we will cover producing understandable but safe text drafts with AI. Basic principle: AI drafts the text; It is up to the specialist to preserve the diagnostic boundary with clinical accuracy, measured language and interpretation.
Two readers, two languages
Report to the clinician: The physician interprets the result in his or her clinical context; What he needs is transparency that allows him to evaluate the reliability of the number. A good clinician's report includes, where necessary: method/unit used, reference range, note of interference if any (hemolysis, etc.), delta alert, sample quality, and whether the laboratory returns the result with confidence or reservation. The physician asks "How much can I trust this result?" should be able to find an answer to the question from the report.
Explanation to the patient: The patient is often someone with no medical training, and a "high" on the report may panic him or a "normal" may falsely reassure him. The text for the patient is written in plain language, away from the language of fear and certainty, without making a diagnosis, and with the guidance "consult your physician for evaluation". Explaining the result to the patient is not making a diagnosis; This distinction must be preserved.
Size
Report to clinician
Explanation to patient
language
Technical, abbreviated
simple, everyday
Method/interference note
Yes, transparent
Simplified if necessary
Reference range
numeric
explained in context
diagnosis
Konmaz (physician reviews)
Konmaz (referred to the physician)
tone
objective
Calm, reassuring but honest
Transparency: hide, indicate border
The hallmark of a good report is that it does not hide the boundaries of the conclusion. If a result comes from a hemolyzed sample and is given with a note of interference, that note should be included in the report — because knowing this could change the physician's clinical decision. If a test was run using a particular method and that method has a known limitation (e.g. cross-reactivity with a particular drug), this should be stated. Transparency increases trust; A hidden boundary may lead to an incorrect clinical decision in the future.
Caution: The "give the result as it is, do not confuse with notes" approach is dangerous. If an interference or method limit could affect the clinical decision, pointing it out is not confusing, but fulfilling responsibility.
How AI helps in reporting
AI significantly reduces the reporting workload: it can draft the same result in different languages for both clinician and patient, translate a complex multi-abnormal report into a simple summary for the patient, produce repetitive interference notes in a standard and consistent format, reorganize a report in a more readable form. But AI easily falls into two mistakes: (1) dramatic language that unnecessarily frightens or falsely reassures the patient, (2) definitive statements that turn the result into a diagnosis. The specialist's job is to correct these two errors in each draft, ensuring clinical accuracy and ensuring that the necessary transparency notes are in place. Additionally, patient data is always processed anonymized.
Weak prompt / Strong prompt
Weak prompt:
Explain these consequences to the patient so that he understands their seriousness: [values].
This request pushes the AI into the language of fear and the implication of recognition by saying "let him understand the seriousness"; There is no anonymization or referral to a physician. The result is a transgressive text that can throw the patient into unnecessary panic.
Powerful prompt:
Your role: assistant who prepares a patient information DRAFT for the laboratory specialist. Rules: plain Turkish; using fear/panic language; Does not imply a definitive diagnosis; ends with "These results are not a diagnosis, consult your physician for evaluation." Anonymous values (with our laboratory reference):- Vitamin D 12 ng/mL (ref 30-100) — low- TSH 5.8 mIU/L (ref 0.4-4.0) — slightly highTask: explain these two values to the patient in a calm, honest and understandable way; convey what they could mean in "possible" language, without exaggeration; refer to physician. Also write a brief, technical version of the same results for the clinician.
The strong prompt defines the language and boundary, is anonymous, prompts the language of “possible,” and produces separate versions for two readers.
three mini cases
Case 1 — Calm and honest patient text. In one patient, vitamin D is 12 ng/mL (low) and TSH is slightly high. The expert receives a draft in two versions with a powerful prompt. Patient text: "Your vitamin D value appears to be below the reference range and your thyroid value is slightly high. These results are not a diagnosis of disease on their own; there may be many reasons. It would be appropriate to consult your physician for evaluation." The expert confirms this. No panic, no false assurance, the border is protected.
