LASSO-Based Identification of Risk Factors and Development of a Prediction Model for Sepsis Patients

Figures & data

Table 1 Demographic Characteristics of the Patients

	Survival Group	Non-Survival Group	p value
Case, n (%)	1351	382	-
Gender, n (%)			<0.001
Male	761(56.3)	252(66)
Female	590(43.7)	130(34)
Age, M±SD (y)	61.48±19.32	59.99±20.58	<0.001
Diagnosis
Biliary infection, n (%)			0.142
No	1328(98.3)	379(99.2)
Yes	23(1.7)	3(0.8)
Severe pneumonia, n (%)			0.052
No	481(35.6)	154(40.3)
Yes	870(64.4)	228(59.7)
Abdominal infection, n (%)			0.469
No	1287(95.3)	363(95)
Yes	22(4.7)	19(5)
Urinary tract infection, n (%)			0.555
No	1345(99.6)	380(99.5)
Yes	6(0.4)	2(0.5)
Septicemia, n (%)			<0.001
No	1233(91.3)	280(73.3)
Yes	118(8.7)	102(26.7)
MDR, n (%)			<0.001
No	1240(91.8)	285(74.6)
Yes	111(8.2)	97(25.4)
Severity of illness
NE needed, n (%)			<0.001
No	956 (70.8)	111 (29.1)
Yes	395 (29.2)	271 (70.9)
MV, n (%)			<0.001
No	591(43.7)	75(19.6)
Yes	760(56.3)	307(80.4)
Lactic acidosis, n (%)			<0.001
No	1006(74.5)	244(63.9)
Yes	345(25.5)	138(36.1)
Thrombocytopenia, n (%)			<0.001
No	978(72.4)	143(37.4)
Yes	373(27.6)	239(62.6)
Hyperbilirubinemia, n (%)			<0.001
No	1161(85.9)	251(65.7)
Yes	190(14.1)	131(34.3)
HF needed, n (%)			<0.001
No	1164(86.2)	224(58.6)
Yes	187(13.8)	158(41.4)
SOFA, M±SD	10.46±4.38	11.07±4.52	<0.001
Laboratory indicators
WBC, M±SD (×10^9/L)	14.54 ± 7.90	13.70 ± 9.08	0.404
N, M±SD (%)	84.45 ± 12.77	85.85 ± 10.13	0.349
L, M±SD (%)	10.10 ± 11.13	8.90 ± 8.57	0.353
CRP, M±SD (mg/L)	106.65 ± 96.56	121.63 ± 117.94	0.718
PCT, M±SD (pg/mL)	16.52 ± 33.68	17.50 ± 34.79	0.814
BCR, M±SD	19.96 ± 12.79	23.61 ± 18.15	0.040
TBIL, M±SD (μmol/L)	23.13 ± 29.65	17.86 ± 11.62	0.565
PLT, M±SD (×10^9/L)	193.33 ± 116.68	168.55 ± 106.99	0.078
ATIII, M±SD (%)	64.54 ± 18.92	54.93 ± 21.78	<0.001
APTT, M±SD (s)	44.95 ± 12.28	52.12 ± 24.79	<0.001
HCT, M±SD (%)	30.03 ± 5.96	28.03 ± 4.81	0.006

Abbreviations: M, mean; SD, standard deviation; MDR, multiple drug-resistant bacteria infection; NE, norepinephrine; MV, mechanical ventilation; HF, hemofiltration; SOFA, sequence organs failure assessment score; WBC, white blood counts; N, percentage of neutrophil; L, percentage of lymphocytes; CRP, C reactive protein; PCT, procalcitonin; BCR, Blood urinary nitrogen to Serum creatinine ratio; TBIL, total bilirubin; PLT, platelet counts; AT III, anti-thrombin III; APTT, activated partial thromboplastin time; HCT, hematocrit.

Table 2 Comparisons of Clinical Outcomes Between Survival and Non-Survival Groups

	Survival Group	Non-Survival Group	p value
LOS of ICU, day median^†	4 (2, 11)	8 (3, 15)	<0.001
MV duration, hours, median^†	34 (14, 116)	99.5 (34, 268)	<0.001

Note: ^†Non-normal distribution data were described using the percentile method.

Abbreviations: LOS, length of stay; MV, mechanical ventilation.

Table 3 Predicting Indicators Screened from LASSO Regression

Lambda (Log)	0.023 (−3.776)	0.002 (−6.381)
(Intercept)	0.405	−1.590
MV = No	−0.239	−0.577
MV = Yes	4.513e-15	1.187
Lactic acidosis =No	0	0.103
Lactic acidosis = Yes	0	−1.903
Septicemia = No	−0.249	−0.381
Septicemia = Yes	0.008	0.021
MDR = No	−0.193	−0.021
MDR = Yes	4.377	7.193
Thrombocytopenia =No	−0.458	−0.512
Thrombocytopenia = Yes	0.003	0.015
HF requirement = No	−0.634	−0.730
HF requirement = Yes	1.070	1.478
NE requirement = No	−0.972	−0.971
NE requirement = Yes	3.988	9.336
RDW-CV	0.012	0.127
HCT	−0.005	−0.007
CRP	0.0002	0.002
AT III	−0.0002	−0.003
APTT	0	0.007

Abbreviations: MV, mechanical ventilation; MDR, multiple drug-resistance infection; HF, hemofiltration; NE, norepinephrine; RDW-CV, red-cell deviation width of coefficient of variation; HCT, hematocrit; CRP, C reaction protein; AT III, anti-thrombin III; APTT, activated partial thromboplastin time.

Figure 1 The LASSO regression analysis identified variables correlated with ICU death.

Notes: (A) Number of non-zero coefficients in the model. (B) Number of variables corresponding to different λ values.

Figure 2 The predictive values of the model of machine learning algorithm.

Notes: (A) Missing data. By calculation, about 7% of data in the cohort were missing. (B) the AUCs of training and validation sets. The AUC of training data to predict ICU mortality was 0.801, while that of validation group was 0.791. (C) the calibration of training set. (D) the calibration of validation set. (E) the DCA curve of training set. (F) the DCA curve of validation set. Within this probability range of 0.25 to 0.90, the predictive performance of the model surpassed that of individual predictors within the cohort. The DCA curve of the validation data resembled that of the training data, further confirming the model’s favorable predictive effect.