Online Supplements

Online Supplements

Late-onset moderate-to-severe ARDS is associated withshorter survival and higher mortality: a two-stage association study

Ruyang Zhang1,2,3, Zhaoxi Wang1, Paula Tejera1,Angela J. Frank4, Yongyue Wei2,3, Li Su1, Zhaozhong Zhu1, Yichen Guo1, Feng Chen2,3, Ednan K. Bajwa4, B. Taylor Thompson4, David C. Christiani1,3,4

Affiliations:

1Department of Environmental Health, Harvard School of Public Health, Boston, MA, USA

2Department of Biostatistics, Ministry of Education Key Laboratory for Modern Toxicology, School of Public Health, Nanjing Medical University, Nanjing, Jiangsu, China

3Joint Laboratory of Health and Environmental Risk Assessment (HERA), Nanjing Medical University School of Public Health / Harvard School of Public Health

4Pulmonary and Critical Care Division, Department of Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA

Statistical methods to determine cut-point of early- and late-onsetmoderate-to-severe ARDS

We estimated hazard ratio (HR) of onset time (day 0 to day6) on overall survivaltime after ARDS onset toidentify cut-point of early- and late-onsetmoderate-to-severe ARDS. The HR was derived from a Cox proportional hazards model adjusted for covariates (Equation 1). Covariates were 13 variables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation and highest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level) and ICU 28-day ventilation-free days. These 13 variables which were not components of APACHE III score, meanwhile were significantly distributed between early- and late-onset ARDS patients (Table 3, Table S4, Table S5 and Table S6), were adjusted in all statistical models in our study.

Equation 1

Firstly, the onset time was included in the model as a dummy variable[1]. We set day0 as a reference group, and compared day 1 to day6 with day0. Based on the values of adjusted HR, onset times were categorizedinto two groups: early- and late-onset. ARDS patientswere assigned into these twogroups, according to their onset time. Furthermore, we used penalized smoothing splines in a multivariateCox proportional hazards model to confirm the effect pattern of onset time [2]. Analysis was performed using the “pspline” package in R software (version 3.3.1).

Statistical method to estimate risk effect of late-onset moderate-to-severe ARDS

The HR of late-onset ARDSwas evaluated using a multivariate Cox proportional hazards model (Equation 2).

Equation 2

The OR of late-onset ARDS was evaluated using a multivariate logistic regression model (Equation 3).

Equation 3

Prediction accuracy of multivariate Cox proportional hazards model

We predicted survival for ARDS patients fromthe 1stday to the 28thday of ICU admission, as well as60thday of ICUadmission, using nearest neighbor estimation (NNE) method for censored survival data [3].Time-dependent receiver operating characteristic curve(ROC) and time-dependent area underROC (AUC) were derived from the “survivalROC” package of R software (version 3.3.1). ARDS onset timeand 13 covariates were included in Cox proportional hazards mode.

General statistical methods

Continuous variables with normal distribution were described using mean ± standard deviation (SD) and were compared by student-t test. Otherwise, median and inter-quintile range (IQR) was used to describe non-normally distributed variables. The difference between groups was tested by Wilcoxon rank-sum test. Categorical variables were described using frequencies and proportions, and were tested by Chi-squared test.Kaplan-Meier survival curves for early- and late-onset ARDS patients were presented to exhibit difference of survival distribution between two groups of population.

Reference

1.Polissar L, Diehr P, (1982) Regression analysis in health services research: the use of dummy variables. Medical care 20: 959-966

2.Eilers PHC, Marx BD, (1996) Flexible smoothing with B-splines and penalties. Statistical Science 11: 89-121

3.Heagerty PJ, Lumley T, Pepe MS, (2000) Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics 56: 337-344

Table S1. Distribution of time forARDS patients receivinginitial arterial blood gas (ABG) and chest X-ray (CXR) examinationsat ICU admission.

