Notes on Care

Everyone means well. Meaning well is the cheapest thing in healthcare. There is no shortage of people who genuinely, sincerely, achingly want the best for the person in front of them — and who are nonetheless part of a system that loses their labs, calls at the wrong hour, and asks them to explain their own history four times in one week.

So we do not get to treat care as a sentiment we feel. We treat it as a thing we make. Something with edges and materials, that can be done well or badly, that gets better with practice and worse with neglect.

Kindness that doesn’t show up on time is not kindness yet. It’s an intention.

A craft has standards you can name. A follow-up that lands the same day. A message written so that a tired person at 11pm can understand it on the first read. A handoff where nothing falls on the floor. None of these are warm feelings. All of them are what warmth looks like once it has been built.

This is also why we are allowed to be exacting with each other. When you push back on a flow that makes someone repeat themselves, you are not being difficult, you are doing the craft. The person who says “this screen will make a caregiver cry at midnight” is doing quality control.

And like every craft: you get to be proud of it. Not modest, not self-effacing — proud. Go look at a discharge summary you rewrote until a human being could actually read it, and feel the way a carpenter feels about a joint that fits.

The trap, of course, is confusing craft with polish. A beautiful interface that nobody can reach at 3am is a decorative object. We are not making decorative objects.

Make the thing that holds weight.

Before the diagnosis. Before the workflow, the claim, the cohort, the funnel. Before the word “patient,” which is a fine word that has quietly done a lot of damage by turning a person into a role they did not audition for.

There is a specific and very ordinary miracle available to us, which is that a person will tell you almost everything you need to know if you begin by being interested in them. Not in their condition. Them.

In practice this is embarrassingly literal. Start user research with what their week is actually like. Start a support call by asking how they are, and then — this is the hard part — wait for the answer. Start a spec with one real named human, not a persona with a stock photograph and a plausible age range.

The system will always offer you a category to put someone in. Decline, politely, for as long as you can.

Categories are useful. We will build many of them; you cannot run a company on anecdotes. But a category is a compression, and compression throws things away, and the things it throws away are usually the things that were making the person miserable.

So: hold the category loosely, and hold the person tightly. When they disagree, the person wins, and the category is what needs fixing.

This scales worse than the alternative. We know. We’re doing it anyway, because the companies that optimized the other direction are the reason we exist.

You are. There is always a piece of context sitting just outside the frame — the sibling who actually makes the decisions, the pharmacy that closes at four, the shift nurse who invented a workaround so elegant that nobody upstream ever learned the problem existed.

Healthcare is unusually good at punishing confidence. It is a field where the obvious answer is frequently correct and occasionally catastrophic, and where the person with the least institutional power very often has the most accurate picture of what is happening.

The people closest to the pain are rarely the people closest to the microphone. Go get them.

So we go looking. Before we are sure, and especially after we are sure. We ask the aide, the biller, the daughter, the person who quit. We treat “I don’t understand yet” as a completely respectable thing for a senior person to say in a meeting — in fact, it is mostly senior people who need to say it, because everyone else is watching to see if it’s allowed.

Assuming you’re missing something is not the same as being timid. Move fast, ship, decide with incomplete information — all of that. But hold your model of the world with an open hand, and go out of your way to find the thing that would prove you wrong, because that thing is going to find you eventually and it is cheaper to meet it early.

The failure mode here is the meeting where everyone nods. Nodding is not agreement, it is often just fatigue. If a room agrees with you immediately, you have either found something true or you have found a room that has stopped arguing. Check which.

Curiosity, in the end, is just humility with somewhere to go.

Call them back yourself. Send the thing you said you’d send, on the day you said you’d send it. Write out the four steps by hand, in their words, because the standard version assumes a reading level and a calm afternoon that this person does not currently have.

None of this scales. That is not a bug in the value; it is most of the point.

The unscalable act is how you find out what to build. Every good system we have ever shipped started life as somebody doing it manually, badly, for one person, and noticing. If you skip that step you will build something that is efficient at the wrong thing, which is the most expensive kind of software there is.

Do it by hand until it hurts. The place where it hurts is the product.

It is also, separately and without any strategic justification, just the right thing to do. Someone is having one of the worse weeks of their life and you are the part of the week that went okay. That is worth doing even if it never turns into a feature.

This extends to each other. The colleague who is underwater does not need a process improvement, they need you to take one thing off the pile today. Notice who is quiet. Notice who has been carrying the on-call for three weeks. The team is also a thing that requires care, and it will not tell you when it stops getting any.

We are wary of the word “family” here — it has been used to ask people for too much. So let’s say something smaller and more honest: we are people who notice each other. That is a real thing, it costs something, and it is one of the few advantages a company can have that cannot be copied.

Most software in health does the opposite by accident. It shrinks people. It turns a life into a row, a fear into a dropdown, a nurse with twenty years of judgment into a person clicking through a form designed by someone who has never met her.

The test we would like to apply to everything we make: did the person on the other side come away feeling more like themselves, or less?

More like themselves usually means: they understood what was happening. They were not made to feel stupid. They had a choice, and the choice was real. Somebody used their name and it was spelled right. The thing remembered them, so they didn’t have to explain their whole life again from the beginning.

Nobody remembers the interface. They remember whether they felt like a person while they were using it.

This is where the whimsy comes in, and we are unembarrassed about it. A small warmth in a piece of copy. A moment of color where the industry expects beige. An animation that does nothing except suggest that a human being made this and was, for a moment, enjoying themselves.

Health is heavy enough. We are not going to make it heavier by matching its mood. Leave your fingerprints on the work — so that somebody, somewhere, at the end of a bad day, gets the distinct impression that another person was thinking about them.

If we get everything else wrong, let’s get this right. Care is the oldest technology we have. All we are trying to do is help it travel further.