Case 2 — Correcting fear language. In the first draft, AI writes to a patient, "Your thyroid value is seriously impaired, treatment is required urgently." The expert catches this: TSH 5.8 is a slight elevation, the language of "serious" and "urgent" is wrong and scary. He corrects the text as "it seems slightly high, it would be appropriate to have it evaluated by your physician." Lesson: AI can dramatize; Moderate language is the responsibility of the expert.
Case 3 — Transparency note. A potassium result comes from a slightly hemolyzed sample and is given with an interference note. AI omits this note in the initial clinician draft. The expert adds: "Note: there is mild hemolysis in the sample; the potassium value may therefore be slightly high, a new sample is recommended in case of clinical incompatibility." With this note, the physician evaluates the result in the correct context. Lesson: boundary information is not deleted from the report, it is specified.
Copiable prompt templates
PATIENT INFORMATION DRAFT TEMPLATE Your role: assistant who prepares a simple information draft for the patient. Rules: daily Turkish; no language of fear/certainty; making a diagnosis; Use “possible” language; End with "these results are not a diagnosis, consult your physician." Anonymous values (by reference): [list]. Keep the text calm and honest.
CLINICIAN NOTE TEMPLATEYour role: assistant preparing a short, technical report note for the clinician. Provide the following information transparently: method/unit, reference range, note of interference (hemolysis/lipemia/icterus) if any, delta warning, sample quality reservation. Making a diagnosis; Add a note about the reliability of the result. Data: [list].
TWO READERS COMPARISON TEMPLATEProduce two separate drafts for the same anonymous set of results: (1) technical and transparent note for the clinician, (2) simple and calm explanation for the patient. Don't make a diagnosis in both. Specify which information remains only in the clinician version. Results: [list].
LANGUAGE AUDIT TEMPLATECheck the following patient information text: flag for fear/panic language, insinuation of diagnosis, false reassurance, exaggeration, and suggest a more measured alternative. Check for a "consult physician" note. Text: [text].
Common mistakes
- Diagnosing the patient. The resulting statement is not a diagnosis; The patient is directed to the physician.
- Language of fear or false reassurance. The exaggeration of “serious/urgent” or the reassurance of “everything is fine” is out of proportion; Honest and calm language is essential.
- Hiding the interference/method boundary. The limit that may affect the clinical decision is stated in the report.
- Giving the same text to two readers. Clinician and patient require different languages; One text does not fit both.
- Bypassing anonymization. The patient ID is not entered into the tool when generating report text.
Tip: Before sending every patient text, look for three words: "certain," "urgent," "all OK." These words convey either fear or false reassurance; Balance each with “possible,” “must be evaluated by your physician,” “not a diagnosis on its own.”
In summary
The report is the visible output of the laboratory's entire effort and speaks to two different readers—the clinician and the patient—in different languages. Transparency to the clinician regarding method, interference and limits; The patient needs a simple, calm language that does not diagnose and directs the patient to the physician. AI sketches these two languages quickly but easily dramatizes or slips into diagnostic implication; The specialist's job is to maintain clinical accuracy, measured language, and diagnostic boundaries with interpretation. Transparency increases trust: the boundary of the result is stated, not hidden. And all texts are generated with anonymous data.
Application task
Select an anonymised result set and produce both a clinician note and a draft patient description with the "Two Readers Comparison" template. Check and correct patient text for fear/certainty language with the "Language Check" template. For an interfering result, include the appropriate transparency note in the clinician's note and write in one sentence why it is necessary.
checklist
- [ ] I produced separate, appropriate language text for the clinician and the patient.
- [ ] I did not make a diagnosis in the patient text; I referred him to the doctor.
- [ ] I balanced the language of fear, certainty, and false reassurance with “possible/should be considered.”
- [ ] I have clearly stated the interference/method limit in the clinician's note.
- [ ] I produced the report text with anonymous data.
- [ ] I have expertly approved clinical accuracy and measured language.