Examination / Onset
day / N / Time of initial ARDS diagnosis through ABG and CXR examinations
Before
24 hours / Within
24 hours / Within
48 hours / Within
72 hours / Before or within
24 hours
ABG / 0 / 381 / 0 / 381 / 0 / 0 / 100.00%
1 / 241 / 0 / 241 / 0 / 0 / 100.00%
2 / 120 / 0 / 114 / 6 / 0 / 95.00%
3 / 76 / 0 / 72 / 3 / 1 / 94.74%
4 / 53 / 0 / 51 / 2 / 0 / 96.23%
5 / 30 / 0 / 30 / 0 / 0 / 100.00%
6 / 23 / 0 / 22 / 0 / 1 / 95.65%
Total / 924 / 0 / 911 / 11 / 2 / 98.59%
CXR / 0 / 381 / 4 / 377 / 0 / 0 / 100.00%
1 / 241 / 2 / 220 / 19 / 0 / 92.12%
2 / 120 / 1 / 112 / 7 / 0 / 94.17%
3 / 76 / 0 / 71 / 4 / 1 / 93.42%
4 / 53 / 0 / 48 / 5 / 0 / 90.57%
5 / 30 / 0 / 29 / 1 / 0 / 96.67%
6 / 23 / 1 / 22 / 0 / 0 / 100.00%
Total / 924 / 8 / 879 / 36 / 1 / 96.00%
ABG+CXR / 0 / 381 / 0 / 381 / 0 / 0 / 100.00%
1 / 241 / 0 / 222 / 19 / 0 / 92.12%
2 / 120 / 0 / 108 / 12 / 0 / 90.00%
3 / 76 / 0 / 67 / 7 / 2 / 88.16%
4 / 53 / 0 / 47 / 6 / 0 / 88.68%
5 / 30 / 0 / 29 / 1 / 0 / 96.67%
6 / 23 / 0 / 22 / 0 / 1 / 95.65%
Total / 924 / 0 / 876 / 45 / 3 / 94.81%

ARDS onset time within 24 hours of ICU admission was defined as day 0, and 24 hours to 48 hours was day 1, etc.; ARDS patients who did not receiveinitial ABG and CXR examinations before or within 24 hours after ICU admission were removed from this study. The final sample size in association analysis was 876.

Table S2. Description of populations in the derivation dataset, the validation dataset, and the combined dataset for detailed chronic diseases, arterial blood gas (ABG), vital signs,hematology and chemistry at time of admission.

