Hypotension and shock in the emergency department (ED) is associated with increased mortality (1). Early recognition and appropriate treatment of patients in shock decreases this mortality substantially (2,3). The priorities in the ED include restoration of intravascular volume, infusion of vasopressors, and respiratory support. Yet, both inadequate and over-resuscitation have been shown to be associated with a higher risk of mortality (3–15), which has led to debate over the individual components of resuscitation. This is particularly true for fluid resuscitation, where debate exists over determining volume responsiveness, fluid tolerance, the optimal resuscitation fluid (16–25).
Traditionally, invasive monitoring of cardiac filling pressures and cardiac output were used to guide resuscitation. These invasive measures have faded due to lack of correlation with volume responsiveness (26,27) and potentially harmful patient outcomes (28). Point-of-care ultrasound (POCUS) is now widely available in the ED and ICU and provides easily obtainable noninvasive evaluations of cardiopulmonary function. In an uncontrolled before and after study, POCUS shortened ED length-of-stay, reduced time to laboratory testing, and reduced time to completion of CT imaging in critically ill patients (29). POCUS performed on hypotensive patients presenting to the ED is considered an appropriate diagnostic tool to aid in the preliminary diagnosis (30,31). POCUS in patients with hypotension or shock has been reported to reduce diagnostic uncertainty and has potential to guide resuscitation (31–35); however, outcomes data are lacking. The objective of this study was to examine the association between POCUS and outcomes in critically ill nontraumatic ED patients. Our hypothesis was that early POCUS evaluation prior to resuscitation would be associated with decreased mortality.
MATERIALS AND METHODS
This study was conducted at two academic medical centers. Each hospital has fully staffed EDs and ICUs. The Banner University Medical Center-Tucson campus has an annual ED census of 85,000 and is a Level 1 Trauma center. The Banner University Medical Center-South campus has an annual ED census of 54,000. Emergency medicine (EM) faculty staff both EDs and each medical center supports its own EM residency, although the entire faculty and residents are shared between both EDs. The medical ICUs at each hospital are staffed with residents, pulmonary and critical care fellows, and faculty from the University of Arizona Department of Medicine.
An electronic quality improvement registry is maintained at each facility that includes all patients that are either admitted from the ED to an ICU or die while in the ED. This study was a retrospective cohort study of all adult (≥ 18 yr old) nontrauma patients with hemodynamic instability in the registry at both EDs from November 1, 2013, to October 31, 2016. Hemodynamic instability was defined as a shock index (heart rate/systolic blood pressure) greater than 0.6 (36). Pediatric and trauma patients were excluded. This project adhered to the standards outlined by the Patient-Centered Outcomes Research Institute for the conduct of registry studies (37). This study was reviewed and approved by the University’s Institutional Review Board.
Each ED has dedicated ultrasound systems for bedside use, and the department supports an emergency ultrasound section of faculty and fellows that provide a rigorous ultrasound education to EM residents and faculty. Residents on average graduate with 550 logged POCUS examinations. Ultrasound examinations performed in the ED are archived in the web-based workflow solution, Qpath (Telexy, Maple Ridge, BC, Canada), and undergo a structured quality control process.
Eligible patients were categorized into one of three groups depending on if and when POCUS was performed in relation to a key therapeutic intervention, which include any fluid bolus, vasopressor bolus, or continuous vasopressor infusion. The timing of POCUS examination was determined from the ED arrival time to the first POCUS performed. If no, POCUS was performed, patients were categorized into cohort 1. When the POCUS was performed within 6 hours but prior to an intervention, the patient was categorized into cohort 2. The remaining patients had POCUS performed after an intervention and were categorized into cohort 3. An a priori subgroup analysis of patients in each cohort with an admission diagnosis of sepsis, severe sepsis, or septic shock was planned.
The primary outcome was death at any point in the hospital stay. Secondary outcomes included proportion of patients that required tracheal intubation, vasopressors, or intubation and the time to each intervention.
