Results

Model outputs, diagnostics, and visualisations. All figures are export-ready.

Moderate confidence

· Overall model confidence

Model fit is adequate (AUC 0.781, HL p = 0.39, Nagelkerke R² 0.241). Two caveats flagged: EPP below recommended threshold (7.8) and 2.6% missing data on key predictors. Primary conclusions are robust; a multiple imputation sensitivity run is recommended before finalizing.

Key Findings

Patients in the intervention arm (Group B) had 85% higher odds of 90-day survival vs standard care (OR 1.85, 95% CI 1.31–2.62, p < 0.001).

Each additional point on the baseline clinical score was associated with a 4.2% increase in survival odds (OR 1.042 per point, 95% CI 1.021–1.063, p < 0.001).

Higher comorbidity burden was strongly associated with reduced survival (OR 0.73 per CCI unit, 95% CI 0.66–0.82, p < 0.001).

AUC

0.781

n (analysis)

820

Nagelkerke R²

0.241

AIC

651.4

Coefficient Table

Logistic Regressionn = 820

Term	β	OR	95% CI	p-value
Intercept	-2.419	0.089	[0.031, 0.257]	<0.001
treatment_group[B]	0.615	1.851	[1.308, 2.620]	<0.001
age	-0.027	0.973	[0.958, 0.988]	<0.001
sex[F]	-0.209	0.811	[0.571, 1.152]	0.244
baseline_score	0.041	1.042	[1.021, 1.063]	<0.001
comorbidity_index	-0.309	0.734	[0.658, 0.819]	<0.001
site[S2]	0.223	1.250	[0.782, 1.998]	0.350
site[S3]	-0.181	0.834	[0.513, 1.353]	0.463
treatment_group[B]:baseline_score	0.018	1.018	[0.996, 1.041]	0.107

Significant (p < 0.05)Non-significant

ROC Curve

Bootstrap 95% CI: [0.737, 0.826] · 1 000 replicates

Calibration Plot

Hosmer–Lemeshow χ²(8) = 8.43, p = 0.392

Coefficient Forest Plot

Odds ratios with 95% CI. Reference: treatment_group=A, sex=M, site=S1.

Diagnostics

Checks complete

Variance Inflation Factors (VIF)

Predictor	VIF	Status
treatment_group	1.42	OK
age	1.18	OK
sex	1.09	OK
baseline_score	2.31	OK
comorbidity_index	1.07	OK
site	1.22	OK
treatment_group:baseline_score	2.48	OK

Max VIF = 2.48 (threshold for concern: 10). No action required.

Influence (Cook's D): max = 0.047 across 820 observations. No highly influential records detected (threshold: 4/n = 0.0049).

⚠

EPP = 7.8, below recommended threshold of 10. Interpret with appropriate caution.

ℹ

22 observations excluded via listwise deletion. Multiple imputation sensitivity analysis recommended.

Export Bundle

Ready

All outputs packaged and ready to download.