Variable / Derivation dataset
(N = 520) / Validation dataset
(N = 356) / Combined dataset
(N = 876) / P
Chronic disease / 194(38.11%) / 130(36.62%) / 324(37.50%) / 0.6553
Chronic dialysis / 36( 7.07%) / 24( 6.76%) / 60( 6.94%) / 0.8591
Non-Hodgkin’s lymphoma / 1( 0.20%) / 3( 0.85%) / 4( 0.46%) / 0.1670
Solid tumor with metastasis / 14( 2.75%) / 5( 1.41%) / 19( 2.20%) / 0.1857
Immune suppression / 36( 7.07%) / 27( 7.61%) / 63( 7.29%) / 0.7669
Hepatic failure with coma / 25( 4.91%) / 11( 3.10%) / 36( 4.17%) / 0.1895
Cirrhosis / 40( 7.86%) / 22( 6.20%) / 62( 7.18%) / 0.3519
Diabetes mellitus / 109(21.41%) / 78(21.97%) / 187(21.64%) / 0.8448
1st ABG test within 24 hours
PaO2 (mmHg) / 91.00(75.00~126.00) / 91.00(76.00~121.00) / 91.00(75.00~123.00) / 0.7026
FiO2 / 0.80(0.60~1.00) / 1.00(0.60~1.00) / 0.90(0.60~1.00) / 0.1605
PaCO2(mmHg) / 43.00(37.00~50.00) / 42.00(37.00~51.00) / 42.00(37.00~51.00) / 0.8609
PH / 7.33(7.26~7.40) / 7.34(7.27~7.40) / 7.33(7.26~7.40) / 0.2417
PaO2/FiO2 / 136.67(91.00~189.17) / 134.00(92.34~185.00) / 135.50(91.00~187.50) / 0.6049
Intubated / 388(74.62%) / 255(71.63%) / 643(73.40%) / 0.3259
Vital signs
Temperature (F) / Lowest / 97.48±1.93 / 97.52±1.84 / 97.49±1.89 / 0.7446
Highest / 100.89±1.89 / 100.76±1.80 / 100.84±1.85 / 0.3074
SBP(mmHg) / Lowest / 83.38±16.14 / 83.37±17.30 / 83.38±16.61 / 0.9939
Highest / 146.47±24.38 / 143.91±25.26 / 145.43±24.76 / 0.1342
MAP (mmHg) / Lowest / 57.65±10.07 / 58.81±9.68 / 58.12±9.93 / 0.0954
Highest / 99.63±17.11 / 99.06±21.76 / 99.40±19.12 / 0.6671
Heart Rate / Lowest / 79.32±19.20 / 78.37±18.67 / 78.94±18.98 / 0.4672
Highest / 118.17±22.60 / 118.02±24.40 / 118.11±23.33 / 0.9261
Respiratory Rate / Lowest / 16.68±5.90 / 16.89±5.85 / 16.76±5.88 / 0.6010
Highest / 31.00±9.15 / 30.46±8.29 / 30.78±8.81 / 0.3592
Urine / 24 hours (mL) / 1645.00(888.50~2547.50) / 1585.00(900.00~2440.00) / 1625.00(898.00~2525.00) / 0.6893
Transfused inpast 7 days / 268(52.65%) / 195(55.56%) / 463(53.84%) / 0.4012
Hematology
Hematocrit (%) / Lowest / 30.45±6.31 / 29.93±6.15 / 30.24±6.25 / 0.2345
Highest / 34.74±6.36 / 34.05±6.21 / 34.46±6.30 / 0.1155
White blood cell / Lowest / 11.20(7.10~16.50) / 11.15(7.30~16.50) / 11.20(7.20~16.50) / 0.8388
Highest / 15.40(10.90~22.50) / 16.00(11.30~21.90) / 15.70(11.10~22.30) / 0.5326
Platelet (109/L) / 166.00(92.00~259.00) / 183.00(106.00~257.50) / 175.00(99.00~258.00) / 0.2433
Chemistry
Sodium(mEq/L) / Lowest / 136.26±5.79 / 136.00±5.53 / 136.16±5.69 / 0.4961
Highest / 139.85±6.65 / 139.65±5.72 / 139.77±6.29 / 0.6378
Potassium(mEq/L) / Lowest / 3.76±0.58 / 3.69±0.66 / 3.73±0.61 / 0.1469
Highest / 4.62±0.97 / 4.52±0.78 / 4.58±0.90 / 0.1010
BUN(mg/dl) / 24.00(16.00~43.00) / 25.00(16.00~40.00) / 24.50(16.00~41.00) / 0.7446
Creatinine(mg/dL) / Lowest / 1.10(0.70~1.80) / 1.07(0.70~1.60) / 1.10(0.70~1.73) / 0.5873
Highest / 1.30(0.90~2.30) / 1.30(0.90~2.10) / 1.30(0.90~2.20) / 0.5036
Glucose(mg/L) / Lowest / 113.00(96.00~142.00) / 118.00(96.00~144.00) / 116.00(96.00~143.00) / 0.2768
Highest / 174.00(142.00~233.00) / 177.00(139.00~237.00) / 175.00(141.00~233.50) / 0.9161
Albumin(g/dL) / Lowest / 2.36±0.67 / 2.44±0.72 / 2.39±0.70 / 0.1325
Highest / 2.49±0.72 / 2.57±0.71 / 2.53±0.72 / 0.1695
Bilirubin (mg/dL) / 0.90(0.50~2.00) / 0.85(0.50~1.90) / 0.90(0.50~2.00) / 0.2298
Bicarbonate (mEq/L) / 21.20±5.74 / 21.56±5.86 / 21.35±5.79 / 0.3650

SBP: systolic blood pressure; MAP: mean arterial pressure; BUN: blood urea nitrogen; Normal and non-normal distributed continuous variables weredescribed using mean ± standard and median (25thquantile~75thquantile), respectively. ARDS patients between two phases are comparable if P value >0.05.

