Phase Space AI

Research

U.S. Physical Therapy [USPH]

U.S. Physical Therapy [USPH] — Research

Tier 2 · as of 2026-07-30 · spot $79.675 · FY ends 31 December

Screen record: reports/scan_universe/USPH_analysis.json — Quality PASS, Valuation PASS, +8.0pp. This memo finds the screen's EV understated by 23.6% because it omits $313.9m of non-controlling interest, and the valuation margin flips from +8.0pp to −8.9pp. That single correction is the largest defect found in this three-name run.

---

1. NON-CONTROLLING INTERESTS — the large, easily-missed claim on earnings

USPH does not own its clinics outright. It buys majority interests — 60%, 65%, 75% — in clinic partnerships and consolidates them in full, with the partners' share carried as non-controlling interest. From the FY2025 10-K's own acquisition table: July 2025 (60%, 3 clinics), April 2025 (40%*), February 2025 (65%, 3 clinics), November 2024 (75%, 8 clinics).

How large the claim is

FY2025 (10-K, verbatim table):

NCI takes 31.8% of consolidated net income. On the TTM to 2026-03-31 it is 34.1% (consolidated ProfitLoss $52.691m, NCI $17.969m, attributable $34.722m).

There is a second-order effect that is easy to miss and USPH spells it out: the tax provision is computed on income after NCI, because the partnerships are pass-throughs. The 10-K shows an effective rate of 33.4% against income-after-NCI of $59.391m — while the rate on consolidated pretax income is 25.5%. A model that applies the disclosed 33.4% to consolidated EBIT double-charges tax. Both rates are stated in this memo and the consolidated one is used.

The balance-sheet claim, and why the screen missed it

$313.9m against a market capitalisation of $1,212.6m — 25.9% of the equity value. And it is growing: $260.0m (2025-03-31) → $263.3m → $277.7m → $293.3m → $313.4m in five quarters, +20.5%.

Redeemable NCI sits in temporary equity — the mezzanine — between liabilities and equity on the balance sheet. That is exactly why an automated EV build misses it: it is neither in a Liabilities roll-up nor in StockholdersEquity. The screen's net_cash_components names only CashAndCashEquivalentsAtCarryingValue and LongTermDebtNoncurrent. NCI never entered the calculation.

The correct EV, and the correct EV/EBIT

The rule is consistency: if you value consolidated EBIT, you must use an EV that includes every claim on it. USPH's reported operating income is fully consolidated — it includes the NCI partners' share — so the EV must include NCI.

The multiple is 37% higher than the screen reported. The screen's ev_ebit of 15.6x is not approximately right; it is a different number about a different company.

Lease basis, stated explicitly: the above is lease-EXCLUSIVE. USPH carries $115.212m non-current + $42.779m current = $157.991m of operating lease liabilities — 9.3% of the corrected EV, and unavoidable for a business that is 776 leased clinic locations. A lease-inclusive EV would be $1,860.85m. Lease-exclusive is used as the primary basis because under ASC 842 the rent is already inside operating cost and therefore inside EBIT; adding the liability to EV while leaving the rent in EBIT would double-count. Stated because the CRWV precedent (6.6x quoted, really 7.95x) invalidated a seven-name ladder in this project for exactly this omission.

Per-share treatment

Every per-share figure in this memo is on shares attributable to USPH shareholders only (15,218,789 outstanding, diluted weighted-average 15,167k) against income attributable to USPH shareholders only. The two must not be mixed: consolidated net income ÷ USPH shares would overstate EPS by 51.7%.

AV's OVERVIEW does exactly that mixing in reverse and gets it wrong the other way — it reports EPS 0.50 and TrailingPE 159.7, which is a single quarter's attributable EPS annualised against nothing coherent. Ignored.

---

2. THE ECONOMICS: volume, rate, and clinician cost

USPH discloses these directly, and the disclosure is unusually good.

