KEY POINTS
Question: What is the best target time from the emergency department (ED) visit to the ICU admission (i.e., ED length of stay [LOS]) for patients with sepsis requiring intensive care?
Findings: Patients with sepsis in the higher ED-LOS quartiles (i.e., > 3.3 hr) had higher 28-day mortality compared with those in the lower ED-LOS quartile (i.e., < 3.3 hr).
Meanings: Patients with sepsis who require intensive care may benefit from a more immediate ICU admission (e.g., 3 hr) than 6 hours based on the suggestion by the Surviving Sepsis Campaign Guidelines 2021.
Sepsis is a leading cause of death and critical illness worldwide (1,2), and rapid assessment and appropriate intervention based on sepsis bundles are necessary to decrease the mortality of sepsis (3,4). Delayed admission of patients with suspected sepsis from the emergency department (ED) to the ICU is associated with lower compliance with the sepsis bundle, higher mortality, longer duration of ventilator use, and longer ICU and hospital stay (5). Therefore, the Surviving Sepsis Campaign Guidelines (SSCG) 2021 suggested that patients with sepsis who require intensive care should be admitted to the ICU within 6 hours of their ED visits (3,4).
The suggestion by the SSCG 2021 is grounded on a previous study that estimated a 1.5% mortality increase for every hour delay in ICU admission for critically-ill patients (6), and several studies have proposed specific time targets from the ED visit to the ICU admission (i.e., ED length of stay [ED-LOS]) of 2.4, 3, and 6 hours (7–9). However, as these findings are based on studies analyzing the data on heterogeneous critically-ill patients with and without sepsis, there are no studies on the optimal ED-LOS in patients with sepsis to directly support the recommendations by the SSCG 2021. Furthermore, there are conflicting findings as to whether ICU admissions within 6 hours of ED visits are associated with better prognostic outcomes among patients with sepsis (10,11). Thus, there is a dearth of evidence that quantifies within how many hours patients with sepsis should be admitted to the ICU.
To address these knowledge gaps, we aimed to investigate the association between ED-LOS and mortality and identify the optimal ED-LOS for patients with sepsis who require intensive care based on the sepsis bundles.
MATERIALS AND METHODS
Study Design and Setting
This is a retrospective cohort study using the data from the Medical Information Mart for Intensive Care Emergency Department (MIMIC-IV-ED) dataset version 1.0 (12) and the Medical Information Mart for Intensive Care (MIMIC-IV) dataset version 1.0 (13). MIMIC-IV is an extensive, publicly available database consisting of de-identified health-related data from over 60,000 patients admitted to the ICUs of the Beth Israel Deaconess Medical Center in the United States from 2008 to 2019 (13). MIMIC-IV-ED is a large publicly available dataset of over 40,000 patients who visited the ED of the Beth Israel Deaconess Medical Center from 2011 to 2019 (12). Approval by the research ethics committee was not needed for this study because MIMIC-IV and MIMIC-IV-ED databases are de-identified according to the HIPAA (HIPAA) Safe Harbor provision, and only credentialed authors who signed and conformed to the specified data use agreement accessed and analyzed the data (12,13). Because of this, the TXP Medical Ethical Review Board waived the requirement for the ethical approval statement and informed consent (TXPREC-008).
Study Participants
The inclusion and exclusion criteria are shown in Figure 1. We identified adult patients (aged ≥ 18 yr) who were admitted to the ICU directly within 9 hours of the ED visit and met the diagnostic criteria for sepsis based on the Sepsis-3 criteria (14) within 24 hours of ICU admission from MIMIC-IV and MIMIC-IV-ED databases. Our analysis was limited to patients initially triaged at the resuscitated or emergency level, which requires compliance with the sepsis bundle (3,4). We excluded patients who were transported from other hospitals, those who had a cardiac arrest or trauma at the ED arrival, those who died at the ED, and those who had missing data on all vital signs (i.e., systolic blood pressure [SBP], diastolic blood pressure [DBP], heart rate [HR], respiratory rate [RR], and body temperature [BT]) at the ED triage or all initial laboratory variables (i.e., sodium, potassium, lactate, blood urea nitrogen [BUN], creatinine, WBC, hemoglobin, and platelet count) on ICU admission.