Table S3. The adjusted hazard ratio (HR) of ARDS onset time using dummy variable in Cox proportional hazardsmodelin the derivation dataset, the validation dataset, and the combined dataset.

Dataset / Onset day / N / % / 28-day survival time / 60-day survival time
HR / 95%CI / P / HR / 95%CI / P
Derivation / 0 / 242 / 46.54 / Ref / Ref
1 / 135 / 25.96 / 1.11 / 0.73 / 1.69 / 0.6153 / 1.07 / 0.73 / 1.57 / 0.7391
2 / 53 / 10.19 / 2.17 / 1.16 / 4.07 / 0.0159 / 1.73 / 0.97 / 3.11 / 0.0656
3 / 37 / 7.12 / 2.86 / 1.52 / 5.38 / 0.0011 / 1.82 / 0.99 / 3.36 / 0.0542
4 / 27 / 5.19 / 1.89 / 0.83 / 4.31 / 0.1320 / 1.43 / 0.67 / 3.04 / 0.3563
5 / 13 / 2.50 / 2.84 / 1.06 / 7.57 / 0.0371 / 2.36 / 0.90 / 6.17 / 0.0807
6 / 13 / 2.50 / 1.85 / 0.42 / 8.13 / 0.4163 / 1.94 / 0.56 / 6.67 / 0.2922
Validation / 0 / 139 / 39.04 / Ref / Ref
1 / 87 / 24.44 / 0.98 / 0.55 / 1.77 / 0.9564 / 0.90 / 0.52 / 1.56 / 0.7125
2 / 55 / 15.45 / 1.99 / 1.04 / 3.80 / 0.0381 / 1.59 / 0.90 / 2.83 / 0.1120
3 / 30 / 8.43 / 2.61 / 1.22 / 5.55 / 0.0130 / 2.15 / 1.07 / 4.31 / 0.0317
4 / 20 / 5.62 / 2.55 / 0.96 / 6.78 / 0.0604 / 1.75 / 0.71 / 4.28 / 0.2236
5 / 16 / 4.49 / 1.54 / 0.50 / 4.82 / 0.4536 / 1.57 / 0.59 / 4.18 / 0.3691
6 / 9 / 2.53 / 1.88 / 0.40 / 8.88 / 0.4239 / 0.90 / 0.20 / 4.07 / 0.8888
Combined / 0 / 381 / 43.49 / Ref / Ref
1 / 222 / 25.34 / 1.09 / 0.78 / 1.53 / 0.6110 / 1.04 / 0.77 / 1.42 / 0.7869
2 / 108 / 12.33 / 1.88 / 1.22 / 2.92 / 0.0045 / 1.54 / 1.04 / 2.29 / 0.0312
3 / 67 / 7.65 / 2.70 / 1.67 / 4.36 / 0.0001 / 1.90 / 1.20 / 2.99 / 0.0057
4 / 47 / 5.37 / 1.97 / 1.06 / 3.66 / 0.0324 / 1.49 / 0.85 / 2.64 / 0.1675
5 / 29 / 3.31 / 1.96 / 0.94 / 4.05 / 0.0716 / 1.86 / 0.96 / 3.64 / 0.0679
6 / 22 / 2.51 / 1.46 / 0.52 / 4.11 / 0.4748 / 1.13 / 0.44 / 2.85 / 0.8025

Ref: reference; HR: hazard ratio; 95%CI: 95% confidence interval; HR, 95%CI and P values were adjusted forvariables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation andhighest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level) and ICU 28-day ventilation-free days in the Cox proportional hazardsmodel; ARDS onset time within 24 hours of ICU admission was defined as day 0, and 24 hours to 48 hours was day 1, etc.

Table S4. Comparison of demographics, past history, detailed chronic diseases and components of lung injury scoreand ICU enrollmentfor early- and late-onset ARDS patients at time of ICU admission, as well as ICU duration days and ventilator-free days.