All statistics were performed in Stata 15.1 (StataCorp LLC, College Station, TX). Categorical data were compared using Fisher exact test, and continuous data were compared with the Kruskal-Wallis test. Proportions and 95% CIs were calculated using the Pearson-Clopper exact method for all binary and categorical data, and medians and 95% CIs were calculated using median regression for all continuous variables. A two-sided p value of less than 0.05 was considered significant. For the primary multivariable logistic regression model, a purposeful backward stepwise regression analysis with a threshold for retention in the model of p value of less than 0.20 was constructed for the primary outcome, the primary independent variable (POCUS use) and clinically relevant potential risk factors and confounders. The threshold for inclusion into the multivariable model was a p value of less than 0.2. All variables initially not included in the preliminary model were included one at a time for a final check of statistical significance and confounding (variables that changed the coefficient for the use of POCUS before the first intervention vs no POCUS > 10% were considered significant confounders).
We conducted several sensitivity analyses to account for the potential bias due to the nonrandom assignment of the use of POCUS and the potential bias due to missing data for various risk factors and confounders.
First, we used inverse-probability-weighted regression-adjustment (IPWRA) to estimate the effect of using ultrasound before interventions compared with not using ultrasound to correct for potential bias due to the nonrandom distribution of POCUS use. IPWRA uses the reciprocals of the estimated treatment probability, calculated as a propensity score from a treatment model with the treatment assignment (POCUS before intervention vs no POCUS) as the outcome variable, as weights to estimate the effect of the treatment on the main outcome, death before hospital discharge, in a logistic regression outcome model (38). We also used propensity score matching, using the propensity scores to match each case (POCUS before—cohort 2) to 1 to 3 controls (no POCUS—cohort 1). We also stratified cohorts 1 and 2 into quintiles based on the propensity scores and conducted our logistic regression analyses separately for each quintile to assess if results were consistent across the range of propensity scores. We formally tested that covariates were balanced for the IPWRA analysis using the overidentification test for covariate balance. For the propensity score analyses, we visually examined the distribution of the propensity scores and covariates.
In addition, we used multiple imputation using chained equations (39) to impute data for variables with missing values to control for bias due to missing data and we repeated all analyses (multivariable logistic regression, IPWRA, propensity score-matched analysis) using the imputed datasets and Rubin’s rules (40) to combine estimates and variances across imputed datasets. In our imputation models, we included all variables that could be related to the outcome or treatment (see outcome and treatment model covariates above), using predictive mean matching (oxygen saturation, respiratory rate, lactate level), least squares regression (mean arterial pressure, body temperature, heart rate), and logistic regression (sex, triage acuity), for the chained equations. We created 25 imputed datasets for analysis.
Model diagnostics for influential observations were performed, and model performance was evaluated with the Hosmer-Lemeshow test, calibration belt test (41), link test, and the area under the receiving operator characteristics curve. All continuous independent variables were tested to ensure a linear relationship with the outcome variable in the log-odds (logit) scale and were transformed using fractional polynomials to ensure a linear association with the outcome variable in the logit. To ensure that covariates used to calculate the propensity score were balanced between treatment and control groups, we used the overidentification test for covariate balance (42) and visually inspected the distribution of propensity scores and covariates for both the treatment and control groups to ensure overlapping distributions.
A total of 9,325 adult patients were included in the registry during the study period. Of these, 1,591 trauma patients were excluded. Of the 7,734 study eligible patients, 2,293 patients were excluded for not meeting shock index threshold. The remaining 5,441 patients were included in the analysis: 4,165 in cohort 1 (no POCUS), 614 in cohort 2 (POCUS prior to intervention), and 662 in cohort 3 (POCUS after intervention) (Fig. 1). Patient characteristics are listed in Supplementary Table 1 (Supplemental Digital Content 1, https://links.lww.com/CCX/A48).
Mortality was 22% in cohort 1, 29% in cohort 2, and 26% in cohort 3 (p < 0.001) (Table 1). In patients with an admission diagnosis of sepsis, severe sepsis, or septic shock, mortality was 29%, 44%, and 27%, respectively (p = 0.02). POCUS prior to an intervention was associated with a 37% increase in the odds of death compared to patients with no POCUS performed (adjusted odds ratio [aOR], 1.37; 95% CI, 1.09–1.72) (Table 2). POCUS after intervention was not associated with an increased odds of death compared to patients with no POCUS (aOR, 0.98; 95% CI, 0.79–1.24). Dropping overly influential observations had no meaningful changes in model variables.
