Predictive markers related to local and systemic inflammation in severe COVID-19-associated ARDS: a prospective single-center analysis

Table 3 Multiple logistic regression models for local and systemic damage as well as quality of life

		AUC	PPP	NPP	Tjur‘s R2	n =	Variable	CCI	Sex [w]	1st PCT	IL-8 BAL
Model local damage	Horovitz min	0.9091	100.00	80.00	0.5754	19	OR	0.8183	7.423	593.8	0.9998
	> 72 mmHg = “negative”	95%-CI: 0.7668 to 1.000					95%CI	0.1032 to 2.989	0.2649 to 645.5	3.385 to 4.01E + 06	0.9991 to 1.000
		p = 0.0030					p	0.7941	0.2670	0.0601	0.3881
	Horovitz mean	0.7460	83.33	80.00	0.2456	16	OR	1.493	2.532	1.468	1.000
	> 150 mmHg = “negative”	95%-CI: 0.4774 to 1.000					95%CI	0.5036 to 4.709	0.1524 to 47.53	0.1687 to 8.105	1.000 to 1.001
		p = 0.1009					p	0.4392	0.5048	0.6477	0.1512
	ECMO/ECLS	0.9659	100.00	91.67	0.7312	19	OR	0.3400	0.07332	57.054	0.9989
	No ECMO/ECLS = “negative”	95%-CI: 0.8889 to 1.000					95%CI	0.01735 to 1.634	1.74E-07 to 34.45	8.282 to 6.22E + 13	0.9968 to 0.9998
		p = 0.0007					p	0.3368	0.5010	0.0787	0.0914
	Dialysis	0.9889	100.0	90.91	0.8074	19	OR	1.49E-17	4.00E + 20	77.444	1.004
	No Dialysis = “negative”	95%-CI: 0.9538 to 1.000					95%CI	6.21E-76 to 0.07919	1.060 to 3.49E + 81	2.082 to 2.29E + 18	0.9999 to 1.015
		p = 0.0003					p	0.3276	0.3572	0.3261	0.3611

		AUC	PPP	NPP	Tjur‘s R2	n =	variable	CCI	Sex [w]	TLR3 blood	IL-6 BAL
Model systemic damage	RASS mean	0.9167	87.50	77.78	0.5197	17	OR	1.146	1.650	0.1243	0.9997
	> − 4 = “negative”	95%-CI: 0.7822 to 1.000					95%CI	0.1782 to 8.219	0.01546 to 339.3	0.005356 to 0.6342	0.9970 to 1.001
		p = 0.0039					p	0.8797	0.8306	0.0693	0.6077
	Sedation	0.9722	100.0	88.89	0.7544	17	OR	2.700	2.07E-05	0.02232	1.007
	< 0.5 = “negative”	95%-CI: 0.9034 to 1.000					95%CI	3.286 to 5.59E + 16	2.28E-20 to 0.3901	1.61E-10 to 0.7858	1.001 to 1.028
		p = 0.0011					p	0.2463	0.2338	0.2972	0.1879
	mRS 6 m	0.9429	90.00	85.71	0.6199	17	OR	41.05	0.003212	0.008457	0.9987
	< 2 = “negative”	95%-CI: 0.8363 to 1.000					95%CI	1.712 to 1.38E + 06	2.45E-12 to 2.016	5.20E-09 to 0.3103	0.9952 to 0.9999
		p = 0.0025					p	0.1554	0.2462	0.1396	0.1540
	Thromboembolic events	0.7917	80.00	85.71	0.2705	17	OR	0.8522	0.6872	1.449	0.9988
	No event = “negative”	95%-CI: 0.5403 to 1.000					95%CI	0.1786 to 3.677	0.01146 to 23.33	0.5591 to 4.175	0.9967 to 1.000
		p = 0.0433					p	0.8270	0.8360	0.4452	0.1840
	Mortality	0.9444	87.50	77.78	0.5711	17	OR	28.99	0.04563	0.05985	0.9990
	Not deceased = “negative”	95%-CI: 0.8406 to 1.000					95%CI	1.951 to 4733	2.76E-05 to 4.201	5.79E-04 to 0.4601	0.9974 to 1.000
		p = 0.0021					p	0.0651	0.2416	0.0710	0.0895

		AUC	PPP	NPP	Tjur‘s R2	n =	Variable	CCI	Sex [w]	1st PCT
QoL 6 m	EQ-5D-5L index	0.8942	87.50	92.31	0.5371	21	OR	2.813	8.630	1.291
	> 0.7 = “negative”	95%-CI: 0.7393 to 1.000					95%CI	1.080 to 11.45	0.4123 to 384.3	1.086 to 1.672
		p = 0.0030					p	0.0661	0.1893	0.0138

Local damage \(\sim\) Intercept + CCI + Sex [w] + 1st PCT + IL-8 BAL. Dependent variables for local damage: Horovitz min > 72 mmHg [favorable] vs. < 72 mmHg [unfavorable], Horovitz mean > 150 mmHg [favorable] vs. < 150 mmHg [unfavorable], ECMO/ECLS support yes/no, dialysis yes/no. [unfavorable] = “positive” outcome
Systemic damage \(\sim\) Intercept + CCI + Sex [w] + TLR3 blood + IL-6 BAL. Dependent variables for systemic damage: RASS mean > − 4 [favorable] vs. RASS mean < − 4 [unfavorable], need for sedation < 0.5 [favorable] vs. > 0.5 [unfavorable], mRS after six months [mRS 6 m] 0–2 [favorable] vs. mRS 6 m 3–6 [unfavorable], thromboembolic events yes/no, mortality yes/no. [unfavorable] = “positive” outcome
QoL \(\sim\) Intercept + CCI + Sex [w] + 1st PCT. EQ-5D-5L index > 0.7 [favorable] vs. < 0.7 [unfavorable]. [unfavorable] = “positive” outcome
Comparison method: Akaike’s Information Criterion, Tjur’s R squared, Hosmer–Lemeshow hypothesis test. Intercept not shown. Statistically significant AUC or OR as well as Tjur’s R² > 0.5 are highlighted in bold. RASS Richmond Agitation-Sedation Scale, mRS modified Rankin Scale, AUC area under the curve, CI confidence interval, PPP positive predictive power, NPP negative predictive power, OR odds ratio

ISSN: 1471-2334