Variable / Early-onset ARDS
(N = 603) / Late-onset ARDS
(N = 273) / P
Body mass index / 28.55±8.88 / 29.07±9.38 / 0.4316
Alcohol abuse within 12 months / 102(16.92%) / 45(16.48%) / 0.8741
Patient source
Operating room / 71(11.77%) / 44(16.12%) / 0.2159
Recovery room / 2( 0.33%) / 4( 1.47%)
Emergency room / 216(35.82%) / 92(33.70%)
Hospital floor / 163(27.03%) / 71(26.01%)
Other special care unit / 24( 3.98%) / 11( 4.03%)
Other hospital floor / 127(21.06%) / 51(18.68%)
ICU readmission / 59( 9.78%) / 31(11.36%) / 0.4782
Chronic disease / 227(38.22%) / 97(35.93%) / 0.5194
Chronic dialysis / 47( 7.91%) / 13( 4.81%) / 0.0969
Non-Hodgkin’s lymphoma / 3( 0.51%) / 1( 0.37%) / 0.7869
Solid tumor with metastasis / 14( 2.36%) / 5( 1.85%) / 0.6389
Immune suppression / 45( 7.58%) / 18( 6.67%) / 0.6338
Hepatic failure with coma / 20( 3.37%) / 16( 5.93%) / 0.0810
Cirrhosis / 41( 6.90%) / 21( 7.78%) / 0.6440
Diabetes mellitus / 128(21.55%) / 59(21.85%) / 0.9201
Components of lung injury score
Chest radiograph score / 2.91±1.12 / 1.83±1.40 / 0.0001
(0) 0 quadrant alveolar consolidation / 25( 4.15%) / 63(23.08%) / 0.0001
(1) 1 quadrant alveolar consolidation / 30( 4.98%) / 52(19.05%)
(2) 2 quadrant alveolar consolidation / 172(28.52%) / 78(28.57%)
(3) 3 quadrant alveolar consolidation / 126(20.90%) / 28(10.26%)
(4) 4 quadrant alveolar consolidation / 250(41.46%) / 52(19.05%)
Hypoxaemia score / 3.19±1.06 / 2.84±1.37 / <0.0001
(0) PaO2/FiO2 ≥300 / 36( 5.97%) / 33(12.09%) / 0.0032
(1) PaO2/FiO2 225~299 / 8( 1.33%) / 17( 6.23%)
(2) PaO2/FiO2 175~224 / 52( 8.62%) / 30(10.99%)
(3) PaO2/FiO2 100~174 / 215(35.66%) / 74(27.11%)
(4) PaO2/FiO2 <100 / 292(48.42%) / 119(43.59%)
Ventilated PEEP score / 1.51±1.40 / 0.75±1.16 / 0.0001
(0) ≤ 5 cm H2O / 206(34.16%) / 174(63.74%) / 0.0001
(1) 6~8 cm H2O / 109(18.08%) / 32(11.72%)
(2) 9~11 cm H2O / 140(23.22%) / 41(15.02%)
(3) 12~14 cm H2O / 68(11.28%) / 13( 4.76%)
(4)≥ 15 cm H2O / 80(13.27%) / 13( 4.76%)
Respiratory system compliance score / 1.79±1.50 / 1.15±1.45 / 0.0001
(0) ≥ 80 mL/cm H2O / 232(38.47%) / 161(58.97%) / 0.0001
(1) 60~79 mL/cm H2O / 7( 1.16%) / 5( 1.83%)
(2) 40~59 mL/cm H2O / 66(10.95%) / 25( 9.16%)
(3) 20~39 mL/cm H2O / 251(41.63%) / 70(25.64%)
(4)≤ 19 mL/cm H2O / 47( 7.79%) / 12( 4.40%)
ICU28-day duration days / 12.00(7.00～21.00) / 15.00(10.00～21.00) / 0.0001
ICU 28-day ventilator-free days / 1.00(0.00～3.00) / 4.00(2.00～6.00) / 0.0001

ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage.

Table S5. Comparison of vital signs, hematology, chemistry and arterial blood gas (ABG) at time of admission for early- and late-onset patients.