Sensitivity analyses also showed that cohort 2 had higher mortality compared with cohort 1. Supplementary Figure 1 (Supplemental Digital Content 2, https://links.lww.com/CCX/A49; legend, Supplemental Digital Content 4, https://links.lww.com/CCX/A51) shows a forest plot comparing the odds ratio for death across all primary and secondary analyses as well as the difference in mortality between cohorts 2 and 1. The crude odds ratio for death was the highest (1.43 [1.19–1.73]), whereas the IPWRA analysis using multiple imputation showed the lowest (1.20 [1.001–1.41]). Similarly, the difference (treatment effect) in mortality between cohort 1 and cohort 2 (mortality of cohort 2–mortality of cohort 1) was similarly highest for the crude analysis (mean difference in absolute mortality, 0.068; 95% CI, 0.03–0.11) and lowest for the IPWRA analysis using multiple imputation (difference, 0.032; 95% CI, 0.001–0.066). Odds ratios and absolute mean differences were in between these values for the IPWRA and propensity score-matched analysis (complete case analysis), as well as the adjusted analysis and the propensity score-matched analysis using multiple imputation. Covariates were adequately balanced between the two groups (cohorts 1 and 2) for the IPWRA and propensity-matched analyses based on visual inspection of distributions for each covariate (data not shown), distributions for the propensity scores (Supplementary Fig. 2, Supplemental Digital Content 3, https://links.lww.com/CCX/A50; legend, Supplemental Digital Content 4, https://links.lww.com/CCX/A51) and the overidentification test for covariate balance for the IPWRA analysis(p = 0.42). A sensitivity analysis focused only upon those receiving vasopressors demonstrated that while unadjusted mortality differences remained, adjusted mortality differences are not significant between groups. (Table 3).
Finally, we also stratified cases based on the quintiles of the propensity scores calculated for the above analyses and calculated stratum-specific odds ratios for mortality before hospital discharge for cohort 2 (POCUS before) versus cohort 1 (no POCUS) as well as a pooled estimate. The test for homogeneity (p = 0.35) suggests that ORs are homogenous across the five strata (quintiles) of propensity scores. Odds ratios ranged from 2.6 (quintile 1) to 1.2 (quintile 5), with a Mantel-Haenszel pooled estimate of 1.32 (95% CI, 1.07–1.63). We also repeated the adjusted analysis using only individuals with a propensity score greater than 0.1 and less than 0.9. The aOR for death comparing cohort 2 to cohort 1 was 1.29 (95% CI, 1.004–1.658).
Interventions with IV fluids, vasoactive agents, and intubation are listed in Table 4. There was a significant difference in time to fluids, with cohort 2 having the longest time to fluid administration.
In this study, we explored the impact of POCUS performed in the ED on the outcomes in critically ill patients in our two large academic EDs. We found critically ill patients that had POCUS performed prior to an intervention (cohort 2) had a higher in-hospital mortality. Although a recent prospective clinical trial suggests the ED POCUS has no impact on clinical outcomes (43) in patients with undifferentiated shock patients, ours is the first to suggest an association with potentially adverse outcomes of ED POCUS in critically ill patients.
A key question that needs exploration is: Why POCUS in the ED would result in higher in-hospital mortality? In this cohort study, ED patients that had POCUS prior to intervention had less aggressive treatment interventions in the ED when measured by both the volumes and timing of IV fluid and rates of intubation. These findings suggest POCUS influenced the aggressiveness of the immediate resuscitation in the ED.
Potentially, POCUS results in significant delays in treatment in the ED enough to adversely affect outcomes. Another potential explanation is that there is a group of patients with diagnostic uncertainty that POCUS identified ED patients who otherwise would not have been considered critically ill. Specifically, the etiology of shock upon initial evaluation in the ED is inherently challenging. There are not established characteristics of nontraumatic hypotension and shock in the ED (1). Thus, there may be differences in the role of POCUS in ED patients with undifferentiated shock versus those with obvious etiologies. There also may be differences in the use of POCUS for diagnosis and categorization than for therapy titration. Clearly, POCUS has been proposed not only for diagnostic purposes in shock but also to determine central venous pressure (44) and to guide fluid resuscitation (45).
