"Let Me Know If You Need Anything" Is Not Help: What to Actually Do When Someone You Love Has Cancer

You heard the diagnosis and something in you immediately went looking for a job. You offered to help, and you meant every word of it. What came back was "thanks, I'll let you know." Weeks later you are still waiting to be told what to do, and you have started to suspect that the offer itself was the problem.

It was.

The offer that costs them energy

"Let me know if you need anything" sounds like support and works like a task. It hands the decision back to the person least able to make it. Now they have to take inventory of their own life, choose something, judge whether it is too much to ask, and find the words to ask it. Given four steps, most people take none.

A specific offer removes all four. I'm bringing dinner Thursday, is seven okay? I can drive you Tuesday. I'll collect the prescription on my way home. A named thing at a named time can be declined with one word, and declining it costs nothing.

The rule underneath all of it: make yourself easy to accept.

What you are actually afraid of

The fear is saying the wrong thing. So you say a little less, then less again, and the distance you were trying to close opens up instead.

Some sentences reliably land badly, and they are the ones that arrive fastest. It could be worse. Everything happens for a reason. Stay positive. I know exactly how you feel. However kindly meant, each one edits their experience into something more comfortable for you to be near.

What works is duller and harder. I don't know what to say. This is unfair. I'm not going anywhere. You are not required to improve the situation. You are required to stay in it.

The two jobs nobody assigns

A diagnosis creates two workloads at once.

The first is logistics — appointments, driving, meals, insurance calls, the school run. Families take this one instinctively, and they take it well.

The second is information. Someone has to read the pathology report and understand what the subtype changes. Someone has to notice that a line in a scan from March was never followed up. Someone has to remember what was actually said at the last appointment, and walk into the next one with questions instead of a blank.

Nobody assigns that job, and it lands on families anyway. Caregiving in the US 2025, the national survey from the National Alliance for Caregiving and AARP, found that 56% of people caring for an adult over 50 now perform medical and nursing tasks — and that only 21% of them were ever trained for it. The same report counts 63 million American caregivers, one adult in four, up 20 million in a decade.

Where it lands nowhere, it falls back on the patient — the one person in the room who is exhausted, frightened, and being asked to make decisions in a vocabulary they learned last week.

That is the job you can take, and take deliberately.

Bring questions, not conclusions

Research is where supporters do their most damage and their most good. The difference is entirely in what you bring back.

Conclusions do harm. I read that this treatment has bad outcomes. My colleague's mother did a completely different protocol. You have handed them fear and homework in the same sentence.

Questions do the opposite. The useful ones are unglamorous and specific:

• What is the exact subtype, not just the site?
• What is the stage based on — imaging, pathology, or both?
• What else was on the differential, and what ruled it out?
• What is this plan optimised for, and what would change it?
• Has anyone outside this practice read the whole file?

None of these challenge the oncologist. A good one welcomes them, because they are the questions they would want asked.

The last one carries more weight than it looks. A Johns Hopkins analysis published in BMJ Quality & Safety in 2024 put the annual toll of diagnostic error in the United States at roughly 795,000 people left dead or permanently disabled, and cancers sit in the top three categories of harm alongside vascular events and infections. That is a population-level estimate, not a prediction about anyone's case. What it does establish is that a fresh reading of a whole file is a reasonable thing to want, not an insult to the person who wrote the first one.

If you want to arrive fluent, start with the documents themselves: how to read a pathology report, how to read an MRI report, and what a tumour marker actually tells you. If the plan itself is the question, the first plan is a hypothesis, not a sentence.

Where the file actually lives

The information job is hard for one structural reason: the file does not exist in one place. The pathology report is a PDF in someone's email. The scans are on a disc or a hospital portal. Last year's bloodwork is somewhere else entirely. The only thing connecting them is a tired person's memory.

Healz holds all of it in one chat. Frontier AI reads each report in the context of everything else rather than line by line, cross-checks the case against 1M+ rare cases, and pushes at what the label skipped instead of confirming it. It remembers across conversations, so a value from March is still there in September, sitting next to the thing it explains.

One place, not ten apps. And when you want a human read of the whole picture, you can bring a board-certified doctor into the same chat for a second opinion.

You cannot cure this. You were never going to. What you can do is make sure the person you love is not also their own researcher, archivist and translator — so that when they walk into the next appointment, someone has already read everything and written down the questions.

Show up with the file, not with advice.