Variable / Early-onset ARDS
(N = 603) / Late-onset ARDS
(N = 273) / P
Vital signs
Temperature (F) / Lowest / 97.46±1.83 / 97.56±2.04 / 0.5002
Highest / 100.87±1.77 / 100.77±2.03 / 0.4746
Systolic blood pressure (mmHg) / Lowest / 81.92±15.88 / 86.61±17.75 / 0.0002
Highest / 144.94±24.21 / 146.51±25.93 / 0.3895
Mean arterial pressure (mmHg) / Lowest / 57.50±9.49 / 59.51±10.73 / 0.0090
Highest / 99.51±19.83 / 99.17±17.46 / 0.8035
Heart Rate / Lowest / 78.50±19.40 / 79.91±18.03 / 0.3097
Highest / 118.69±23.79 / 116.80±22.28 / 0.2672
Respiratory Rate / Lowest / 16.91±5.86 / 16.44±5.92 / 0.2820
Highest / 31.35±8.82 / 29.52±8.67 / 0.0047
Urine / 24 hours (mL) / 1590.00(859.00～2455.00) / 1700.00(952.00～2680.00) / 0.1360
Hematology
Hematocrit (%) / Lowest / 30.60±6.25 / 29.44±6.17 / 0.0117
Highest / 34.91±6.27 / 33.46±6.28 / 0.0016
White blood cell / Lowest / 11.50(7.20～16.80) / 11.00(7.20～15.90) / 0.4808
Highest / 15.85(11.10～22.75) / 15.50(10.90～21.10) / 0.4147
Platelet(109/L) / 177.00(105.00～267.00) / 169.00(88.00～247.00) / 0.0875
Chemistry
Sodium (mEq/L) / Lowest / 136.37±5.31 / 135.68±6.43 / 0.1246
Highest / 139.98±6.32 / 139.30±6.19 / 0.1418
Potassium (mEq/L) / Lowest / 3.73±0.63 / 3.75±0.59 / 0.6576
Highest / 4.62±0.90 / 4.48±0.87 / 0.0380
Blood urea nitrogen (mg/dl) / 24.00(16.00～41.00) / 25.00(16.00～42.00) / 0.8932
Creatinine (mg/dL) / Lowest / 1.03(0.70～1.70) / 1.10(0.70～1.80) / 0.6064
Highest / 1.30(0.90～2.20) / 1.30(0.90～2.20) / 0.9278
Glucose (mg/L) / Lowest / 114.00(96.00～142.00) / 117.50(98.00～144.00) / 0.3695
Highest / 177.00(143.00～239.00) / 170.00(136.00～225.00) / 0.1216
Albumin (g/dL) / Lowest / 2.38±0.68 / 2.43±0.74 / 0.3852
Highest / 2.54±0.68 / 2.50±0.79 / 0.6130
Bilirubin (mg/dL) / 0.90(0.50～1.70) / 1.00(0.60～2.90) / 0.0465
Bicarbonate (mEq/L) / 21.31±5.96 / 21.43±5.39 / 0.7748
Initial ABG test within 24 hours
PaO2(mmHg) / 91.00(75.00～119.00) / 94.00(75.00～134.00) / 0.1311
FiO2 / 1.00(0.60～1.00) / 0.80(0.50～1.00) / 0.0015
PaCO2(mmHg) / 44.00(37.00～52.00) / 41.00(36.00～47.00) / 0.0005
PH / 7.33(7.25～7.40) / 7.34(7.29～7.41) / 0.0055
PaO2/FiO2 / 126.15(89.00～176.00) / 156.00(100.27～213.00) / 0.0001
PaO2/FiO2 decline from admission to onset / 13.00(0.00～60.00) / 29.00(0.50～94.00) / 0.0445

Normal and non-normal distributed continuous variables weredescribed using mean ± standard and median (25thquantile~75thquantile), respectively;ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage.

Table S6. Comparison of detailed Marshall'smultiple organ dysfunction score for early- and late-onset patients at time of ARDS onset.