There may also be a difference in the impact of POCUS between fluids and vasopressors. A sensitivity analysis focused only upon those receiving vasopressors demonstrated that while unadjusted mortality differences remained, adjusted mortality differences are not significant between groups. Yet looking at only patients receiving vasopressors is a limitation. Fluids and vasopressors are the two main interventions available in the ED for a hemodynamically unstable patient. ED POCUS has been widely touted to distinguish the fluid status these patients (hypovolemic, euvolemic, hypervolemic) and whether contractility is compromised. Thus, the purpose of POCUS in the ED is to guide whether to give more fluids, less fluids, or start a vasopressor/inotrope. It is not the ultrasound that changes the outcomes; it is the treatment decisions that come from doing the ultrasound (fluids or vasopressors).
POCUS in critically ill patients has widespread enthusiasm given the potential for rapid, noninvasive, and easily repeatable assessments of hemodynamics. In the ICU setting, there is evidence that POCUS can positively impact care outcomes. In a recent study by Kanji, a hemodynamic-guided echocardiogram was performed in patients admitted to the ICU with undifferentiated shock at a median time of 11 hours from admission. They found an improvement in mortality, which appears due to a reduction in total fluids in day 1 (44). A recent analysis of the Medical Information Mart for Intensive Care-III database showed reduced odds of mortality in patients that received formal echocardiography, interpreted by a cardiologist, in the ICU (45).
Extrapolation of these findings with POCUS in the ICU setting to the ED setting has been acceptable in our institutions given that POCUS allows for earlier diagnosis and treatment, improves confidence in the diagnosis (29–35,43,46), and leads to changes in resuscitation strategy 25–30% of the time (35). Indeed, this has led to general recommendations and endorsements for POCUS to be performed in critically ill patients, including in the ED (46,47). Furthermore, recent literature has shown that a positive fluid balance, generally considered as “over-resuscitation,” is associated with a higher mortality in shock (4,5,7,8,10). An extension of this logic is that patients at risk of fluid overload, such as those with heart failure or renal failure, will be at greater risk of fluid overload. The best estimates of the prevalence of comorbid heart failure and renal failure are 20% and 10%, respectively (8,9,11). Additionally, the utility of the fluid bolus has recently come under scrutiny as large database analyses are conflicting over the contribution of the fluid bolus to the mortality reduction seen with early resuscitation in septic shock (9–11). For these reasons, we hypothesized that POCUS prior to any intervention in the ED should reduce mortality.
Yet, what is less clear is the accuracy of the diagnoses or the impact of the resulting therapeutic interventions. Sekiguchi et al (35) performed focused cardiac ultrasounds on septic patients presenting to their medical ICU. The therapeutic plan was changed in 27%, and confidence in the diagnosis was enhanced in 37%. However, on independent review, providers incorrectly classified left ventricular (LV) function in 40% and right ventricular function in 50%; and no mortality data were reported. In another recent study by Hu et al (48), there is only moderate agreement between physician sonographers in the interpretation of cardiac standstill. In light of our results, this raises the question if physicians are making incorrect decisions based on what they think they see with POCUS? If LV function is impaired, which can be accurately estimated by emergency physicians (49), what is the threshold for using an inotrope in contrast to vasopressors or fluids? A possible explanation for our findings is inaccurate interpretation of POCUS images, and the therapeutic decisions based on those images. Thus, POCUS-related changes in fluid administration and vasoactive agents due to incorrect interpretation could have contributed to the differences in mortality. Perhaps it is not the first POCUS study that matters, but the one that comes after an initial period of resuscitation.