;)
Figure 1.: Flowchart of patient selection. ED = emergency department.
Measurements
For each patient, we collected the following covariates: patient demographics (age, sex, and race/ethnicity [non-Hispanic White, non-Hispanic Black, Hispanic, and others]), initial vital signs (SBP, DBP, HR, RR, and BT) and the counts of Systemic Inflammatory Response Syndrome [SIRS] qualifying categories at ED arrival, initial laboratory data on ICU admission (sodium, potassium, lactate, BUN, creatinine, WBC, hemoglobin, and platelet count), the maximum Sequential Organ Failure Assessment [SOFA] score within 24 hours of ICU admission. The exposure was time from the ED visit to ICU admission (i.e., ED length of stay [ED-LOS]) and the primary outcome was 28-day mortality. We considered patients who were discharged or were transferred from the ICU within 28 days as those who survived for 28 days after ICU admission.
Statistical Analyses
First, we treated outliers as missing and imputed all missing values using the MissForest algorithm, a nonparametric machine learning-based imputation method (15). The valid ranges of variables defining outliers and the percentage of missing values are shown in Supplemental Table 1, https://links.lww.com/CCX/B191.
We used summary statistics to delineate the patient characteristics of all patients and each patient in the ED-LOS quartiles. Continuous variables were expressed as medians and interquartile ranges (IQRs). Categorical variables were summarized as counts (n) and percentages (%).
Second, we visualized the association between ED-LOS and 28-day mortality using a locally weighted scatterplot smoother (Lowess) curve given that the relationship between the exposure and outcome could be nonlinear. Lowess is a widely used statistical method for estimating local regression surfaces to remove noise from raw data and clarify graphical presentations (16).
Third, we constructed a multivariable logistic regression model to investigate the association between ED-LOS and 28-day mortality, adjusting for the following covariates: patient demographics (age, sex, and race/ethnicity), initial vital signs, the counts of SIRS qualifying categories at ED arrival, initial laboratory data at least 24 hours after ICU admission (sodium, potassium, lactate, BUN, creatinine, WBC, hemoglobin, and platelet count), the maximum SOFA score within 24 hours of ICU admission. Furthermore, we repeated the logistic regression analysis after categorizing all patients into ED-LOS quartiles to identify the optimal ED-LOS.
All analyses were conducted using Google BigQuery, Python version 3.8.8, and R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). A p value of less than 0.05 was considered statistically significant.
RESULTS
Patient Characteristics
We identified 1,849 patients eligible for analysis. Table 1 showed summary statistics of all patients and each patient in the ED-LOS quartiles. The median (IQR) age of overall patients was 68 (57–80) years, and 1,025 (55%) were male; 1,244 (67%) were non-Hispanic White, 286 (15%) were non-Hispanic Black, and 91 (5%) were Hispanic. The median (IQR) lactate level at least 24 hours after ICU admission and maximum SOFA score within 24 hours of ICU admission were 2.3 mmol/L (1.6–3.3) and 6 (4–8), respectively. The median ED-LOS was 4.6 (3.3–6.1) hours, and 258 (14%) died within 28 days of the ED visit. All patients were divided based on ED-LOS quartiles (i.e., < 3.3 hr, 3.3–4.5 hr, 4.6–6.1 hr, and > 6.1 hr), but we did not find any clinically meaningful differences in prognostic parameters of sepsis (e.g., the lactate level, the count of the SIRS criteria qualified, and the SOFA score) and outcome (i.e., 28-d mortality) across each quartile group.
