Most concierge practices I've worked with can tell you their panel size and their average revenue per member. Ask them how many clinical minutes it takes to serve one member per year, and you get a shrug or a guess. That gap is where margin quietly disappears.

Throughput in a concierge model doesn't mean cramming more visits into a day. It means understanding where your clinician minutes actually go so you can decide, with real numbers, whether you can add fifty more members without hiring, or whether you're already underwater at your current panel. Here are the metrics I've found genuinely useful, and the ones that mostly waste dashboard space.

Start with clinician minutes, not visit counts

Visit volume is a legacy metric from fee-for-service. In a concierge practice, a "visit" can mean a 90-minute annual executive physical, a 12-minute video check-in, or a message thread that pulls the physician in five separate times over a week. Counting them as one unit each hides everything that matters.

The unit I track is clinician minutes per member per month. Total up every minute a licensed clinician spends on behalf of a member: synchronous visits, chart review, message responses, refill approvals, care coordination calls, results interpretation. Divide by active members. In practices I've operated, this number ranges from about 25 minutes per member per month in a light-touch model to over 90 in a high-acuity longevity or complex-care panel. Neither is inherently right or wrong. What matters is whether the number matches what you priced for.

Once you have that baseline, break it down by channel:

  • Synchronous clinical time (in-person and video)
  • Asynchronous messaging (portal, secure text, email)
  • Documentation and chart prep
  • Refill and prior-auth work
  • Care coordination (specialists, labs, pharmacy)

The mix tells you where to invest in workflow. A practice spending 35% of clinician time on documentation has a very different problem than one spending 35% on messaging.

The metrics that actually move decisions

Panel load index

Take your average clinician minutes per member per month and multiply by panel size. Compare that to the available clinical minutes your physician actually has (subtract admin blocks, PTO, CME, meetings). If you're at 85% or higher, you have no capacity for growth or absence coverage. Most owners discover they've been running at 95%+ for months and wondering why their doctor is short with staff.

Message turnaround time, segmented

Aggregate response time is a vanity metric. Segment by message type. A refill request should close in under four working hours. A clinical question with symptoms needs triage within one hour and a clinician response within the day. A general admin question can wait until end of day. When you segment, you find out that your "4-hour average response" is really 30 minutes for easy stuff and 26 hours for the messages that actually needed the doctor.

Refill cycle time

From member request to medication in hand. For compounding-pharmacy-adjacent practices, this is one of the highest-signal metrics you have. It touches intake, clinical review, pharmacy communication, and shipping. If your median is above 72 hours, something in the handoff is broken. Track P50 and P90, not just the average, because the tail is what generates the angry messages.

Documentation lag

How many hours between visit end and note signed. Anything over 24 hours means your clinician is finishing notes at night or on weekends. That is the single most reliable predictor of burnout I've seen. It also creates downstream drag: unsigned notes delay billing, coding review, and any AI agent or staff member trying to act on the encounter.

Touch count per resolved issue

How many separate interactions (messages, calls, chart entries) does it take to close one member request? A refill should be one to two touches. A new symptom workup might reasonably be five to eight. If simple things are averaging four touches, you have a routing problem, not a volume problem.

Metrics that mislead

Utilization rate, defined as booked visit hours divided by available hours, is popular and mostly useless in concierge. Your calendar can be 60% booked while your physician is drowning in async work. Members in this model expect that most of their access happens outside scheduled slots.

Net Promoter Score is fine as a lagging outcome metric, but it doesn't tell you where time is going. A practice can have a 78 NPS and still be six months from a clinician quitting.

Raw message volume without context is another trap. More messages can mean higher engagement (good) or that your intake and education materials are failing and members are asking things they shouldn't need to ask (bad). Volume alone can't distinguish the two.

How to actually capture the data

The honest problem: most EMRs weren't built to measure clinician minutes across channels. They log visits and notes. Messages sit in one queue, refills in another, phone notes somewhere else, and none of it rolls up.

You have three practical options. First, sample manually. Have your clinician track time in 15-minute blocks for two representative weeks per quarter. Painful but revealing. Second, instrument your EMR with event timestamps: message opened, message responded, refill received, refill approved, note started, note signed. Even without perfect duration data, timestamps let you calculate cycle times and identify bottlenecks. Third, use an infrastructure layer that captures agent and human activity across the workflow, so refill approvals, message triage, and intake all produce structured events you can query.

We built Qintara around the assumption that operators need this data as a native output of the system, not a side project. If you're rebuilding your stack or evaluating tools and want to see what that looks like in practice, talk to our team.

Turning the numbers into decisions

Metrics only matter if they change what you do. Some patterns I've seen repeatedly:

When documentation lag climbs past 24 hours, the fix is almost never "the doctor needs to be faster." It's ambient scribing, template restructuring, or offloading note components to a well-designed agent workflow. Adding pressure to a burned-out clinician makes the lag worse, not better.

When refill cycle time is bimodal (fast for some meds, slow for others), the slow bucket almost always involves an external pharmacy handoff or a prior auth. That's a workflow design problem, not a staffing one.

When touch count per resolved issue creeps up, look at your first-touch routing. Front desk sending everything to the clinician instead of resolving what they can, or an intake form that doesn't collect enough to let the clinician answer in one pass.

When panel load index crosses 85%, stop selling new memberships until you've either added capacity or reduced minutes per member. Growing past that line is how practices end up with three-month waits for annual physicals and members quietly leaving.

Frequently Asked Questions

How often should we review these metrics?

Weekly for operational metrics (message turnaround, refill cycle time, documentation lag). Monthly for panel load and minutes per member. Quarterly for the mix analysis and any strategic decisions about pricing or hiring. Reviewing panel load daily creates noise; reviewing message turnaround monthly is too late to fix a bad week.

What's a reasonable target for clinician minutes per member per month?

It depends entirely on your model and price point. A $2,500/year primary care concierge practice with a panel of 400 probably needs to stay under 40 minutes per member per month to remain viable. A $25,000/year longevity practice with a panel of 100 can absorb 120+ minutes. Back into the number from your pricing and your clinician's total available hours, then measure against it.

Should we measure staff time the same way?

Yes, though with a different purpose. Clinician minutes drive panel capacity and burnout risk. Staff minutes drive operating cost and identify where automation or agent workflows pay back. Track both, but don't blend them into a single "labor per member" number: they answer different questions.

How do we handle work that spans multiple members, like protocol updates or CME?

Keep it out of per-member calculations. Track it as a separate overhead bucket that reduces available clinical minutes. This is why panel load index uses actual available minutes rather than raw scheduled hours. If protocol maintenance is eating six hours a week, that's a real constraint on capacity and should show up in your denominator.