← Back to Resources

Practice Operations

The Cost of Desk-Bound Patient Communication

Desk-bound patient communication makes every reply wait for a workstation. What a multi-specialty group loses to that, and which parts never need a chair.

7 min read

Desk-bound patient communication is the quiet constraint in most multi-specialty groups. Every reply, every confirmation, every rescheduling message waits for a specific person to be at a specific workstation. The queue does not move because nobody is sitting in front of it, and the queue does not care what time it is.

That constraint is invisible in the numbers most groups track. Message volume looks fine. Response time looks acceptable on average. What the average hides is that the whole queue moves in two bursts, once mid-morning and once after the last patient leaves, because those are the only hours anybody has a chair.

Meanwhile the patient who asked a Tuesday afternoon question about a Wednesday morning appointment gets an answer Wednesday afternoon. The appointment already happened, or it already did not.

The queue waits on a chair, not on a decision

Sort a week of outbound patient communication in a multi-specialty group and most of it needs no judgment at all. Confirmations, reschedule offers, pre-visit instructions, form status, balance questions, and directions to the right building are administrative from end to end.

They are waiting on availability, not on expertise. A staff member has to be logged in, in the right department context, with the patient record open, to send a message that has one correct answer.

That is the cost worth naming. Not staffing cost, which is what these conversations usually become. Latency cost. The gap between when the patient asked and when the practice was physically able to answer, multiplied across every specialty in the group.

In a March 10, 2026, MGMA Stat poll of practice leaders, the most time-intensive phone tasks were eligibility and prior authorization at 45%, scheduling at 31%, intake at 9%, prescription refills at 6%, and an other category at 9%. Every one of those categories has a messaging equivalent, and the messaging version is the one that gets deferred because it has no ringing phone attached.

The paired appointment that only gets one reminder

Here is the complication that makes desk-bound communication actively expensive in a group with imaging or procedures alongside clinic visits.

Native reminders fire on the chronologically first appointment. A patient with a 9:30 imaging slot and a 10:00 provider visit gets reminded about one of them. They show up for the scan, or they show up for the doctor, and the other half of the day is a hole in a schedule that was booked correctly.

At one vascular group, roughly 90% of visits were an imaging study plus a provider visit, booked as a pair with specific spacing. Nothing in the scheduling template enforced the pairing. It existed as convention, and the patient just called asking to see the doctor.

Catching that by hand means somebody reading the day’s booked appointment list, spotting the pairs, and sending a second message. That is exactly the work that only happens when there is time, which means it happens least on the busiest days.

Done automatically, the booked list is read continuously, pairs are detected by patient and date, and the reminder describes the whole visit rather than the first leg of it. When the pairing looks wrong, one leg missing or the spacing violating the department’s rule, it goes to a scheduler rather than being auto-corrected. The system flags the broken pair. A person decides what the day is supposed to look like.

Groups that decide to move volume onto text run into the same wall two weeks later. Which patients have actually agreed to receive messages, on which number, for which purpose, and where is that recorded.

In a multi-specialty group this is messier than in a single practice, because the patient may have consented at one department’s front desk and never at another, and the mobile number on file may be four years old. Sending anyway is not an option, and checking by hand at send time defeats the point.

The workable pattern is to make consent state a property of the record rather than of somebody’s memory. The messaging terms and conditions the practice operates under are readable programmatically, consent is captured once through a documented flow, and every outbound message checks state before it sends rather than after somebody complains.

What that buys you is the ability to run outbound campaigns at all. Groups that never sorted this out end up limited to inbound replies forever, which is the smallest and least valuable half of the channel.

Sort the queue before it reaches a person

The mistake groups make when they finally automate this is automating the sending and leaving the sorting alone. Volume goes up, the human half of the queue gets bigger, and the front office is worse off.

The useful design does the opposite. Everything arriving is classified first: administrative requests that can be completed, administrative requests that need a specific person, and anything that has to reach a clinician. Only the third category touches a clinical inbox, and it arrives with the record context attached.

The AMA has covered clinician inbox burden at length, noting that most organizations can reduce inbox volume by 50% by eliminating redundant notifications and other noise. In a multi-specialty group that noise is structural. The same patient generates messages from three departments and every department copies somebody.

The AI completes what is complete and routes what is not. It does not answer anything clinical, does not interpret a symptom a patient describes in a message, and does not decide what a patient needs. Those pass straight through with the context assembled so the clinician does not have to go find it.

Measure latency, not volume

If you change one number in your reporting after reading this, change it from messages handled to time from patient question to practice answer, split by category.

Volume rewards activity. Latency rewards the thing patients actually experience and the thing that determines whether a message prevented a no-show or arrived after it. Split it by category and the picture gets actionable fast, because the administrative categories should approach zero and the clinical ones legitimately should not.

Then watch the tail rather than the mean. A group with a two-hour average and a forty-hour ninety-fifth percentile does not have a two-hour practice. It has a small number of patients having a terrible experience and an average that hides them.

That tail is almost always the desk-bound part. Find it, name which category it is, and you know exactly which piece of the queue should never have needed a chair.

Key Takeaways

  • Name the cost as latency rather than staffing, because the queue is waiting on availability and not on expertise.
  • Read the booked appointment list continuously so paired visits get a reminder covering both legs, not just the earlier one.
  • Send a broken or oddly spaced appointment pair to a scheduler instead of auto-correcting it.
  • Treat messaging consent as a state on the record checked at send time, not a checkbox somebody remembers.
  • Classify inbound messages before they reach anyone, so only genuinely clinical items land in a clinical inbox.
  • Replace messages-handled with time-from-question-to-answer by category as your headline communication metric.
  • Watch the ninety-fifth percentile rather than the mean, since the tail is where the desk-bound work hides.

A multi-specialty group does not have a patient communication problem so much as a geography problem. The work is fine. It is stuck to a chair. Move the administrative half onto athenaOne messaging that runs without anyone being logged in, keep consent state on the record so outbound is actually possible, sort everything before it reaches a person, and measure the delay rather than the volume. What is left for staff is the part that genuinely needed them, and it arrives with the context already gathered.

Sources

Ready to See It in Action?

See how PGA clears the administrative half of a multi-specialty message queue without anyone sitting down to it

Schedule a Demo →

Written by Kevin Henrikson