Every tool in this series so far assumes I'm there. Khai runs when I run it. The watchers alert me. The dashboard waits for me to look at it. All of it is built around a person at the center who sees the signal and acts. This post is about the hours when I'm not there, because I'm asleep, or with my family, or on a plane, and about the one problem in running a solo practice that you genuinely cannot engineer your way out of.
You can automate the monitoring. You can automate the testing, the reminders, the reports. What you cannot automate is being a doctor at three in the morning. A telehealth practice quietly implies availability, a website is open all night whether or not anyone's behind it, and the gap between what the website implies and what one human can actually provide is where patients get let down, or worse. Coverage is not a feature I added. It's an obligation I had to design around honestly, starting from the fact that I will, predictably and regularly, be unreachable.
The problem you can't engineer away
The instinct of someone who builds software is to treat every problem as a software problem, and for most of this series that instinct has been right. Coverage is where it stops being right. There is no code I can write that makes me available when I'm asleep, and the failure mode I care about most is a patient who needs something during the hours I'm not there and has no idea what to do about it.
So the honest question is not "how do I make sure I'm never unavailable," which has no answer for a practice of one. It's "what happens to a patient who reaches out when I'm not there." That reframing is the whole post. Once you accept that your own absence is a certainty rather than a failure, coverage becomes a design problem with an actual solution, instead of a guilt you carry around.
Three kinds of after-hours need
The key move is realizing that "a patient needs something after hours" is not one situation, it's three, and they could not be more different in what they require.
The first, and by far the most common, can wait until morning. A question about a refill, a scheduling request, a general "how does this work." The right response is to acknowledge it right away so the patient knows it landed, and then handle it when I'm back. Nobody is harmed by a few hours' wait as long as they know the message was received and roughly when to expect me.
The second needs a human today, but not this exact second, and not necessarily me. Some of it is logistical, the kind of thing office staff can handle with a callback. Some of it is clinical enough that it needs another clinician, which is what a covering arrangement is for. Either way it needs a person, just not an emergency room.
The third is a genuine emergency, and this is the one that matters more than the other two combined, because getting it wrong is catastrophic. A solo telehealth practice is not an emergency service, it cannot be, and the single most important thing the whole coverage design does is make sure a patient in a real crisis is routed immediately to something that can actually help them, which means calling 911 or a crisis line, not waiting on a message to a practice that might not be read for hours. The worst possible outcome is a patient in crisis politely waiting for a reply. Everything about how the practice communicates after hours is built so that never happens.
Sorting an incoming need into those three buckets, and making sure the third one always has an instant, human-independent path to real help, is the entire job.
Setting expectations before they're needed
The kindest and safest tool for all of this turns out not to be a clever piece of software. It's telling patients, in advance, what after-hours actually looks like.
Silence is frightening when you don't know it's coming and reassuring when you do. A patient who knows I don't respond overnight, and knows exactly what to do if something can't wait, is not scared by an unanswered message at 2am, because it's what I told them would happen. A patient who expected an answer and got silence is left to invent an explanation, and the explanations people invent at 2am are not kind ones. So expectations get set up front, plainly: here's when you'll hear back, here's what after-hours looks like, and here, in bold, is what to do in an emergency. That last line is a safety net that works even when every system I built is down and I'm asleep, because it doesn't depend on any of them. It just depends on the patient knowing, before the bad night, to call for real help.
The covering doctor
For the middle category, the thing that needs a human but isn't a 911 situation, there's a covering physician: a colleague who handles what can't wait for me when I'm genuinely unreachable for a stretch. I'm not going to say much about the person, out of respect for their privacy, except that this is the one part of the practice that is not software and never will be. It's a professional relationship and a human arrangement, the old kind of infrastructure, and it's load-bearing.
What software can do is make the handoff clean. A covering doctor walking into a situation cold is only as good as the context they have, and the same patient history that lets me see what we've sent and done, from the last post, is what lets someone covering for me get up to speed quickly. The tooling doesn't replace the human backstop. It makes the human backstop effective.
What happens to a message when I'm away
When I'm away, the practice doesn't go silent, it just stops depending on me for the first response. Two things carry the load.
The first is an automatic reply. Most telehealth patients reach out by email, a few by phone, and the automatic email reply is the workhorse: the moment a message comes in during away hours, the patient gets an immediate acknowledgement that sets expectations and, in bold, says what to do if it can't wait. It costs nothing, it fires every time, and it turns silence into a known quantity, which as the last section said is most of the battle.
The second is a callback. An auto-reply buys time, but a lot of what patients need after hours is not a clinical decision, it's a human one: rescheduling, a logistics question, a bit of reassurance that someone is on it. Those don't require me specifically, they require a person, and they get handled by office staff calling the patient back. The auto-reply says we've got this and here's what to do if it's urgent, and then an actual human closes the loop for the things that need a voice but not a doctor.
That leaves only the genuinely clinical, can't-wait cases for the covering physician, and the true emergencies for 911, which is exactly where I want the hard things to land: with a real clinician, or with real emergency services, never sitting in a queue.
What I'd do differently
I spent real effort on an elaborate in-app away mode, a toggle meant to flip the whole practice into an after-hours posture, and most of what actually provides coverage turned out to be much plainer: an automatic email reply and a person making callbacks. The elaborate version wasn't wrong, it was just a lot of machinery in front of a job that a simple reply and a phone call already do most of.
If I were starting over, I'd design coverage from the patient's actual behavior outward. Start from what patients really do, which is send an email, and work backward: make sure that email gets an instant, expectation-setting reply with an emergency instruction, make sure the human follow-up has a clear owner, and make sure the clinical and emergency cases each have their own path. Match the coverage to how patients actually reach out, not to how the admin panel is organized.
Why this matters for a solo practice
Coverage is where "solo" runs into its hard limit, and the responsible thing is to meet that limit honestly instead of papering over it. You will be unreachable sometimes. That's not a flaw to hide, it's a fact to design around. So you assume your own absence, you make sure the truly urgent always has a path that doesn't depend on you being awake, you set expectations so that silence is understood instead of frightening, and you keep a human backstop for the things that need a person but not an emergency room.
The goal was never to pretend I'm always on, because pretending is exactly how a patient ends up waiting on a practice when they should be calling for help. The goal is to make my absence safe and predictable, so that being a practice of one never quietly becomes a practice of none at the moment someone needs it.
Next time
Coverage is about the hours I'm not there. The next post is about the everyday machinery of the hours I am: how a practice this small keeps up with the constant inbound stream of patient messages across every channel they use, and makes software do the work a front desk would, so that almost nothing needs a dedicated person to catch it. A front desk without a front desk.