FY2025 — physical therapy segment (10-K)

The three-part decomposition, exactly as the brief asks for it:

1. Volume: +14.9% — and it is the whole story. But 96 owned clinics were added in 2024 and 47 in 2025, so this is clinic count, not clinics getting busier. Same-clinic productivity (average daily visits per clinic) rose only +5.9%.
2. Rate: +1.0%. One percentage point. Against clinician wage inflation this is the binding constraint on the model.
3. Clinician cost: +0.5% per visit. Better than rate growth by 0.5pp — which is the only reason gross margin expanded 0.8pp. USPH earned its margin expansion by holding wages below a 1% rate increase, not by pricing. That is not a durable source of leverage.

Q1 2026 — and here the volume story deteriorates

The 10-Q segments revenue into mature clinics, additions, and closures. This is the same-clinic number and it is the most important table in the file:

Same-clinic (mature) revenue growth is +2.5%. The consolidated 7.9% headline is 2.5pp same-clinic plus 6.5pp of clinic additions less 1.1pp of closures. Average daily visits per clinic decelerated from +5.9% (FY2025) to +1.9% (Q1 2026).

And PT gross profit grew only +2.1% on +7.2% revenue — margin compressed from 16.6% to 15.8% in the segment. Corporate office costs rose to $18.3m from $16.2m, which the 10-Q attributes to "increased clinic count, expenses related to acquisition integration, and the implementation of a new financial and human resources system." Consolidated Q1 2026 operating margin was 6.29% against 10.00% TTM and 11.10% in FY2025.

The one genuinely good line: IIP same-store revenue +8.2% with margin up 1.8pp. Industrial injury prevention is contracted directly with employers, not payors, so it escapes the Medicare rate problem entirely. It is 15.4% of Q1 2026 revenue and it is the best-growing part of the company.

---

3. REIMBURSEMENT — dated, disclosed, and finally turning

USPH's Medicare disclosure is the best of the three names in this cluster (RMD discloses no rate schedule at all; UFPT has no reimbursement exposure).

FY2025 10-K, verbatim:

"For calendar years 2021, 2022, 2023 and 2024, the MPFS had decreases in Medicare reimbursement of approximately 3.5%, 0.75%, 2.0% and 1.8%, respectively. The MPFS for 2025 resulted in a decrease in Medicare reimbursement for therapy services…"

and, separately:

"For services provided in 2026, we expect our reimbursement rates under the MPFS to increase by approximately 1.75% as compared to the applicable reimbursement rates during 2025."

Five consecutive years of Medicare rate cuts, then a +1.75% increase for 2026 — the first positive year in six. That is a dated, disclosed, company-quantified catalyst and it is already partly visible: net rate per visit is +0.8% in Q1 2026.

Exposure: Medicare net patient revenues were $213.5m in FY2025 against $183.4m in FY2024 — 32.8% of PT net patient revenue ($650.4m), and rising in share. So a 1.75% rate increase on 32.8% of the base is worth roughly +0.57% to consolidated revenue, or about $4.5m — meaningful against $79.5m of TTM EBIT, but not transformative. The rate lever is real and it is small. Anyone underwriting a Medicare-driven re-rating should hold that number in mind.

The annual rate-setting cycle is itself a standing risk: the 10-K's first listed risk factor is "changes in Medicare rules and guidelines," and rates are reset every calendar year with the proposed rule typically published mid-year and finalised in the autumn.

---

4. De novo versus acquired clinic growth

USPH does not use the term "de novo" anywhere in the FY2025 10-K or the Q1 2026 10-Q — searched explicitly, zero hits. It discloses "clinics added," which conflates newly opened with acquired, and the acquisition table separately lists the interests purchased.

What can be established:

• 96 owned clinics added in FY2024, 47 in FY2025 (10-K footnote to the segment table).
• 45 underperforming clinics closed in FY2024, 23 in FY2025. Clinic closure costs $4.4m (2024) and $0.3m (2025).
• Net clinic additions are therefore roughly 51 (2024) and 24 (2025) — the pace of expansion has halved.
• Acquisitions in the last three years each bought a partial interest (60%, 65%, 75%, 40%), which is exactly why NCI compounds at 20% a year while attributable earnings do not.

