My son's medical records in my pocket: how I used AI to never get lost again

1. Introduction: When paper becomes an enemy

My son is 14 years old, a quadriplegic, and ventilated. It's a sentence I write naturally, because it's been our normality for years. But behind that sentence lies an impressive medical history: hospitalizations, medical reports, specialist visits, ventilation parameters, discharge letters, various tests, and neuromotor evaluations. Years of paperwork.

For a long time, I managed everything with a paper folder. Then two. Then a trolley. Before every specialist appointment, the ritual was always the same: open it, look for that 2019 report, hope I hadn't left it at home. Once, in front of the pulmonologist, I couldn't remember the ventilator settings from the previous year. The doctor waited. I leafed through.

The turning point came from an unexpected place: work. I had begun using artificial intelligence tools in my professional field, to organize documents, extract information, and ask questions of large amounts of text. At a certain point, I thought: why not apply the same logic to my son's medical records? Medical reports are text. Often, a lot of text. And I needed exactly what these tools can do: not diagnose, but the ability to "converse" with documents.

The tool that changed everything is called NotebookLM, developed by Google. It's not a generic AI that invents answers: it works exclusively on the documents you upload. It essentially becomes an assistant that knows your child's medical history better than you can remember it.

2. Setting up: prepare the ground before leaving

Document recovery and digitization

The first step was to collect all the existing material. I followed two parallel paths:

• For paper documents: I used one of the many smartphone scanning apps (in my case, Adobe Scan), which saves as PDF and activates OCR (optical character recognition). This is crucial: a simple photo isn't perfectly readable by software, but a PDF with OCR is.

• For documents already digital: I downloaded all available reports from my region's Electronic Health Record (FSE). In Lombardy, you can access the portal and directly download reports, including those for children under your care.

It's worth adding a practical consideration: many hospitals today allow you to download reports directly from their portal. Furthermore, I always ask doctors who prepare documentation to also send me the documents via email. These are small habits that, over time, significantly reduce the amount of paperwork I have to manage.

File Organization

The next step was to choose where to store these files. I use Google Drive, Google's free cloud storage service: all you need is a Google account (the same as Gmail) to get 15 GB of space, accessible from any device (computer, tablet, or smartphone). The advantage over keeping files solely on your phone is that they're always available, even during an appointment, even if you change devices. You can also grant the doctor access to the cloud storage if needed.

Once I created the folder on Drive, I renamed each file with a precise chronological logic:

Year-Month-Day_Healthcare_Facility_Specialization

Example: 2023-10-15_OspedaleBuzzi_Pneumology.pdf

This system ensures automatic sorting in any folder. I then created a structure on Google Drive divided by specialty: Pulmonology, Neurology, Cardiology, Exams, Discharges... This can be varied based on specific needs.

Creating and configuring NotebookLM with a custom role

Once Google Drive was set up, I created a space dedicated to my son on NotebookLM and entered Google Drive as the data source.

At this point, I configured NotebookLM with a customization prompt in the "Objective, style, or role" field. This dramatically changes the quality of the responses. Here's the text I used, which you can copy directly:

“ Act as an assistant in analyzing and organizing clinical documentation.

Help me review, summarize, and compare medical reports, diagnostic tests, and health documents related to my child's medical history.

Support me in monitoring clinical evolution over time.

Highlights changes, trends in clinical parameters and differences between subsequent tests.

Use clear language that is understandable even to non-doctors.

Do not diagnose or replace a doctor. ”

With this configuration, NotebookLM behaves like a clinical monitoring assistant: it already knows how to structure the responses, which elements to underline, and above all it knows what its limits should be (I will return to this later).

Once the upload was complete, I asked the first orientation question:

"Analyze all the documents present and describe the medical history in chronological order."

The result was a synthesis of years of clinical history, with direct references to the original documents. For the first time, I had before my eyes a complete and navigable map.

3. Data Use: What You Can Do Now

This is the most concrete part. Once the system is up and running, here's how I use it in my daily practice.

Before any specialist visit

I ask NotebookLM to prepare a briefing for me:

"Generate a one-page clinical summary to show the physician, focused on the [pulmonary] specialty."

The system produces a summary sheet with the most relevant data, citing the original reports. I can print it or show it directly from my smartphone.

To compare the evolution over time

With chronically ill patients, the key question isn't "how are they doing now" but "how are they doing compared to before." The prompt I use:

"Compare the respiratory parameters of the last three years.

It indicates improvements, worsening and stable values."

To understand technical terms

Medical reports use language I'm often unfamiliar with:

"Analyze this report and explain the out-of-range values to me in a way that a non-physician can understand."

4. Conclusions and important warning

Using NotebookLM to manage my son's medical records has significantly improved the quality of our doctor visits. I no longer arrive unprepared. I no longer waste time leafing through paper while the doctor waits. I have a longitudinal view of his medical history that I wasn't able to have before.

For other caregivers in similar situations, the message I feel like conveying is simple: the clinical complexity of a child with a severe disability requires adequate tools. These exist, are free or nearly free, and don't require advanced IT skills.

Warnings that are not optional

I need to be clear about this, because the stakes are high.

• NotebookLM is not a medical tool. It does not diagnose. It does not replace clinical judgment. Therefore, use its answers as a starting point, not a destination.

• Always verify. Each AI response should be compared with the original document. NotebookLM always cites sources: use them to double-check.

• Pay attention to privacy. Health data is among the most sensitive information in existence. Before uploading any document to any platform, read the privacy policy and carefully assess the risk. Never share access to your notebook with unauthorized persons.

• It's not a replacement for a doctor's clinical memory. The attending physician has an understanding of the clinical picture that no AI can replicate. The tool helps you, as a caregiver, be more prepared and organized.

Technology doesn't solve the complexity of disability. But it can take away some of the bureaucratic and cognitive burden that often weighs on caregivers. This, for me, is worth everything.

---

Support ForAllWe

ForAllWe is an independent project on digital accessibility and inclusive technology. If you find our content useful, you can support us with a voluntary contribution. Your support helps us remain independent and involve people with disabilities as authors and testers.