Skip to main content

Practice Operations

When a Patient Sends an After-Hours Text at 11pm

An after-hours text is a promise the practice did not make on purpose. What an internal medicine group has to decide before the messaging channel stays open.

8 min read

The phone tree has a script for after hours. The messaging channel usually does not, which is how an internal medicine group ends up with an after-hours text arriving at 11pm on a Saturday that nobody agreed to answer and nobody agreed to ignore. The patient hears silence. The practice hears about it Monday.

Adult primary care panels have moved a large share of their traffic to asynchronous channels, and the volume did not politely confine itself to business hours. Refill requests, results questions, insurance changes, and the message that starts I know it is late but arrive continuously across evenings and weekends.

What makes this harder than the phone is that a text does not feel closed. A caller who reaches a recorded message at 11pm knows the office is shut. A patient who sends a message into a channel that has answered her within minutes for the last six weeks has no reason to think tonight is different, and she waits.

So the practice is running an implied service level it never wrote down, staffed by whoever happens to look, with no record of what was promised. That is the actual problem to solve, and it is a policy problem before it is a technology one.

Decide what the channel promises, then make it say so

The first decision is whether the messaging channel is open after hours at all, and both answers are defensible as long as the patient knows which one is true.

A closed channel needs an immediate automatic reply that says so in plain language, states when it will be read, and gives the exact number to call for anything urgent. The reply has to go out in seconds, because a delayed autoresponder is worse than none; it teaches patients that something is watching and still not answering.

An open channel needs a real definition of open. Which categories get handled overnight, which get acknowledged and queued for the morning, and which trigger a live escalation. Writing that list is the entire program, and it is short enough to fit on one page.

The important detail is that the acknowledgment is not a courtesy, it is the mechanism that keeps a waiting patient from calling the on-call line to ask whether her message was received. Practices that add same-minute acknowledgment usually see the after-hours call volume fall on its own, because most of those calls were status checks.

The categories that actually arrive overnight

Overnight message traffic in an adult primary care panel is more predictable than it feels, and it sorts into four buckets.

Refill requests are the largest. Most are routine, some involve controlled substances, and nearly all of them have a days-remaining answer sitting in the chart that determines whether this can wait until Monday. Creating the refill request case with the medication matched and the days remaining attached is work that can be done at 11pm and reviewed in ninety seconds at 8am.

Scheduling comes second, and it is the category with the clearest overnight resolution. Booking, rescheduling, and cancelling against the practice’s own templates does not need a person and does not need to wait.

Third is administrative: insurance changes, balance questions, forms, records requests. These become routed cases with the details already captured, and nothing about them needs a clinician.

Fourth is the clinical question, and that is the bucket where the practice’s escalation rules do the work. The AI’s role is to recognize that it is in that bucket, stop, and follow the practice’s written path, which is either connect now or hold for the morning with the message preserved as written.

Escalation has to be a path, not a page

The weakest link in most after-hours setups is the last twenty feet: what happens between deciding to escalate and a clinician actually reading something.

A page with a callback number puts the entire burden on the clinician, who now calls a patient blind. The better pattern hands the on-call clinician the message as written, the patient’s chart, the last visit, the active medication list, and whatever the practice’s protocol already collected, so the first thirty seconds of the call are not spent reconstructing the situation.

Escalation also needs a failure path. If the on-call clinician is not reached inside a defined window, something has to happen next, and the something has to be written down before the night it matters. Practices that skip this discover the gap during an incident review rather than during design.

Record both events separately. Escalation sent and clinician reached are different facts, and the interval between them is the number that tells you whether the after-hours program is real.

The complications nobody plans for

Three recurring ones, all of which show up inside the first month of an open channel.

The message that is not from the patient. Adult children of older panel members send a lot of overnight messages, frequently from their own phone or portal account, about a parent’s care. Relationship capture and access rules matter here in a way that is easy to get wrong quietly, because the person messaging genuinely is involved and genuinely does not have a documented right to everything in the chart.

The thread that switches topics. A message about a balance ends with a sentence about a symptom, and a router that classified on the first sentence sends the whole thing to billing. Classifying on the full message and splitting it into two cases when it carries two requests is the fix, and it is the difference between a router that helps and one staff learn to distrust.

The patient who messages the same thing four times because nothing answered. Deduplication against an open case for the same patient and category prevents the morning worklist from showing four entries that are one request, which is how a fifteen-item queue looks like sixty.

What the morning should look like

The measure of an overnight program is the state of the practice at 7:45am, not the volume of activity between midnight and dawn.

A good morning is a single ordered worklist. Refill cases with the medication matched and the days remaining shown. Scheduling requests already resolved and marked as such. Administrative cases with the detail captured. Clinical messages at the top, in the order they arrived, with the patient’s own words intact.

A bad morning is a channel history that someone has to read. If staff have to scroll to find out what happened overnight, the program moved work in time rather than removing it.

Track three numbers. Share of overnight messages resolved without staff, median time from message to acknowledgment, and the count of messages that were still unanswered when the office opened. The third one is the service level the practice is actually running, whether or not it ever wrote it down.

Key Takeaways

  • Decide whether the messaging channel is open after hours and make the channel say which one is true within seconds of a message arriving.
  • Same-minute acknowledgment usually reduces after-hours call volume on its own, because a large share of those calls are status checks.
  • Sort overnight traffic into refills, scheduling, administrative, and clinical, and write what happens to each before turning anything on.
  • Refill requests can be built overnight with the medication matched and days remaining attached, then reviewed in seconds in the morning.
  • Hand the on-call clinician the message, the chart, and the context rather than a callback number, and define what happens when the clinician is not reached.
  • Capture the relationship when the sender is an adult child or caregiver, and apply access rules rather than assuming involvement equals authorization.
  • Deduplicate repeat messages against an open case so a fifteen-item morning queue does not look like sixty.
  • Report unanswered-at-open count. That number is the service level the practice is actually running.

An internal medicine group does not get to decide whether patients send messages at 11pm. It only gets to decide what happens next, and leaving that undefined is itself a decision that the practice pays for in Monday complaints and on-call fatigue. An AI team working the overnight queue inside athenaOne can acknowledge instantly, resolve the scheduling and administrative traffic outright, build the refill case with the chart detail attached, and escalate the rest along a path somebody wrote down in daylight.

Sources

Ready to See It in Action?

See how PGA covers the after-hours message queue and escalates only what needs a clinician

Schedule a Demo →

Written by Kevin Henrikson