Development of a model by LASSO to predict hospital length of stay (LOS) in patients with the SARS-Cov-2 omicron variant

Figures & data

Table 1. General characteristics of the 384 patients on admission.

Characteristics	Aggregate (n = 384)	LOS<15 days (n = 85)	LOS 15–25 days (n = 212)	LOS>25 days (n = 87)	P for trend
Age (years)	19 ± 6	13 ± 1	19 ± 3	29 ± 3	0.3914
Female, n (%)	222 (57.8)	50 (58.8)	125 (59.0)	47 (54.0)	0.5216
Familial cluster	166 (43.2)	24 (28.2)	100 (47.2)	42 (48.3)	0.0084
Number of vaccination doses
0	2 (0.5)	0 (0.0)	1 (0.5)	1 (1.1)	0.2951
1	12 (3.1)	1 (1.2)	8 (3.8)	3 (3.4)	0.3964
2	349 (90.9)	78 (91.8)	193 (91.0)	78 (89.7)	0.6306
3	21 (5.5)	6 (7.1)	10 (4.7)	5 (5.7)	0.7102
Coexisting conditions
Diabetes	10 (2.6)	2 (2.4)	7 (3.3)	1 (1.1)	0.6149
Hypertension	23 (6.0)	5 (5.9)	14 (6.6)	4 (4.6)	0.7193
Cardiovascular disease	6 (1.6)	1 (1.2)	2 (0.9)	3 (3.4)	0.2267
Chronic liver disease	6 (1.6)	1 (1.2)	3 (1.4)	2 (2.3)	0.5519
Fever	96 (25.0)	17 (20.0)	58 (27.4)	21 (24.1)	0.5382
Cough	240 (62.5)	56 (65.9)	136 (64.2)	48 (55.2)	0.1457
Runny nose	39 (10.2)	8 (9.4)	20 (9.4)	11 (12.6)	0.4808
Nasal congestion	96 (25.0)	27 (31.8)	49 (23.1)	20 (23.0)	0.1872
Fatigue	15 (3.9)	2 (2.4)	9 (4.2)	4 (4.6)	0.45
Expectoration	112 (29.2)	23 (27.1)	64 (30.2)	25 (28.7)	0.8124
Headache	7 (1.8)	1 (1.2)	6 (2.8)	0 (0.0)	0.5552
Diarrhoea	6 (1.6)	1 (1.2)	3 (1.4)	2 (2.3)	0.5519
Systolic pressure, mmHg	119 (110, 125)	120 (111, 125)	117 (111, 125)	117 (110, 125)	0.2724
Diastolic pressure, mmHg	72.8 (66, 80)	75 (68, 80)	73 (65.5, 80)	70 (65, 78)	0.1821
Heart rate	80 (75, 82)	79 (75, 81)	80 (74, 82)	80 (75, 82)	0.3772
Respiratory rate	19 (18, 20)	19 (18, 20)	19 (18, 20)	19 (18, 20)	0.392
Oxygen saturation	99 (98, 99)	99 (98, 99)	99 (98, 99)	99 (97, 99)	0.5432
Time from illness onset to first hospital admission (days)	2 (1, 3)	2 (1, 3)	2 (1, 3)	2 (1, 3)	0.9355

Data are median (IQR), mean ± SD or n (%). IQR: interquartile range.

Figure 1. Flow diagram of the study. A total of 384 patients infected with Omicron variant from January 8 to February 13 in 2022 were enrolled. The final predictive model was constructed by fitting a linear regression model using the predictors selected by the LASSO method. Due to small sample size, bootstrap validation was used to test model performance. A total of 2,000 bootstrap samples were taken and replaced with the same sample size as the original, resulting in the final model.

Table 2. Treatments and outcomes for 384 patients.

Characteristics	Aggregate (n = 384)	LOS<15 days (n = 85)	LOS 15–25 days (n = 212)	LOS>25 days (n = 87)	P for trend
TCM treatment	381 (99.2)	83 (97.6)	211 (99.5)	87 (100.0)	0.0079
Immunotherapy	143 (37.2)	3 (3.5)	80 (37.7)	60 (69.0)	<0.0001
Heparin	20 (5.2)	0 (0.0)	8 (3.8)	12 (13.8)	<0.0001
Interferon inhalation	22 (5.7)	0 (0.0)	13 (6.1)	9 (10.3)	0.0036
Antibiotics	11 (2.9)	0 (0.0)	8 (3.8)	3 (3.4)	0.1791
Digestive treatment	12 (3.1)	1 (1.2)	8 (3.8)	3 (3.4)	0.3964

TCM treatment: traditional Chinese medicine treatment.

Figure 2. Screening process and analysis of results. (a) Histogram of length of hospital stay. Because the patients’ LOS showed a skewed state, log transformation with base e was made to it. (b) Using all the sample and candidate predictors, we employ LASSO to select the primitive predictors. Optimal lambda is 0.395477. (c) Box plot of real LOS and its predicted value. (d) Scatterplot of real LOS and its predicted value.

Table 3. Candidate variables selected by the LASSO simulation.

Number	Variables	Coefficients
1	Familial cluster	1.2470368406
2	Runny nose	0.5292789600
3	SBP	−0.0002272907
4	PLT	0.0008747604
5	APTT	0.0175742136
6	ALT	0.0066736904
7	AST	0.0171238952
8	Immunotherapy	5.3377636058
9	Heparin	1.6286117378

ASSO: Least Absolute Shrinkage and Selection Operator; SBP: Systolic blood pressure; PLT: platelet count; APTT: activated partial thromboplastin time; ALT: alanine transaminase; AST: aspartate aminotransferase.

Table 4. Candidate variables again selected by a linear regression model.

Label	Estimate	StdErr	tValue	P
Immunotherapy	0.3155	0.0284	11.1265	0.0000
Familial cluster	0.1127	0.0274	4.1201	0.0000
Heparin	0.1542	0.0614	2.5108	0.0125
Runny nose	0.0941	0.0446	2.1082	0.0357
APTT	0.0035	0.0017	2.0302	0.0430
Systolic pressure	−0.0019	0.0010	−1.8372	0.0670
AST	0.0017	0.0011	1.5114	0.1315
PLT	0.0004	0.0003	1.4203	0.1563
ALT	0.0010	0.0007	1.4047	0.1609

APTT: activated partial thromboplastin time; AST: aspartate aminotransferase; PLT: platelet count; ALT: alanine transaminase.

Table 5. A prognostic model to predict the length of stay of Omicron patients.

Label	Estimate	StdErr	tValue	p-value	OR(95%CI)
Intercept	2.66263	0.05879	45.29	<0.0001
Immunotherapy	0.3077	0.0286	10.7475	0.0000	36.0305 (28.6072–43.8822)
Familial cluster	0.1158	0.0275	4.2076	0.0000	12.2754 (6.3802–18.4973)
Heparin	0.1496	0.0624	2.3981	0.0170	16.1422 (2.7716–31.2524)
Runny_nose	0.0989	0.0451	2.1920	0.0290	10.3953 (1.0521–20.6023)
APTT	0.0036	0.0017	2.0656	0.0396	0.3570 (0.0182–0.6969)

APTT: activated partial thromboplastin time.

Data availability statement

The data that support the findings of this study are available from the corresponding author [RZG], upon reasonable request.