Corpus Intelligence Regression Analysis 2026-04-26 00:39 UTC
Regression Analysis
OLS: Net Patient Revenue ~ 17 features | R² = 81.0% | n = 4,907 | 15 significant
🛡️ Public data only — no PHI permitted on this instance.
81.0%
80.9%
Adj R²
4,907
Observations
17
Features
1225.9
F-Statistic
$218.8M
RMSE (avg error)

R² = fraction of variance explained (1.0 = perfect, 0 = no signal). RMSE = root mean squared error (average prediction miss in target units). F-statistic tests whether the model explains more than random chance.

Intercept Interpretation

Value: $254.8M
SE: $3.1M
Target Mean: $254.8M
Target Range: 386,584 — $8.94B

When all features are at their mean values, the predicted net patient revenue is 254,770,179.37

Coefficients (Standardized)

Target: Net Patient Revenue. Standardized coefficients (-1.0 to +1.0): a one-SD increase in the feature produces this fraction of the strongest effect. *** p<0.001, ** p<0.01, * p<0.05.

VariableStrengthtp-valueSig95% CIImpact
Total Patient Days+1.00011.40.0000***[324618339.745, 459377824.321]
Revenue Per Bed+0.87429.30.0000***[319831117.133, 365627349.894]
Expense Per Bed-0.61321.00.0000***[-262894582.962, -218065024.587]
Medicare Days+0.5056.20.0000***[135104793.894, 260819624.751]
Beds+0.4404.60.0000***[99002783.909, 245581257.134]
Bed Days Available-0.4306.60.0000***[-218578065.302, -118736128.721]
Medicare Intensity-0.3414.20.0000***[-195191397.950, -71909323.796]
Occupancy Rate-0.17213.80.0000***[-77209433.056, -58001483.780]
Revenue Per Day-0.0728.50.0000***[-34733794.583, -21752026.401]
Operating Margin-0.0606.40.0000***[-30912694.504, -16366274.440]
Medicaid Days-0.0584.50.0000***[-32599739.876, -12718214.715]
Size Quartile+0.0402.60.0084**[4018470.765, 27339581.020]
Medicaid Day Pct+0.0353.80.0002***[6535101.795, 20830932.977]
Net To Gross Ratio+0.0282.90.0039**[3538700.062, 18496681.237]
Commercial Pct-0.0192.60.0096**[-13310185.633, -1844413.092]
Payer Diversity-0.0171.50.1279ns[-15556357.553, 1954287.625]
Medicare Day Pct-0.0030.40.6920ns[-6609697.347, 4387163.271]

Univariate Correlations with Target

Pearson r for each feature vs Net Patient Revenue. Shows raw linear relationship before controlling for other variables.

FeaturerStrength
Total Patient Days0.857Strong
Bed Days Available0.827Strong
Beds0.816Strong
Medicare Days0.795Strong
Medicare Intensity0.739Strong
Medicaid Days0.523Moderate
Size Quartile0.472Moderate
Occupancy Rate0.376Weak
Revenue Per Bed0.281Weak
Medicare Day Pct-0.222Weak
Expense Per Bed0.217Weak
Commercial Pct0.215Weak
Net To Gross Ratio-0.164Weak
Operating Margin0.072Weak
Payer Diversity-0.038Weak
Medicaid Day Pct0.012Weak
Revenue Per Day-0.003Weak

Hospital Outliers (Residual Analysis)

Hospitals with the largest standardized residuals. >2σ = model underpredicts/overpredicts — investigate for deal opportunities or data quality issues.

HospitalStateActualPredictedResidual
ST. LUKES HOSPITALPA$8.94B$3.39B+25.37σ
STANFORD HEALTH CARECA$6.76B$2.77B+18.25σ
CLEVELAND CLINIC HOSPITALOH$6.38B$3.08B+15.08σ
UCSF MEDICAL CENTERCA$5.44B$2.51B+13.39σ
NYU LANGONE HOSPITALSNY$7.24B$4.42B+12.89σ
PARKLAND HLTH CTR - BONNE TERREMO$124.2M$2.80B-12.23σ
MEMORIAL HOSPITAL FOR CANCER AND ALNY$4.34B$1.80B+11.62σ
UT MD ANDERSON CANCER CENTERTX$4.90B$2.39B+11.45σ
VANDERBILT UNIVERSITY MEDICAL CENTETN$5.44B$3.08B+10.81σ
UNIV OF MI HOSPITALS & HLTH CTRSMI$4.62B$2.69B+8.81σ
METHODIST HOSPITALTX$2.42B$4.07B-7.58σ
UC DAVIS MEDICAL CENTERCA$3.28B$1.81B+6.73σ
WENATCHEE VALLEY HOSPITALWA$277.5M$1.74B-6.68σ
MIDWESTERN REGIONAL MEDICAL CENTERIL$1.38B$2.79B-6.43σ
JACKSON MEMORIALFL$1.47B$2.87B-6.39σ

Variance Inflation Factors

VIF > 10 = severe multicollinearity (coefficient estimates unreliable). VIF > 5 = moderate. Consider removing high-VIF features.

FeatureVIFStatus
Medicare Day Pct999.0High
Medicaid Day Pct999.0High
Commercial Pct999.0High
Beds143.4High
Total Patient Days121.2High
Medicare Days105.5High
Medicare Intensity101.4High
Bed Days Available66.5High
Revenue Per Bed14.0High
Expense Per Bed13.4High
Size Quartile3.6OK
Medicaid Days2.6OK

Model Fit by State

How well the national model predicts within each state. Low R² states have unique market dynamics not captured by national features — consider state-specific models.

StatenMean Residual
CT3297.8%$27.0M
RI1296.4%$16.4M
NM3695.9%$19.3M
VT1495.8%$23.8M
KY11094.9%$16.1M
KS9593.3%$21.5M
WV5793.2%$10.0M
MN12392.8%$-14.5M
HI1892.3%$-14.7M
NC11892.1%$-16.6M
VA8992.0%$16.4M
ND4191.7%$28.5M
ME3690.6%$25.6M
SD5489.3%$4.5M
LA14389.1%$24.0M

Top Pairwise Correlations

Variable 1Variable 2r
BedsBed Days Available0.984
Total Patient DaysBed Days Available0.976
BedsTotal Patient Days0.962
Medicare DaysMedicare Intensity0.955
Revenue Per BedExpense Per Bed0.948
Total Patient DaysMedicare Days0.899
BedsMedicare Intensity0.879
Medicare DaysBed Days Available0.871
Bed Days AvailableMedicare Intensity0.860
BedsMedicare Days0.859
Total Patient DaysNet Patient Revenue0.857
Total Patient DaysMedicare Intensity0.844
Bed Days AvailableNet Patient Revenue0.827
BedsNet Patient Revenue0.816
Medicare Day PctCommercial Pct-0.799