The composition of "clinic additions" between de novo and acquired is not disclosed and I will not estimate it. Recorded as INDETERMINATE with the reason. It matters: a de novo clinic is 100%-owned and adds nothing to NCI; an acquired clinic is 60–75% owned and adds to the NCI claim on every subsequent dollar of its earnings. The observed 20.5% growth in redeemable NCI over five quarters is indirect evidence that the additions are predominantly acquired.

---

5. Accounting quality — DSO, factoring, retired metrics

Same-quarter DSO — and it is clean

Use EDGAR AccountsReceivableNetCurrent. AV's currentNetReceivables is $96.724m against EDGAR's $69.082m — 40.0% higher (see §7).

A 31.8-day DSO in a reimbursement business is genuinely good and it improved year on year. For context this project has recorded AAOI at 129.7–180.5 days. USPH collects fast.

Receivables-driven quality is the key tell for a reimbursement business because contractual allowances are an estimate: USPH recognises "net patient revenues (patient revenues less estimated contractual adjustments)… at the estimated net realisable amounts," and the auditor designated "Measurement of Patient Revenue Net of Contractual Adjustments" a CRITICAL AUDIT MATTER. That is the auditor flagging the single largest estimation risk in the statements. The provision for credit losses was $7.6m in FY2025 and $6.9m in FY2024 — 1.2% of net patient revenues in both years, i.e. stable. Combined with the falling DSO, the estimate does not appear to be being stretched.

Factoring check

Searched the FY2025 10-K for factoring, securitiz, sold … receivable: zero hits on all three. No receivables-sale line in the cash-flow statement. The DSO improvement is therefore real — which is the statement the CLS finding exists to force, because on CLS a 7.3-day reported improvement was in fact a deterioration of up to 3.6 days once $565m of sold receivables was added back.

Retired or redefined metrics — one found, and it is subtle

"Mature clinics" is redefined every single year, by construction. FY2025 10-K: "Mature clinics are clinics… opened or acquired prior to January 1, 2024." Q1 2026 10-Q: "opened or acquired prior to January 1, 2025."

That is a legitimate rolling definition, and USPH is transparent about it. But it means the same-clinic growth rate is not comparable across years, and it means a clinic acquired in 2024 that is now underperforming gets folded into "mature" the following year, diluting the base. The +2.5% mature-clinic growth in Q1 2026 is measured against a base that now includes the 2024 acquisitions. Recorded, not alleged as manipulation.

Nothing else was found to have been withdrawn: net rate per visit, patient visits, average daily visits per clinic, adjusted salaries per visit and adjusted operating costs per visit are all still disclosed, quarterly, with prior-year comparatives. On disclosure quality USPH is the best of the three names in this run, and that deserves to be said alongside the valuation conclusion.

---

6. Company state

STATE A — mature and structurally stable.

Not B — there is no exogenous cycle; the Medicare rate schedule is a policy variable, not a cycle, and it is set administratively each year. Not C — profitable with an observable margin, not scaling. Not D — the reverse DCF solves.

evidence_grade: B. Filings current (10-Q filed 2026-05-08). Deductions: transcript coverage is 6 of 16 quarters and all six are 2023–2024 — nothing from 2025 or 2026 is available, so the mention-frequency work cannot cover the period that matters; the de novo/acquired split is not disclosed; and AV's statement data for USPH is the most defective of the three names in this run (§7).

---

7. Defects found — and USPH is where AV broke worst

The USPH lesson for the corpus: AV's annual statements were reliable and its quarterly series was not, on the same name, in the same response. A pipeline that builds TTM from AV quarterlies would have reported $729.2m of revenue against a true $795.5m — 8.3% low — and a TTM net income of $1.2m against $34.7m. The screen escaped this only because it used annual figures for margins and rebuilt TTM revenue from EDGAR. Neither source is safe on its own; the disagreement is the signal.