DX / Differentials / MULTI-SPECIALTY MSO

D-04 — MULTI-LOCATION MARKETING · MULTI-SPECIALTY MSO

One playbook, different results by location — MSOs

Same staffing ratio. Sevenfold different economics.

DIFFERENTIAL — LIKELY CAUSES, MOST LIKELY FIRST

1. 01

   Task alignment and staff skill, not staffing quantity

   Two sites can carry identical support staff per FTE physician and deploy them at completely different points in the encounter. What the staff do, and whether the right work is delegated off the physician, determines the outcome — not how many of them there are.

   Across 141 multispecialty groups with primary and specialty care, at the median staffing level of 5.0 total support staff per FTE physician, total medical revenue after operating cost per FTE physician ranged from as low as $250,000 to as high as $1,800,000 — a sevenfold difference at identical staffing. MGMA's own conclusion: staffing quantity matters, but staff skill and task alignment matter more.

   MGMA DataDive Cost and Revenue, via MGMA, "Data Mine: The secret of staffing success" (David N. Gans, 2019-06-10).

   NOT THIS IF — Not this if the spread across sites tracks staffing ratio closely. If the highest-staffed sites are simply the best-performing ones, the variance is a resourcing decision, and the next cause is the one to read.

2. 02

   Staffing composition — the mix, not the count

   High-performing groups do not merely staff more; they staff differently, and the difference concentrates in business operations rather than clinical support.

   Comparing 4th-quartile to 1st-quartile-productivity multispecialty groups: total support staff per FTE physician 73% greater, business operations staff 1.21 versus 0.58 per FTE physician (more than double), clinical support 2.02 versus 1.54. Separately, practices with 6.1 or more FTE employees per physician produced roughly three times the output of practices with 2.1 to 3.0, and generated more than three times the median total medical revenue after operating cost per FTE physician.

   MGMA DataDive Pro Cost and Revenue, 2016 data, via MGMA "Threading the needle" (structure valid, dollar levels dated); MGMA DataDive Cost and Revenue via "Data Mine: The secret of staffing success."

   NOT THIS IF — Not this if business-operations staffing per FTE physician is already even across sites. Then composition is not the variable and the difference is in ownership structure or mix.

3. 03

   Ownership structure differs across sites and carries its own cost curve

   Acquired sites arrive with the staffing model and cost trajectory of their prior owner, and those models are structurally different. A single brand can be running two economies.

   More than 59% of hospital-owned multispecialty groups have four or fewer FTE employees per FTE physician, against only 4% of physician-owned groups. Hospital- and IDS-owned multispecialty groups report 62% of the total employees per FTE physician that physician-owned counterparts do, while physician-owned groups carry 28% more nursing staff. Operating cost per FTE physician grew 15.8% (2021) and 19.9% (2022) at hospital-/IDS-owned groups against 7.3% and 7.6% at physician-owned groups. Nationally, 44.3% of physicians in hospital-owned practices are in multi-specialty practices, against 20.5% of private-practice physicians and 27.7% of PE-owned-practice physicians.

   MGMA 2014 Cost Survey and MGMA DataDive Cost and Revenue (2014 and 2022 data — the 2014 ratios are dated; the structural pattern persists in the 2022 growth rates); AMA Physician Practice Benchmark Survey 2024 (n=5,000, NORC-weighted, 43% response rate).

   NOT THIS IF — Not this in a group that grew de novo under one ownership model. Then every site inherited the same design and the variance is operational.

4. 04

   Payer mix and service-line mix differ by site more than anyone has measured

   Two sites with the same encounter volume can carry very different revenue per encounter because their payer mix and service-line composition differ. This is real variance, not fixable variance, and it must be netted out before the operational gap can be seen.

   Nationally, office-based physician visits split 44.2% private insurance, 38.7% Medicare, 11.7% Medicaid, 5.4% uninsured — visit-weighted, not revenue-weighted. Mean expense per office-based visit varies enormously by specialty: psychiatry $159, pediatrics $169, primary care $186, cardiology $335, orthopedics $419. And the overall mean of $265 sits against a median of $116 — any "average revenue per visit" claim that does not say which one it is, is unusable.