Variable / Early-onset ARDS
(N = 603) / Late-onset ARDS
(N = 273) / P
Pulmonary score / 3.11±0.66 / 2.91±0.65 / 0.0001
PaO2/FiO2 / (2) 151~255 / 102(17.06%) / 71(26.30%) / 0.0001
(3) 76~150 / 331(55.35%) / 153(56.67%)
(4) <75 / 165(27.59%) / 46(17.04%)
Hematologicscore / 0.49±0.92 / 0.76±1.11 / 0.0006
Platelet (109/L) / (0) >120 / 430(72.64%) / 167(61.85%) / 0.0005
(1) 81~120 / 75(12.67%) / 39(14.44%)
(2) 51~80 / 54( 9.12%) / 31(11.48%)
(3) 21~50 / 26( 4.39%) / 29(10.74%)
(4) ≤20 / 7( 1.18%) / 4( 1.48%)
Renal score / 0.86±1.00 / 0.73±0.94 / 0.0724
Creatinine (mg/dL) / (0) ≤1.1 / 268(45.12%) / 143(52.57%) / 0.0537
(1) 1.1~2.3 / 201(33.84%) / 79(29.04%)
(2) 2.3~4.0 / 84(14.14%) / 36(13.24%)
(3) 4.0~5.7 / 24( 4.04%) / 9( 3.31%)
(4) >5.7 / 17( 2.86%) / 5( 1.84%)
Hepaticscore / 0.55±0.99 / 0.73±1.14 / 0.0426
Total Bilirubin (mg/dL) / (0)≤1.2 / 319(68.90%) / 140(61.67%) / 0.0402
(1) 1.2~3.5 / 79(17.06%) / 43(18.94%)
(2)3.5~7.0 / 33( 7.13%) / 20( 8.81%)
(3) 7.0~14.0 / 18( 3.89%) / 13( 5.73%)
(4) 14.0 / 14( 3.02%) / 11( 4.85%)
Cardiovascularscore / 1.21±1.10 / 1.05±0.98 / 0.0394
Vasopressors / (0) / 122(21.59%) / 66(26.61%) / 0.0634
(1) / 335(59.29%) / 143(57.66%)
(2) / 14( 2.48%) / 9( 3.63%)
(3) / 56( 9.91%) / 21( 8.47%)
(4) / 38( 6.73%) / 9( 3.63%)

ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage.

1

Table S7. AssociationbetweenARDS onset time and survival time in fiverandomly generated derivation and validation datasets.

Index / Dataset / ARDS onset / 28-day survival time / 60-day survival time
HR / 95% CI / P / HR / 95% CI / P
1 / Derivation / Late vs Early / 1.89 / 1.27 / 2.79 / 0.0015 / 1.52 / 1.06 / 2.17 / 0.0211
Validation / Late vs Early / 2.42 / 1.44 / 4.06 / 0.0008 / 1.94 / 1.21 / 3.13 / 0.0064
2 / Derivation / Late vs Early / 2.20 / 1.46 / 3.30 / 0.0002 / 1.62 / 1.13 / 2.33 / 0.0095
Validation / Late vs Early / 2.32 / 1.40 / 3.84 / 0.0011 / 1.93 / 1.21 / 3.08 / 0.0061
3 / Derivation / Late vs Early / 2.08 / 1.40 / 3.09 / 0.0003 / 1.46 / 1.02 / 2.10 / 0.0383
Validation / Late vs Early / 2.09 / 1.26 / 3.48 / 0.0042 / 1.96 / 1.25 / 3.09 / 0.0035
4 / Derivation / Late vs Early / 2.25 / 1.49 / 3.38 / 0.0001 / 1.71 / 1.18 / 2.49 / 0.0047
Validation / Late vs Early / 2.05 / 1.24 / 3.39 / 0.0050 / 1.69 / 1.07 / 2.67 / 0.0242
5 / Derivation / Late vs Early / 1.84 / 1.25 / 2.70 / 0.0020 / 1.57 / 1.11 / 2.23 / 0.0109
Validation / Late vs Early / 2.57 / 1.50 / 4.39 / 0.0006 / 1.73 / 1.06 / 2.83 / 0.0280

HR: hazard ratio; 95% CI: 95% confidence interval; HR, 95%CI and P values were adjusted for variables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation and highest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level) and ICU 28-day ventilation-free days in the Cox proportional hazardsmodel; ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage.