TABLE 1. -
Characteristics of All Patients and Patients Divided Into Quartiles Based on the Time From Emergency Department Visit to ICU Admission
| Variables |
Overall (n = 1,849) |
< 3.3 hr
(n = 466) |
3.3–4.5 hrn = 459) |
4.6–6.1 hr(n = 462) |
> 6.1 hr
(n = 462) |
| Patient demographics |
|
|
|
|
|
| Age (yr), median (IQR) |
68 (57–80) |
68 (57–80) |
67 (57–81) |
68 (58–80) |
67 (57–78) |
| Male, n (%) |
1025 (55) |
265 (57) |
268 (58) |
251 (54) |
241 (52) |
| BMI (kg/m2), median (IQR) |
26.6 (23.4–30.9) |
26.3 (22.8–29.8) |
27.0 (23.6–31.1) |
26.6 (23.5–31.0) |
26.7 (23.6–32.1) |
| Race, n (%) |
|
|
|
|
|
| Non-Hispanic White |
1,244 (67) |
317 (68) |
313 (68) |
301 (65) |
313 (68) |
| Non-Hispanic Black |
286 (15) |
70 (15) |
66 (14) |
78 (17) |
72 (16) |
| Hispanic |
91 (5) |
15 (3) |
24 (5) |
27 (6) |
25 (5) |
| Triage vital signs, median (IQR) |
|
|
|
|
|
| Systolic blood pressure (mm Hg) |
113 (94–132) |
116 (97–134) |
112 (95–132) |
113 (93–131) |
110 (93–132) |
| Diastolic blood pressure (mm Hg) |
64 (52–76) |
67 (53–78) |
64 (53–75) |
63 (51–76) |
63 (52–76) |
| Heart rate (/min) |
99 (82–115) |
100 (84–116) |
97 (80–116) |
98 (80–114) |
99 (82–113) |
| Respiratory rate (/min) |
20 (18–24) |
20 (18–26) |
20 (18–24) |
19 (18–23) |
18 (16–22) |
| Body temperature (°C) |
36.9 (36.5–37.4) |
36.9 (36.6–37.4) |
36.9 (36.5–37.4) |
36.9 (36.4–37.5) |
36.9 (36.4–37.4) |
| First laboratory results, median (IQR) |
|
|
|
|
|
| Sodium (mEq/L) |
138 (134–141) |
138 (135–141) |
138 (134–141) |
138 (134–141) |
138 (134–141) |
| Potassium (mEq/L) |
4.1 (3.7–4.7) |
4.1 (3.8–4.7) |
4.1 (3.7–4.6) |
4.1 (3.6–4.7) |
4.1 (3.6–4.7) |
| Lactate (mmol/L) |
2.3 (1.6–3.3) |
2.2 (1.6–3.2) |
2.2 (1.5–3.1) |
2.3 (1.6–3.4) |
2.3 (1.6–3.4) |
| Blood urea nitrogen (mg/dL) |
27 (17–46) |
27 (17–46) |
29 (18–50) |
27 (17–45) |
27 (16–45) |
| Creatinine (mg/dL) |
1.2 (0.8–2.1) |
1.1 (0.8–1.9) |
1.2 (0.8–2.3) |
1.3 (0.8–2.0) |
1.2 (0.8–2.1) |
| WBC (103/μL) |
10.6 (6.9–15.5) |
10.8 (7.0–15.2) |
10.6 (6.8–15.3) |
10.0 (6.7–15.2) |
10.8 (6.7–15.9) |
| Hemoglobin (mg/dL) |
10.0 (8.5–11.6) |
9.9 (8.4–11.5) |
10.0 (8.6–11.5) |
10.1 (8.5–11.7) |
9.9 (8.4–11.7) |
| Platelet (103/μL) |
177 (115–253) |
177 (110–262) |
169 (119–250) |
179 (115–249) |
177 (109–253) |
| Clinical scores, median (IQR) |
|
|
|
|
|
| The counts of SIRS |
2 (1–3) |
2 (1–3) |
2 (1–3) |
2 (1–3) |
2 (1–3) |
| SOFA score |
6 (4–8) |
6 (4–8) |
6 (4–9) |
6 (4–9) |
6 (4–8) |
| Time from ED visit to ICU admission (hour), median (IQR) |
4.6 (3.3–6.1) |
2.6 (2.1–3.0) |
3.9 (3.7–4.3) |
5.3 (4.9–5.6) |
7.1 (6.5–7.8) |
| Disposition |
|
|
|
|
|
| 28-d mortality, n (%) |
258 (14) |
59 (13) |
73 (16) |
68 (15) |
58 (13) |
ED = emergency department, IQR = interquartile range, SIRS = systemic inflammatory response syndrome, SOFA = Sequential Organ Failure Assessment.
