AI tool made medical records easier for patients to understand

AI tool made medical records easier for patients to understand

Rating
—/5 0 ratings
Follow us on Google News
Health & Lifestyle8 min read

An artificial intelligence tool tested on breast cancer-related medical notes helped participants better understand what they were reading. It did not make diagnoses or recommend treatment: it added explanations of terms, plain-language rewording, key information, questions and possible actions to discuss with a clinician.

The finding matters as patients gain access to more digital documents, which are often written for communication between professionals. The study does not, however, support the conclusion that a chatbot can safely interpret an individual case. Researchers identified potentially harmful errors, and the evaluation did not measure effects on treatment or health outcomes.

In brief: an AI-generated explanation may be a better starting point for a conversation with a doctor. It is not a second medical opinion and should not be used to change medication, interpret a result in isolation or delay seeking medical advice.

What the study tested about AI for medical records

The study, published in Scientific Reports on 5 November 2026, evaluated a tool that augments medical notes for patients. The wording matters: the research did not test AI’s ability to interpret a person’s entire medical history or decide what care they should receive.

The tool worked on clinical notes and produced several layers of explanation: definitions of jargon, simplified wording, questions and answers, extraction of information considered important, and a list of actions. The original note remained available alongside these additions.

The quantitative part of the evaluation involved 200 healthy women, who reviewed eight medical notes in a breast cancer-related context. Those who received augmented versions scored better on questions testing their understanding of the text and reported greater confidence than the control group. The authors also conducted interviews with seven patients with a history of breast cancer.

The study data and design are set out in the Scientific Reports research article. The most cautious conclusion is this: for certain notes, in a particular context, automated explanations made the document easier for the assessed participants to understand.

A simplified note is not the same as clinical interpretation

The distinction may seem technical, but it has direct consequences. A medical note can record findings, symptoms, hypotheses, investigations under way, conditional recommendations and uncertainties. A system that rephrases the text can make that information more accessible. It does not, however, automatically acquire the full context in which the clinician produced it.

For example, phrases such as “correlate clinically”, “possible”, “monitor” or “depending on progress” are not stylistic details. They indicate that information is not a final verdict. If a simplification omits that limitation or makes it more categorical, a clearer text can, paradoxically, become more misleading.

The study authors asked a team of three clinicians to analyse errors. Most were described as minor and occurred when the tool assumed a context that was not explicitly stated in the note. However, there were also errors considered potentially harmful. Errors were more frequent in additions made to real notes than to synthetic ones.

The difference highlights a limitation that is easily overlooked: the quality of a summary depends on the quality, completeness and clarity of the source document. If a note is ambiguous, fragmentary or contains an error, AI may reproduce it in a more fluent—and therefore more convincing—sentence.

Where the real benefit may lie

The plausible benefit of such a tool is not a “digital doctor”, but preparing a patient for a better discussion. Rather than looking up every abbreviation separately or arriving at an appointment without knowing what to ask, patients may be able to identify unclear terms and details needing clarification more quickly.

In the main study, participants particularly preferred the definitions and simplifications. That is a useful signal: more features do not necessarily mean greater clarity. An interface crowded with questions, “actions” and extra explanations may add information, but also confusion.

A separate randomised study involving 300 adults assessed plain-language AI summaries reviewed before appointments. It reported a modest improvement in some aspects of doctor-patient communication and shared decision-making, without evidence that perceived empathy increased. The result supports AI as preparation for a consultation, rather than a replacement for one.

Another study, involving 2,000 US adults, found that AI-simplified versions of radiology reports increased the accuracy of answers about findings and follow-up steps. But the study used vignettes, not a patient’s full path through real-world care. This type of result says something about comprehension, not about the safety of decisions made at home.

What these findings do not yet show

The Scientific Reports study does not show that people receiving augmented notes adhere to treatment better, see a doctor sooner, avoid errors or have better clinical outcomes. Nor does it show that the tool works equally well in complex illnesses, emergencies, psychiatry, for children, or for people with limited health and digital literacy.