Table S8. Subgroup analysis of association between ARDS onset time and survival time by different ARDS risk factors.

ARDS risk factors / N / 28-day survival time / 60-day survival time
HR / 95% CI / P / HR / 95% CI / P
Bacteremia / 140 / 1.85 / 0.90 / 3.81 / 0.0956 / 1.31 / 0.66 / 2.60 / 0.4331
Sepsis / 756 / 1.98 / 1.47 / 2.67 / <0.0001 / 1.54 / 1.17 / 2.03 / 0.0020
Septic shock / 595 / 1.66 / 1.19 / 2.32 / 0.0028 / 1.38 / 1.02 / 1.88 / 0.0391
Pneumonia / 651 / 1.80 / 1.28 / 2.52 / 0.0007 / 1.54 / 1.13 / 2.10 / 0.0060
Multiple fractures / 34 / 1.81 / 0.11 / 29.73 / 0.6792 / 6.43 / 0.40 / 102.63 / 0.1880
Pulmonary contusion / 43 / 1.37 / 0.10 / 19.12 / 0.8173 / 2.36 / 0.28 / 19.96 / 0.4313
Aspiration / 87 / 2.86 / 0.86 / 9.46 / 0.0860 / 1.85 / 0.61 / 5.62 / 0.2808
Multiple transfusion / 71 / 2.96 / 1.00 / 8.78 / 0.0499 / 2.13 / 0.85 / 5.32 / 0.1073

HR: hazard ratio; 95% CI: 95% confidence interval; HR, 95%CI and P values were adjusted for variables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation and highest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level)and ICU 28-day ventilation-free days in the Cox proportional hazardsmodel; ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage.

Table S9. The 28-day and 60-day mortality difference between early- and late-onset ARDS patients in the combined dataset.

Survival time / Onset stage / N / Alive / Dead / Crude mortality rate / OR / 95%CI / P
28-day / Early / 603 / 433 / 170 / 28.19% / Ref
Late / 273 / 186 / 87 / 31.87% / 1.46 / 1.04 / 2.06 / 0.0305
60-day / Early / 603 / 406 / 197 / 32.67% / Ref
Late / 273 / 174 / 99 / 36.26% / 1.44 / 1.03 / 2.02 / 0.0313

ARDS onset within 48 hours after admission was defined as early-onset, and otherwise as late-onset stage; The OR, 95%CI, z and P values were derived from logistic regression model adjusted for variables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation and highest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level) and ICU 28-day ventilation-free days.

Figure S1. Distribution of onset time for ARDS patients in the derivation dataset, the validation dataset, and the combined dataset.ARDS onset time within 24 hours of ICU admission was defined as day 0, and 24 hours to 48 hours was day 1, etc.

Figure S2. The adjusted hazard ratio of ARDS onset time using dummy variable in Cox proportional hazardsmodelof 28-day and 60-day survival time in the derivation dataset, the validation dataset, and the combined dataset.ARDS onset time within 24 hours of ICU admission was defined as day 0, and 24 hours to 48 hours was day 1, etc. ARDS onset within 48 hours after admission was defined as early-onset, andotherwise as late-onset stage.

1

FigureS3.Adjusted hazards ratio of ARDS onset time using penalized smoothing splines in Cox proportional hazardsmodel of 28-day and 60-day survivaltime in thederivation dataset, the validation dataset and the combined dataset.ARDS onset time within 24 hours of ICU admission was defined as day 0, and 24 hours to 48 hours was day 1, etc.ARDS onset within 48 hours after admission was defined as early-onset, andotherwise as late-onset stage.

1

FigureS4.Time-dependent receiver operating characteristic curve (ROC) for multivariate Cox proportionalhazardsmodel of 28-day and 60-day survival time in the combined dataset. Baseline includes variables collected within 24 hours of ICU admission (APACHE III score, septic shock, pneumonia, multiple fractures, transfusion within 7 days prior admission, vasopressors prior admission, intubation, ventilation and highest potassium level), variables collected at time of ARDS onset (PaO2/FiO2, platelet count, and creatinine level) and ICU 28-day ventilation-free days.

1