My biggest gripe with outpatient medicine is the inbox messages and their management.

They’re a time-suck right now. I say “right now” because I don’t think we (outpatient clinics, health systems) have adopted robust frameworks to triage the influx of inbox messages from patients and caregivers. In other words, patient-facing technology is advancing faster than the physician-facing systems designed to manage care.

The turning point can be traced to the peri-pandemic era. In my first year of medical school, I had “MyChart”—Epic’s patient-facing app that lets users view records, lab results, and even schedule appointments. From what I recall, users could also message their care team, choosing the type of message:

I once used it for a non-urgent medical question that, in hindsight, probably warranted a phone call to the on-call clinician. I mis-triaged myself. Still, at the time, messaging wasn’t widespread enough to become a problem.

In a new JAMA article, Long and colleagues investigated this rise in patient messaging by analyzing 8 billion encounters in Epic from 2020 through 2025, spanning 2,000 hospitals and 47,000 clinics. They found:

As I’ll explain more in the Insights section, increased access to the care team—especially physicians—is both a boon (for patients) and a bane (for physicians, though this may be temporary). It’s a boon for patients because they can quickly reach their care team and have medical questions answered within a day or two. It’s a bane for physicians because inbox-management time has increased without corresponding changes to the administrative system (time allotted for admin work, compensation for after-hours responses). This is especially true because patient and caregiver messages vary widely—from “I have a question about my medication” (which medication?) to “I’ve had worsening shortness of breath for the past week; can you prescribe antibiotics?” (this warrants a phone call to triage whether the patient needs an urgent visit or an emergency-department evaluation!).

The problem: Physicians and care teams are drowning in patient portal inbox work that keeps growing, without a matching redesign of staffing, workflows, and incentives.

There are three straightforward solutions to this, and I’m proposing them from both sides of the portal: as a patient and as a physician.

First, we need a real triage layer (people + protocols) before messages reach the physician. Most inbox work is predictable: refills, routine labs, forms, scheduling friction, and “is this urgent?” symptom questions. The fix is to formalize routing rules and give staff the authority to close the loop on protocolized items, escalating to the physician only when medical decision-making is truly required. If portal messaging is going to be a core modality, it needs an operating model that treats it like one. At my institution, we have some version of this, but I still get inbasket messages forwarded to me that should’ve been clarified upstream—vague “question about lab result” or “question about med” messages that prompt the obvious response: “Which lab result and which med?” Generative AI could help tighten that front-end clarification.

Second, we should use AI to deflect and draft, with physicians acting as reviewers when needed. Several health systems are already deploying tools that sort messages by urgency, draft replies, and deflect administrative requests (scheduling, navigation, med logistics) before they ever land in the clinician’s in-basket, which I wrote about here nearly two years ago. This doesn’t limit access. It protects it by keeping physician attention focused on the minority of messages that require physician-level judgment. The goal is fewer raw messages, better prioritization, and faster “good enough” first drafts for clinicians to edit and send.

Last, we need incentive alignment: dedicate time and bill for asynchronous care when appropriate. If messages are increasingly where care happens, health systems need to fund the work the way they fund visits. That means protected admin time tied to message volume, clear service-level expectations, and operational support so after-hours work doesn’t become the default. It also means capturing revenue when a message thread crosses into billable medical care, rather than treating it as free labor and hoping satisfaction scores make up the difference. Other professions already do this. If I call my lawyer with a 15-minute question, I get a bill for 15 minutes.

In summary, patient portal messaging has become a second clinical workload layered on top of visits—not a replacement—and the growth curve from 2020 to 2025 makes it obvious this isn’t self-correcting. If we want digital access to stay a win for patients without turning into uncompensated, after-hours care for us physicians, health systems need to build an operating model for asynchronous care: real triage, smarter drafting/deflection, and protected time (and payment) that matches the work.

I’ve written previously about this topic. If you want to dive deeper, see below: