Health coaching ROI for medical practices: 4 inputs

Does adding health coaching pay for itself?

What is the ROI of adding health coaching to a medical practice? It can pay for itself, but not through cost savings. The best cost study of primary-care health coaching found no evidence it saved money at one year. When it pays, it pays through program margin, through patients who complete rather than drift, and through clinical time returned to the clinician.

Your practice P&L on one screen, a per-patient quote on the other

The decision stalls in the same place. Month end, the P&L open, a vendor proposal quoting a fee per patient per month open beside it. The clinical logic is not in question. Arithmetic is what stalls it, so the file closes again.

The room has changed. In an MGMA Stat poll on 30 June 2026, with 221 applicable responses, only 47% of medical groups reported year-to-date revenue above the same period in 2025, while 84% reported higher operating costs. A year earlier, 56% had reported growth (MGMA, 2026).

That does not argue against a behavioral layer. It argues against buying one on belief. If you have not worked through why sound plans come apart between appointments, start with what the evidence says actually changes patient behavior.

What health coaching actually costs, measured

A between-visit behavioral layer is a standing labor cost per patient per month that scales with caseload, so you can know that line first.

Wagner and colleagues costed the Health Coaching in Primary Care trial, 441 patients, at $483 per participant per year at San Francisco wages, or about $356 using nationalized costs (Wagner et al., 2016). Divide by twelve and you have the unit a vendor quotes you: roughly $40 per patient per month at the high-wage end, roughly $30 nationalized.

Three details matter more than the headline number. Direct labor was 70% of total cost, with training at 6% and benefits, space and supplies at 24%, so little of that is anything but someone's time. Caseload ran 40 to 60 patients per coach, which is your denominator. Participants received about nine coaching hours, which is the dose the cost buys.

Who delivers those hours, and whether you hire, train or embed, is a separate decision with its own briefing. It changes the cost's shape, not the model below.

The honest headline: health coaching did not save money at one year

Here is the finding most vendor decks leave out. In the same paper's own words, "There was no evidence that the coaching intervention saved money at 1 year." Total healthcare costs came to $3,207 for the intervention group against $3,276 for controls, not a significant difference (Wagner et al., 2016).

Read it plainly. A well-run program did not cut utilization enough to pay for itself in twelve months. One trial is not the whole literature, and a year is short for chronic disease, but it is the best direct costing available.

So a proposal promising the program pays for itself through avoided admissions is promising what the cost evidence does not support. Owners are right to refuse that. The case has to be built somewhere other than savings.

Four ways a coaching service line returns its cost, on four clocks

Owners get this wrong by comparing a monthly cost against an annual, diffuse benefit. Separate the returns, and label each honestly.

1. 01

   Channel 1. Billed: what a payer may already cover

   Behavioral health integration, or BHI, is the Medicare code family that pays for structured between-visit support under a treating clinician. For 2026, general BHI code 99484 carries a national average near $57 per patient per month. Collaborative Care runs higher: roughly $163 for the initial month (99492), $130 to $140 for later months (99493), and about $70 for the add-on (99494) (Nsight Health, 2026; ThoroughCare, 2026). Treat those as orientation, not your number. They move with geographic adjustment, payer contract and contractor policy, and carry eligibility and documentation requirements. The CMS Physician Fee Schedule Look-Up Tool is the primary record, so verify each code for your own locality before it enters the model (Centers for Medicare & Medicaid Services, 2026). Evidenced, and conditional.

2. 02

   Channel 2. Cash-pay margin: a formula, never a benchmark

   Monthly contribution is enrolled patients multiplied by price per patient per month multiplied by completion rate, minus enrolled patients multiplied by delivery cost per patient per month. There is no industry price to plug in. Anyone quoting you one is quoting their own market and patient mix. You must model this.

3. 03

   Channel 3. Retention: revenue you keep rather than revenue you add

   Assume a patient who completes stays longer than one who quietly drifts, then test that against your own attrition data before you count it. That is revenue continuity, not new revenue, and it is easy to overstate. The line is retained patients multiplied by your own average annual revenue per patient. Cannot pull both? Leave it at zero. You must model this.