Association Between ED-LOS and 28-Day Mortality
The Lowess curve had a left-skewed distribution (Fig. 2), and patients who were immediately admitted to the ICU (e.g., ED-LOS of < 2 hr) tended to have a higher 28-day mortality rate.
Figure 2.: Association between time from ED visit to ICU admission and 28-day mortality using a locally weighted scatterplot smoother curve. We depicted the association between time from ED visit to ICU admission (i.e., ED length of stay [ED-LOS]) and 28-day mortality using a locally weighted scatterplot smoother (Lowess) curve. The blue line in the graph shows the estimated regression curve and the gray interval shows the 95% CI. ED-LOS = ED length of stay, ED = emergency department.
In the multivariable logistic regression model using ED-LOS as a continuous variable, ED-LOS was not significantly associated with 28-day mortality (adjusted odds ratio [OR] per hour increase, 1.04; 95% CI, 0.96–1.13; p = 0.30). However, as shown in Figure 3, patients in the higher time quartiles had higher 28-day mortality compared with those in the lowest time quartile (< 3.3 hr) (e.g., adjusted OR for patients in the second time quartile [3.3–4.5 hr], 1.59; 95% CI, 1.03–2.46; p = 0.04).
Figure 3.: Associations between time from ED visit to ICU admission and 28-day mortality using multivariable logistic regression. We examined the association between time from ED visit to ICU admission (i.e., ED length of stay [ED-LOS]) and 28-day mortality using a multivariable logistic regression model. Points and error bars indicate the adjusted OR and 95% CI, respectively, with <3.3 hr from ED visit to ICU admission as the reference. ED-LOS = ED length of stay, ED = emergency department; OR = odds ratio.
DISCUSSION
Using the data on 1,849 patients with sepsis from the MIMIC-IV and MIMIC-IV-ED databases, we found that patients in the lowest ED-LOS quartile (< 3.3 hr) had lower 28-day mortality than those in the higher ED-LOS quartiles (> 3.3 hr).
To our knowledge, this is the first study to statistically ascertain the optimal ED-LOS within 6 hours for patients with sepsis for whom compliance with the sepsis bundle is required. Patients in the lowest ED-LOS quartile (< 3.3 hr) were likely to have higher crude mortality but were significantly associated with lower mortality than those in the higher ED-LOS quartiles after adjustment for prognostic factors of sepsis. Therefore, immediate admission to the ICU to comply with the sepsis bundle may be effective in improving the prognosis of sepsis. Our findings suggest that a more immediate ICU admission (e.g., 3 hr) than 6 hours may be preferable as the target ED-LOS, which is consistent with the findings reported by Groenland et al (7). and Chalfin et al (8) in critically-ill patients that ICU admission within 2.4 hours and 3 hours were significantly associated with better outcomes, respectively.
