dist_km_vs_cox.png

# --- 5) Survival: Kaplan-Meier vs. Cox-Regression ---
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(9.5, 4))
fig.subplots_adjust(wspace=0.35, bottom=0.18)

# Kaplan-Meier: Raw step function curves (unadjusted)
t_km = np.arange(0, 31, 2)
s1 = np.exp(-0.02 * t_km)
s2 = np.exp(-0.06 * t_km)
ax1.step(t_km, s1, where="post", color=PRIMARY, linewidth=2, label="Therapie A (unbereinigt)")
ax1.step(t_km, s2, where="post", color=EVENT, linewidth=2, label="Therapie B (unbereinigt)")
ax1.set_title("Kaplan-Meier & Log-Rank\n(Unbereinigte Überlebenskurven)")
ax1.set_xlabel("Zeit (Tage)")
ax1.set_ylabel("Überlebenswahrscheinlichkeit")
ax1.set_ylim(0, 1.05)
ax1.legend(loc="lower left")
ax1.grid(True, linestyle=":", alpha=0.6)

# Cox-Regression: Hazard ratio plot (Forest Plot, multivariable adjusted)
variables = ["Therapie B (vs A)", "Alter (+10 J.)", "SOFA-Score (+1)"]
hrs = [2.4, 1.25, 1.15]
cis = [[1.5, 3.8], [1.05, 1.5], [1.02, 1.3]]

for idx, (var, hr, ci) in enumerate(zip(variables, hrs, cis)):
    ax2.errorbar(hr, idx, xerr=[[hr - ci[0]], [ci[1] - hr]], fmt="o", color=PRIMARY,
                 markersize=8, elinewidth=2, capsize=4)
ax2.axvline(1.0, color="#6B7178", linestyle="--", linewidth=1.2)
ax2.set_yticks(range(len(variables)))
ax2.set_yticklabels(variables)
ax2.set_title("Cox Proportional Hazards\n(Adjustierter Effekt mehrerer Faktoren)")
ax2.set_xlabel("Hazard Ratio (HR) mit 95% KI")
ax2.set_ylim(-0.5, 2.5)
ax2.set_xlim(0.5, 4.5)
ax2.grid(True, linestyle=":", alpha=0.6)

save(fig, M21_DIR / "dist_km_vs_cox.png")