Corpus Intelligence ML Insights 2026-04-26 00:39 UTC
ML Insights
6,123 hospitals | 7 archetypes | Distress AUC 0.629 | 25 high-risk
🛡️ Public data only — no PHI permitted on this instance.
6,123
Hospitals Analyzed
7
Archetypes
25
High Distress Risk
0.629
Distress Model AUC
-4.8%
Median Op Margin

Hospital Archetypes (K-Means Clustering)

Unsupervised clustering of 6,123 US hospitals into 7 investable archetypes based on size, revenue, margins, payer mix, and occupancy. Each cluster has a distinct risk/return profile for PE evaluation.

Suburban Community Hospital
1457 hospitals
suburban_community
Beds: 70
Revenue: $107.8M
Margin: 6.1%
Medicare: 35%

Community hospitals — the largest PE deal category. Focus on RCM improvement and cost optimization at 9-11x.

Representative: PROMEDICA HICKMAN HO, WATAUGA MEDICAL CENT, ST LUKES HOSPITAL -
Suburban Community Hospital
1272 hospitals
suburban_community
Beds: 256
Revenue: $418.0M
Margin: -1.2%
Medicare: 23%

Community hospitals — the largest PE deal category. Focus on RCM improvement and cost optimization at 9-11x.

Representative: CLARA MAASS MEDICAL , PRESENCE SAINTS MARY, ADVENTIST HEALTH BAK
Under-Performing / Distressed
837 hospitals
distressed
Beds: 93
Revenue: $80.0M
Margin: -29.5%
Medicare: 27%

Deeply negative margins signal severe distress. Evaluate asset-level acquisition at 4-6x normalized EBITDA.

Representative: HOLY FAMILY MEMORIAL, MIDDLESBORO ARH, ATRIUM HEALTH NAVICE
Rural/Critical Access
788 hospitals
rural_cah
Beds: 29
Revenue: $30.4M
Margin: -9.9%
Medicare: 62%

Rural/small hospitals face structural headwinds. Evaluate CAH conversion, telehealth, and rural health funding.

Representative: BRECKINRIDGE HEALTH , SHERIDAN MEMORIAL HO, DAYTON GENERAL HOSPI
Safety-Net/Medicaid Heavy
345 hospitals
safety_net
Beds: 118
Revenue: $143.7M
Margin: -10.4%
Medicare: 15%

High Medicaid dependence creates reimbursement risk. Assess DSH payments and state expansion status.

Representative: WESTERN PEAKS SPECIA, HOSPITAL MENONITA AI, TOPPENISH COMMUNITY
Large Academic Medical Center
202 hospitals
large_academic
Beds: 795
Revenue: $2.06B
Margin: -5.9%
Medicare: 21%

Large medical centers trade at premium multiples (12-14x). Limited PE value creation but strong cash flow.

Representative: NORTH CAROLINA BAPTI, HACKENSACK UNIVERSIT, TEMPLE UNIVERSITY HO
Under-Performing / Distressed
6 hospitals
distressed
Beds: 16
Revenue: $715.5M
Margin: -22.1%
Medicare: 37%

Deeply negative margins signal severe distress. Evaluate asset-level acquisition at 4-6x normalized EBITDA.

Representative: PARKLAND HLTH CTR - , LAGUNA HONDA HOSPITA, WENATCHEE VALLEY HOS

Distress Risk Screening (Logistic Regression)

Hospitals ranked by predicted probability of financial distress (operating margin < -5%). Model AUC = 0.629 on 4,907 training samples. High-distress hospitals are potential turnaround acquisition targets at discounted multiples.