Actually, there are conflicting findings as to whether ICU admissions within 6 hours of ED visits are associated with better prognostic outcomes among patients with sepsis (10,11). Although Angstin et al and Zhang et al (10,11) came to different conclusions from their own results, both studies reported higher mortality risk in patients with an ED-LOS of greater than or equal to 6 hours than those with an ED-LOS of less than 6 hours using adjusted logistic regression models. On top of this, considering our findings that patients who were admitted to the ICU within 3.3 hours tended to have a lower, although not significant, mortality rate than those who were admitted to the ICU within greater than 6.1 hours, it may be more beneficial to transfer patients to the ICU promptly after initial assessment and treatment at the ED, rather than necessarily using 6 hours as an optimal target.
Longer ED-LOS could be largely attributed to the lack of admitting beds resulting in the boarding of critically-ill patients in the ED, a critical issue that needs to be resolved (5,17). However, given that the median ED-LOS of critically-ill patients in the United States was shorter than 6 hours (79 min [IQR, 36–145] in the general population and 197 min [IQR, 112–313] in mechanically ventilated patients) (18,19), 3-hour target from ED visit to ICU admission of patients with sepsis requiring intensive care is feasible. Therefore, if medical resources permit, ICU admission within 3 hours of the ED visits may be beneficial in improving the prognosis of sepsis.
Our study has several limitations. First, the database used in this study covers patients enrolled from 2011 to 2019, suggesting that care for these patients may not have strictly adhered to the sepsis bundle proposed in SSCG 2021. Second, we considered patients with suspected infection and a SOFA score of 2 or greater upon ICU admission to be patients with sepsis according to previous studies using retrospective cohorts (20–22). However, this may have led to misclassification. Third, we had missing data on vital signs and laboratory parameters of patients eligible for analysis, which could be a potential source of bias. However, we believe this issue was minimized by using random forest imputation after excluding patients whose vital signs or laboratory data were all missing (23). In addition, the outcome measure used in this study, 28-day mortality, was limited to in-hospital mortality; we were unable to track patient outcomes after transfer or discharge from the ED to home. Fourth, because of the limited data available in the MIMIC-IV and MIMIC-IV-ED databases, we were not able to obtain data on medical interventions (e.g., administration of antibiotics, the performance of imaging tests) and changes in vital signs over time in the ED, which may be a source of confounding bias. In addition, we could not fully investigate the mechanism of the association between ED-LOS and the 28-day mortality rate in our study. These issues should be addressed in future prospective studies. These issues should be addressed in future prospective studies. Fifth, our study was not preregistered in an authorized clinical trial registry (e.g., UMIN-CTR), but our study protocol was approved by an institutional review board. Finally, the databases we used in this study are from a single center in the United States, which may limit the generalizability of our results. To verify our findings, multicenter studies including a variety of geographic areas (e.g., hospitals in rural areas or outside the United States) are warranted to consider differences in-hospital care systems.
CONCLUSIONS
Earlier admission to the ICU (e.g., within 3 hr of ED visits) was associated with lower 28-day mortality in patients with sepsis. For patients with sepsis who require intensive care, a more immediate ICU admission than 6 hours may result in better patient outcomes.