4. 04

   Channel 4. Reclaimed clinician time: real, and not a dollar figure

   Porter and colleagues estimated that delivering all guideline-recommended adult primary care alone would take 26.7 hours per day, against 9.3 hours per day under a team-based model, a reduction of roughly 65% (Porter et al., 2023). That is not a claim that a behavioral layer returns 65% of your day. It is evidence that much of a primary care panel's work is delegable in principle, and between-visit follow-through sits inside that share. Resist pricing it: capacity only becomes revenue if you fill the reclaimed slots. Evidenced as a mechanism, unpriced. Whether more contact changes outcomes is a clinical question, and so is delivery format; both have their own briefings.

The four numbers that give your practice a breakeven census

Health coaching ROI for a medical practice is four inputs and one line of arithmetic.

1. 01

   Enrolled patients

   How many patients are actually in the program this month, not how many are eligible for it.

2. 02

   Price or billed rate per patient per month

   Your cash price, or the rate you have verified against your own locality fee schedule.

3. 03

   Delivery cost per patient per month

   Anchor to $30 to $40 from the costing evidence, then adjust to your wage market and delivery model.

4. 04

   Completion rate

   The share of enrolled patients who finish a month of the program. This is the input owners guess at.

The one line of arithmetic that answers the question

Breakeven census

The number of enrolled patients at which the program's contribution covers its overhead: monthly program overhead divided by price multiplied by completion rate, minus delivery cost per patient per month.

• The arithmetic that follows is illustrative only. The price, completion rate and overhead are placeholders chosen to show the mechanics. They are not benchmarks and not Clementine's numbers.
• Say $2,000 a month of program overhead, a $150 price, 75% completion, and $40 delivery cost.
• Contribution per enrolled patient is $150 multiplied by 0.75, minus $40, which is $72.50.
• Breakeven census is $2,000 divided by $72.50, or 28 patients.
• Twenty-eight sits inside the 40 to 60 patient caseload one coach carried in the trial. That is the shape of the answer: a census number you can hold against your panel, on one page, in an afternoon.

The uncomfortable input most owners leave out

Completion rate is the number owners skip, and the one that decides the model. It sits in the revenue term, and if you bill BHI it sits in the documentation term too, because an unfinished month is an unbilled month.

So model a pessimistic completion rate, not an aspirational one, and track it from the first enrolled patient. If the line only works at 90% completion, you do not have a service line, you have a hope.

Measuring between-visit outcomes clinically is its own discipline and its own briefing. Here, completion is only a financial input, and the practical ways to hold it up are covered in how to improve patient adherence between visits.

When adding health coaching would be the wrong call for your practice

Name these out loud before you commit.

1. 01

   Your panel cannot reach the breakeven census

   Arithmetic, not ambition.

2. 02

   Nobody owns enrollment

   A program no one is accountable for filling sits at a third of census and loses money quietly.

3. 03

   You are modeling savings

   The one claim the cost evidence does not support.

4. 04

   You are hiring before you have priced the line

   Payroll is the 70% that does not flex.

5. 05

   You have not chosen cash-pay or billed

   Different businesses, different documentation, and deciding late is expensive.

A one-page health coaching ROI model you can build this week

Pull a realistic enrolled census for month three, not month twelve. Set a price, or verify your locality rate for 99484 and Collaborative Care against the CMS schedule. Set delivery cost at $30 to $40, and a deliberately pessimistic completion rate. Compute contribution per patient, then the breakeven census. Name the input that would have to be wrong for the decision to fail.

At Clementine, the embedded coaching service line is built to be modeled this way: a known cost per patient per month, a defined caseload, no new clinical hire. Sleep is usually the entry point, because it is low friction and its downstream effects are well documented, as in what improves when sleep improves.

An owner who can model this in four inputs is in a far stronger position than one who dismisses it or buys it on faith. This layer is infrastructure you price and staff, not enthusiasm you hope for.