HospitalStateBedsRevenueMarginDistress PRisk
HILL HOSPITAL OF SUMTER COUNTYAL33$3.7M-31.6%75.1%Critical
NORTHEAST MONTANA HEALTHCARE-POPLARMT20$12.7M4.8%74.6%High
GREENE COUNTY HOSPITALAL20$9.9M-16.3%74.3%High
PREMIERCARE TENNESSEE INC.TN36$1.2M-50.0%72.9%High
NORTHEAST MONTANA HEALTH SERVICESMT22$35.4M-10.2%72.3%High
SOUTHERN PLAINS MED CTR OF GARVIN COOK43$28.3M-9.8%71.0%High
LAKE TAYLOR TRANSITIONAL CARE HOSPITVA104$51.4M14.6%70.9%High
LINDSAY MUNICIPAL HOSPITAL AUTHORITYOK26$16.5M-2.0%70.9%High
BRENTWOOD BEHAVIORAL HEALTHCAREMS105$23.9M-4.7%70.8%High
PORTERVILLE DEVELOPMENTAL CENTERCA17$193.6M-6.0%70.6%High
ODESSA MEMORIAL HOSPITALWA25$9.7M-23.6%70.2%High
SAGE MEMORIAL HOSPITALAZ25$24.0M-50.0%69.8%High
BULLOCK COUNTY HOSPITALAL29$13.9M-24.4%69.6%High
MEADOWBROOK REHABILITATION HOSPITALKS54$23.2M0.7%69.5%High
RJ BLACKLEY ADATCNC60$13.8M-39.6%69.5%High
FLINT RIVER COMMUNITY HOSPITALGA22$8.3M-10.5%69.5%High
COALINGA REGIONAL MEDICAL CENTERCA22$1.8M-50.0%69.5%High
BIG HORN HOSPITALMT18$12.8M-49.5%69.5%High
MIRAMOUNT BEHAVIORAL HEALTHWI72$9.3M-21.8%69.3%High
CORDOVA COMMUNITY MEDICAL CENTERAK13$12.3M-5.5%69.0%High

RCM Performance Screening (Predicted from Public Data)

Hospitals with the worst predicted RCM metrics — highest denial rates, longest AR days. These are potential PE targets where RCM improvement could create the most value. Predictions use HCRIS financials + payer mix + geography only (no internal data needed).

HospitalStateBedsEst DenialEst AR DaysEst Clean ClaimRCM Score
ASPEN HILLS HEALTHCARE CENTERNJ3025.0%75d98.0%50
FALLBROOK HOSPITALCA025.0%75d98.0%50
SAN GORGONIO MEMORIALCA7925.0%75d98.0%50
ST. JOSEPH HOSPITAL OF ORANGECA31525.0%75d98.0%50
ASCENSION SAINT THOMAS REHABILITATITN4025.0%75d98.0%50
QUINCY VALLEY MEDICAL CENTERWA1025.0%75d98.0%50
LTAC OF LOUISIANA LLCLA1825.0%75d98.0%50
BROOKHAVEN MEADOWBROOK HOSPITALOK025.0%75d98.0%50
ENCOMPASS HEALTH REHABILITATION HOSTX8425.0%75d98.0%50
CENTRAL LA. STATE HOSPITALLA14825.0%75d98.0%50
BAKERSFIELD REHABILITATION HOSPITALCA5025.0%75d98.0%50
KINDRED HOSPITAL SOUTH FLORIDAFL19725.0%75d98.0%50
OCEANS BEHAVIORAL HOSPITAL OF HAMMOLA5225.0%75d98.0%50
SUMMITRIDGE HOSPITALGA9625.0%75d98.0%50
THE PHYSICIAN CENTRETX1625.0%75d98.0%50
REHABILIATION HOSPITAL OF NAPLESFL5025.0%75d98.0%50
MADONNA REHABILITATION LTC HOSPITALNE7725.0%75d98.0%50
SSH - JOHNSTOWN INC.PA3925.0%75d98.0%50
BEHAVIORAL HEALTH CENTERSTN1625.0%75d98.0%50
SSH WILLINGBORONJ6925.0%75d98.0%50

Proprietary Models

Hospital Clustering

K-means on 7 standardized features (beds, revenue, margin, Medicare %, Medicaid %, occupancy, revenue/bed). Clusters labeled by centroid characteristics. Pure numpy — no sklearn dependency.

Distress Predictor

L2-regularized logistic regression predicting P(margin < -5%). Trained on cross-sectional HCRIS data. Features: occupancy, Medicare %, Medicaid %, revenue/bed, net-to-gross ratio, beds. AUC validated on held-out data.

RCM Opportunity Scorer

Gap analysis across 6 RCM levers: denial reduction, AR acceleration, clean claim rate, net-to-gross improvement, payer mix optimization, occupancy. Each lever benchmarked against P75 peers with 60% gap closure assumption and confidence weighting.

Conformal Prediction

Distribution-free 90% prediction intervals via split conformal inference. Guarantees finite-sample coverage — every point estimate comes with a calibrated uncertainty band, not just a standard error.