   AHRQ MEPS Statistical Brief #517, "Expenses for Office-Based Physician Visits by Specialty and Insurance Type, 2016" (Machlin & Mitchell, October 2018) — read at source. 2016 data; the most recent MEPS brief on this topic.

   NOT THIS IF — Not this if payer mix and service-line composition are within a few points across sites. Then the variance is not structural and the first two causes own it.

5. 05

   The sites are not measured on the same definitions

   Variance that lives in the definition rather than the operation. A site that counts encounters, FTEs, or support staff differently will rank differently for reasons that have nothing to do with performance.

   Stated qualitatively rather than with a borrowed figure: HFMA MAP Keys exist precisely because revenue-cycle definitions vary between organisations, and the definitions themselves are member-gated. There is no free public benchmark for contribution margin by service line — the metric MSOs actually run on — from MGMA, AMGA, or anyone else, so cross-site comparison here is necessarily internal.

   HFMA MAP Keys (member-gated definitions); recorded as a documented gap — contribution margin by service line has zero public benchmarks, and both MGMA and AMGA hold adjacent data neither publishes free.

   NOT THIS IF — Not this if a single analyst can reproduce every site's numbers from source data using one written definition. If they cannot, fix this before ranking anything.

HOW TO TELL THEM APART

How to tell these apart in your own numbers

Each of these is a measurement you can run yourself, without us.

WHAT RESOLVES EACH

What resolves this, and how you will know it resolved

Task alignment and staff skill, not staffing quantity	Rx 03 · patient conversion →	Front-desk feedback loop and call-centre instrumentation, applied to the lagging sites specifically. The MGMA finding is the whole argument: same inputs, sevenfold outcomes, so the constraint is operational rather than resource volume — and adding headcount to a misaligned site adds cost without adding output.
Staffing composition — the mix, not the count	Rx 03 · patient conversion →	Composition, not headcount. The published gap between high- and low-productivity groups concentrates in business operations staffing, which is the least intuitive place a clinical organisation looks.
Ownership structure differs across sites and carries its own cost curve	Rx 04 · marketing attribution →	Report acquired cohorts separately until their cost curve converges. Blending them produces a group average that describes no site. This is a reporting fix before it is an operating fix.
Payer mix and service-line mix differ by site more than anyone has measured	Rx 04 · marketing attribution →	Mix-normalised site comparison, monthly. Where the spread is mix, we would say so — that portion is not addressable by anything we do, and pretending otherwise is how a group ends up spending against a market difference.
The sites are not measured on the same definitions	Rx 04 · marketing attribution →	One page, one definition, every site, monthly. Cheapest fix on this list and the one most often skipped. Note that no public benchmark for contribution margin by service line exists, so the group's internal definition is the only standard there will be — write it down.

HOW THIS DIFFERS BY SCALE

How this differs by scale

Single site	One site has no variance problem — it has a benchmark problem, and the only available comparison is the MGMA distribution, which is paywalled beyond the free headline medians. Compare periods, not peers.
Group	Three to fifteen sites is where this differential does its most work: enough sites to rank, few enough that mix can be normalised by hand. Rank within staffing bands, not across them, or the ranking simply reproduces the staffing decision.
Platform	Fifty-plus providers across mixed ownership. Site variance is compounded by ownership variance, and Kaufman Hall's cohort-level pattern — wRVUs per FTE physician up 9% since 2023 while net patient revenue per wRVU is down — means the whole distribution is sliding while the internal spread persists. Measure the spread and the drift separately; a site improving against a falling market can look flat and be winning.

OTHER PRESENTATIONS — MULTI-SPECIALTY MSO

• wRVUs per FTE physician are up. Total medical revenue per FTE physician is not.
• Cost per encounter is rising — and the marketing line is the one number nobody publishes
• Referrals are being placed. Nobody can say whether the patient was seen.
• Growth has to read as referral capture, not as acquisitions