The quantitative sample consisted of healthy women, and the subject was breast cancer. The interviews included only seven patients with a history of breast cancer. Some of the notes were synthetic. These factors do not invalidate the findings, but they limit how far they can be generalised to all documents and patient groups.

The sources consulted for this article provide no evidence that an identical tool is integrated into Romania’s Electronic Health Record system or is widely available to Romanian patients. It is also unclear how such a system would perform in Romanian, with local abbreviations, scanned documents or inconsistent wording.

Access to data does not guarantee understanding

The issue is relevant in Europe, too. The EU regulation on the European Health Data Space aims to give people better access to their own electronic health data, including through access services and interoperable formats. The regulation is available on EUR-Lex.

The regulation neither introduces nor validates an AI summariser for every patient. It does, however, make the difference between two problems more visible: being able to open a document and being able to understand it. A portal with more results, letters and tests may increase transparency, but it does not automatically remove the barrier of medical jargon.

For Romania, the useful question is not whether such a product “will replace doctors”, but under what conditions it could become a responsible aid in a patient portal. That would include an easily accessible original version, a visible distinction between medical text and generated content, a clear route for reporting an error, and precise rules for processing sensitive data.

For more information on health data, see also The invisible doctor: continuous AI health monitoring and who controls your body’s data, which offers complementary details on the subject.

Five checks before using an automatically generated medical summary

  • Keep the original document in view. The summary should be a guide, not the only version of the information.
  • Check what was generated automatically. Definitions, questions and action lists should be clearly marked as distinct from the clinical note’s content.
  • Do not change treatment. A simplified explanation does not justify starting, stopping or changing medication without a doctor’s recommendation.
  • Turn uncertainties into questions. It is safer to note, “What does this result mean for me?” than to treat AI wording as a definitive answer.
  • Ask where the data goes. Medical notes contain sensitive data. It matters who processes the document, where data is stored, how long it is kept and who can access it.

The World Health Organization includes validation, privacy protection, individual autonomy and governance among the central conditions for AI in health. These are not box-ticking formalities to address after launch, but requirements that directly affect patient safety.

A tool for conversation, not decision-making

The most promising use described in the available studies is modest and concrete: making a medical document easier to read and helping the patient arrive better prepared for a consultation. That can have real value, particularly when the original text is dense and full of technical terms.

But its usefulness depends on visible limits. The original note must remain the reference point; the automated explanation must be recognised as an explanation, not additional clinical truth; and the doctor or care team remains the place where a result is interpreted in the context of the individual. New, severe or worsening symptoms, or any uncertainty about treatment, require medical assessment—not an AI summary.

Frequently asked questions

Can an AI tool explain my medical results?

It can rephrase jargon and highlight information in a document, but it cannot replace interpretation of a result in the context of medical history, symptoms, a clinical examination and other investigations.

Can I make treatment decisions after an AI summary?

No. An automated summary should not be used to start, stop or change treatment. Discuss any decision with your doctor or the medical team overseeing your care.

Why does it matter to see the original medical note?

The original version allows the source to be checked and may show nuances, conditions or uncertainties that a simplification could omit or rephrase too categorically.

Is such a tool already available in Romania’s health system?

The sources consulted for this article do not confirm implementation in Romania of the tool evaluated in the study, or of an identical service integrated into the Electronic Health Record system.

Sources consulted

Etichete: Breast cancerHealth Datamedical recordsArtificial intelligencemedical notespatientsEuropean Health Data Space
Share this article
COMUNITATEA INFOBUZ

Discuția cititorilor

Ai o opinie sau o întrebare despre acest articol? Spune-ne ce crezi.

0 comentarii

Fii primul care deschide discuția.

Lasă un comentariu

Email address este vizibilă doar administratorilor și nu va fi publicată. Câmpurile marcate cu * sunt obligatorii.

Te rugăm să păstrezi un limbaj civilizat și să discuți la subiect. Comentariile spam sunt filtrate automat.