REFERENCES
1. Vincent J-L, Marshall JC, Ñamendys-Silva SA, et al.; ICON investigators: Assessment of the worldwide burden of critical illness: The Intensive Care Over Nations (ICON) audit. Lancet Respir Med 2014; 2:380–386
2. Fleischmann C, Scherag A, Adhikari NK, et al.; International Forum of Acute Care Trialists: Assessment of global incidence and mortality of hospital-treated sepsis. Current estimates and limitations. Am J Respir Crit Care Med 2016; 193:259–272
3. Evans L, Rhodes A, Alhazzani W, et al.: Surviving sepsis campaign: International guidelines for management of sepsis and septic shock 2021. Intensive Care Med 2021; 47:1181–1247
4. Evans L, Rhodes A, Alhazzani W, et al.: Surviving sepsis campaign: International guidelines for management of sepsis and septic shock 2021. Crit Care Med 2021; 49:e1063–e1143
5. Mohr NM, Wessman BT, Bassin B, et al.: Boarding of critically ill patients in the emergency department. Crit Care Med 2020; 48:1180–1187
6. Cardoso LT, Grion CM, Matsuo T, et al.: Impact of delayed admission to intensive care units on mortality of critically ill patients: A cohort study. Crit Care 2011; 15:R28
7. Groenland CNL, Termorshuizen F, Rietdijk WJR, et al.: Emergency department to ICU time is associated with hospital mortality: A registry analysis of 14,788 patients from six university hospitals in the Netherlands. Crit Care Med 2019; 47:1564–1571
8. Chalfin DB, Trzeciak S, Likourezos A, et al.: Impact of delayed transfer of critically ill patients from the emergency department to the intensive care unit. Crit Care Med 2007; 35:1477–1483
9. O’Callaghan DJ, Jayia P, Vaughan-Huxley E, et al.: An observational study to determine the effect of delayed admission to the intensive care unit on patient outcome. Crit Care 2012; 16:R173
10. Agustin M, Price LL, Andoh-Duku A, et al.: Impact of delayed admission to the intensive care unit from the emergency department upon sepsis outcomes and sepsis protocol compliance. Crit Care Res Pract 2017; 2017:9616545
11. Zhang Z, Bokhari F, Guo Y, et al.: Prolonged length of stay in the emergency department and increased risk of hospital mortality in patients with sepsis requiring ICU admission. Emerg Med J 2019; 36:82–87
12. Johnson A, Bulgarelli L, Pollard T, et al.: MIMIC-IV-ED (version 1.0). PhysioNet. Available at:
https://doi.org/10.13026/77z6-9w59. Accessed November 17, 2022
13. Johnson A, Bulgarelli L, Pollard T, et al.: MIMIC-IV (version 1.0). PhysioNet, 2021. Available at:
https://doi.org/10.13026/s6n6-xd98. Accessed November 17, 2022
14. Singer M, Deutschman CS, Seymour CW, et al.: The third international consensus definitions for sepsis and septic shock (sepsis-3). JAMA 2016; 315:801–810
15. Stekhoven DJ, Buhlmann P: MissForest—non-parametric missing value imputation for mixed-type data. Bioinformatics 2012; 28:112–118
16. Cleveland WS, Devlin SJ: Locally-weighted regression: An approach to regression analysis by local fitting. J Am Stat Assoc 1988; 83:596–610
17. Mathews KS, Durst MS, Vargas-Torres C, et al.: Effect of emergency department and ICU occupancy on admission decisions and outcomes for critically ill patients. Crit Care Med 2018; 46:720–727
18. Pitts SR, Vaughns FL, Gautreau MA, et al.: A cross-sectional study of emergency department boarding practices in the United States. Acad Emerg Med 2014; 21:497–503
19. Lone NI, Walsh TS: Prolonged mechanical ventilation in critically ill patients: Epidemiology, outcomes and modelling the potential cost consequences of establishing a regional weaning unit. Crit Care 2011; 15:R102
20. Cui L, Bao J, Yu C, et al.: Development of a nomogram for predicting 90-day mortality in patients with sepsis-associated liver injury [published March 4, 2023]. Sci Rep 2023; 13:3662
21. Zhao Q, Xiao J, Liu X, et al.: The nomogram to predict the occurrence of sepsis-associated encephalopathy in elderly patients in the intensive care units: A retrospective cohort study [published February 2, 2023]. Front Neurol 2023; 14:1084868
22. Huo Y, Wang X, Li B, et al.: Impact of central venous pressure on the mortality of patients with sepsis-related acute kidney injury: A propensity score-matched analysis based on the MIMIC IV database. Ann Transl Med 2022; 10:199
23. Shah AD, Bartlett JW, Carpenter J, et al.: Comparison of random forest and parametric imputation models for imputing missing data using MICE: A CALIBER study. Am J Epidemiol 2014; 179:764–774
