Group teletherapy: logistics that make or break it

Group teletherapy platform logistics decide more about whether a virtual group survives than the curriculum does. A well-designed DBT skills group with a broken join link and a roster that leaks names will fold by week four. A mediocre process group with reliable plumbing will run for years. The clinical work is yours; this is about the twenty or so operational decisions sitting underneath it, most of which you only notice when one of them fails in front of nine people.

Say you run a Tuesday evening DBT skills group, twelve weeks, eight members, one facilitator and one co-facilitator. That is the hypothetical I will use throughout, because the problems are easier to see with a specific shape in mind.

The group agreement is where confidentiality actually lives

In a room, confidentiality is mostly enforced by the walls. On video, every member is sitting somewhere you cannot see, possibly with a roommate in the next chair, possibly with a phone recording. Your group agreement has to do work that architecture used to do.

What belongs in it, signed before session one:

• No recording by anyone, members or facilitators, and an explicit statement that screenshots count. Make this the first clause. It is the one people are most tempted to break and the one with the widest blast radius.
• A private-space requirement, with a concrete standard: alone in a room with the door closed and headphones on. Rule out the parked car outside work and the commute home; both come up more than you would expect.
• Cameras on as the default, and a named exception process. Whatever you decide, decide it in advance rather than negotiating it live in week three.
• What members may and may not share outside the group, phrased concretely. "What you hear here stays here, including on group chats you start with each other."
• Location confirmation. Members confirm the address they are joining from at the start of each session, so that if someone becomes unsafe you know where to send help. This one gets skipped constantly and it is the one that matters at 7:40 on a bad night.

On recording specifically, my position is that recording-free should be the default and the exception should be rare. A recording of a one-to-one session is one person's disclosure to manage. A recording of your Tuesday group is eight people's, including people who never got a vote on how carefully the file gets stored. If supervision requires it, take separate written consent from every member, state the retention period, and build a path for the person who says no that does not push them out of the group.

If any member is in your substance-use program, 42 CFR Part 2 applies on top of everything else, and the 2024 final rule has been fully in force since February 2026. Under Part 2 the affiliation itself is protected, so the group name in a calendar invite, an email subject line, or a reminder text can be the disclosure. Neutral naming everywhere the patient's phone can see it.

How do you run a roster without exposing members to each other?

This is the requirement that separates a group teletherapy platform from a video tool with a shared link, and it is worth being blunt about. A group has two layers that must stay separate:

1. The room, where members see each other's first names and faces, because that is the point of a group.
2. The record, where every member has their own chart, their own note, their own measures, and their own claim, and can see none of anyone else's.

Shared links collapse those layers. Anyone who has the URL is in, including the member who dropped out in week two and the partner who found it in a browser history. A roster-based room instead binds the session to a specific list of patients, which gets you three things at once: nobody joins who is not enrolled, attendance is captured as a byproduct of joining, and adds or removes are a roster edit rather than a new link emailed to seven people.

Removing a member mid-cohort is the test case. When someone leaves your twelve-week group in week five, the correct behavior is that their access ends, the remaining members' experience does not change, and nobody has to be told a new link exists. If your setup requires reissuing a link, you have just announced the departure to everyone.

What breaks in the first five minutes?

Almost always the same four things: audio input on a laptop that defaults to the wrong microphone, a browser that has not been given camera permission, a corporate laptop blocking the connection, and someone on cellular in a dead zone. None of these are exotic and all of them are preventable with a ten-minute tech check before session one.

Run it as a real scheduled event, not an email with instructions:

• Five to seven days before the group starts, each member joins a solo test room for ten minutes, on the exact device they plan to use for the group. Not their phone if they will use a laptop.
• Confirm audio in and out, camera, and that the browser remembered the permissions. Have them mute and unmute once so they know where the button is.
• Establish the fallback out loud: if you drop, rejoin from the same link; if you cannot get back in within five minutes, text the practice number and we will call you.
• For work devices, ask them to test from home network and work network if they might use both. VPNs and locked-down laptops are the most common silent failure.

A browser-based join with no app install and no account creation removes most of the remaining friction. Every additional step between a text message and a face on screen costs you attendance, and in a group, one person fumbling with a download at 6:02 pulls the facilitator away from the other seven.

Attendance itself responds to the same reminder mechanics as individual sessions. A Cochrane review of mobile phone reminders put the risk ratio for non-attendance at roughly 0.77 against no reminder, and a study in The Permanente Journal found a second text cut mental-health no-shows by about 11% relative to a single one. Groups have a sharper edge here than individual work, because a session with three of eight members present is a different clinical event, and members who show up to a thin room start deciding the group is dying.

Co-facilitation, and who does what

Two facilitators in a video group are worth more than two in a physical room, because one of them can absorb the operational load that would otherwise interrupt the clinical work. Split it explicitly before the cohort starts: one runs content, the other watches the chat, tracks who has not spoken, handles the member whose audio failed, and manages arrivals after the start.

Both facilitators need real access to the group in the system, including documentation rights, rather than one being a guest who joins through a forwarded link. When a co-facilitator has to ask for the link every week, they are not co-facilitating, they are attending.

How do you bill it?

Group psychotherapy bills per member, per session attended, generally under 90853. The mechanical requirement that surprises people moving groups online is that one clinical event has to produce one claim per attending member, driven by who actually joined rather than who was enrolled. If your roster has eight and six showed, you bill six.

Telehealth place-of-service codes and modifier expectations vary by payer and have changed repeatedly over the past few years, so verify current requirements with each payer instead of carrying your in-person configuration across. I am not going to quote reimbursement figures here, because they are payer-specific and anyone quoting a national average for group telehealth is guessing.

Documentation follows the same shape: one note per member in their own chart, describing that member's participation, with shared group content described once and individual response described individually. Copy-pasting the same paragraph into eight charts is the fastest way to make an audit unpleasant. If you are running measures, the AIMS Center treat-to-target cadence works in group settings too, with PHQ-9 or GAD-7 at baseline, every two weeks during active treatment, and a flag on anyone under 50% improvement by week 10 to 12. In a twelve-week group that flag lands right around the point where you still have time to plan aftercare.

What to skip

Waiting rooms that require the facilitator to admit each person individually. With eight members arriving in a two-minute window, admitting people is a full-time job for the first five minutes, and a roster-bound room removes the need for it.

Breakout rooms in a clinical group of eight, unless the curriculum genuinely calls for pairs. The transition costs ninety seconds each way and the facilitator loses sight of everyone during it.

Email chains for scheduling changes. Reply-all on a group email is a confidentiality incident waiting for its moment.

Valence handles group telehealth natively rather than as an individual session with extra people in it: roster-based rooms, co-facilitators with full access, a group appointment that links multiple patients while keeping their charts and notes entirely separate, and browser join with no app and no patient account. Reminders for group sessions run through the same care pathways as everything else, with quiet hours, weekly caps, and opt-out handling built in.

None of which fixes the part that actually kills groups. A platform can hold the roster, the notes, and the claims apart correctly and you can still lose a cohort by week four, because the group agreement was vague about recording or nobody ran the tech check and two people spent the first session troubleshooting audio. The plumbing buys you the chance to do the clinical work. It does not do any of it for you.