There are several limitations to our study, and these results must be interpreted with caution. We developed our registry to prospectively follow the care and outcomes of all critically ill patients admitted to the ICU from our EDs. Yet there are inherent limitations in these types of registries, including the dependence upon available clinical data rather than standardized research data obtained in an investigator controlled study. We used our formal ED ultrasound program to assure compliance with training and competence of providers performing POCUS and certification of the quality of ultrasound images. Yet, we did not review each individual study for formal cardiac functional measurements. Our methods were designed to reduce bias, however even though our sensitivity analyses retained the significant association with mortality in the POCUS group, the effect sizes were smaller and may be a result of other unmeasured confounders. Our registry also does not include patients that were resuscitated in the ED and avoided ICU admission, which could potentially bias against POCUS. The care the patients received after ICU admission could have been different between the groups as well and could have affected our results.
Furthermore, there are limitations in the analysis of observational studies that require special attention (50) including: 1) Causal inference requires careful consideration of confounding, 2) Interpretation of results should not rely on the magnitude of p values, and 3) Results should be presented in a granular and transparent fashion. We do not assert in this report that POCUS is causal. We merely suggest that the impact of POCUS on care processes and outcomes needs to be further interrogated. Our methods were designed to address known confounders and reduce bias, however even though our sensitivity analyses retained the significant association with mortality in the POCUS group, the effect sizes were smaller and may be a result of other unmeasured confounders.
There may be unknown clinical and system confounders in this retrospective cohort analysis. We attempted to control for potential confounders with our IPWRA, which maintained an increase in probability of mortality with POCUS before any intervention. Yet the resulting study data are counter to our original hypothesis that the POCUS-guided resuscitation would improve outcomes. A key known confounder is severity of illness. We used the Emergency Severity Index (ESI) as a measure of severity of illness as the ESI is a good predictor of critical care outcomes (51). ESI has the advantages of reflecting the patient at the time of the initial encounter in the ED (before interventions, including POCUS), being available on all patients (not just a subset) and being uniformly applied across this all-inclusive patient registry of critically ill patients in the ED. With the exception of ESI, all severity of illness scales are calculated by the worst value for each variable in a 24-hour period. When studying an intervention such as POCUS, separating if any difference in severity of illness is a potential confounder or if it is a result of the intervention will be impossible.
Further explorations of the impact on patient outcomes and the optimal role of POCUS in the ED appear warranted despite the widespread adoption of POCUS. Previous technologies that were widely adopted by providers as valuable clinical aids, including MAST trousers, pulmonary arterial catheters, and mixed venous oxygen saturation monitors were eventually found not to impact patient outcomes and even to be harmful (52). Given that a recent randomized controlled trial found no improvement in mortality in undifferentiated shock patients with POCUS use in the ED (52), we propose a larger prospective investigation to define the optimal role and impact of POCUS in the ED resuscitation of critically ill patients
POCUS in the ED has been widely adopted by the providers in our academic hospital EDs. Contrary to our hypothesis, POCUS prior to a key intervention in the ED appears to be associated with a higher mortality in critically ill patients with hemodynamic instability. Key questions that need further exploration include how POCUS in the ED may impact resuscitation strategies, treatment times, and patient outcomes through larger prospective studies.
1. Holler JG, Bech CN, Henriksen DP, et al. Nontraumatic hypotension and shock in the emergency department and the prehospital setting, prevalence, etiology, and mortality: A systematic review. PLoS One 2015; 10:e0119331
2. Sebat F, Musthafa AA, Johnson D, et al. Effect of a rapid response system for patients in shock on time to treatment and mortality during 5 years. Crit Care Med 2007; 35:2568–2575
3. Rivers E, Nguyen B, Havstad S, et al.; Early Goal-Directed Therapy Collaborative Group: Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001; 345:1368–1377
4. Boyd JH, Forbes J, Nakada TA, et al. Fluid resuscitation in septic shock: A positive fluid balance and elevated central venous pressure are associated with increased mortality. Crit Care Med 2011; 39:259–265
5. Maitland K, Kiguli S, Opoka RO, et al.; FEAST Trial Group: Mortality after fluid bolus in African children with severe infection. N Engl J Med 2011; 364:2483–2495
6. Seymour CW, Cooke CR, Heckbert SR, et al. Prehospital intravenous access and fluid resuscitation in severe sepsis: An observational cohort study. Crit Care 2014; 18:533
7. Andrews B, Semler MW, Muchemwa L, et al. Effect of an early resuscitation protocol on in-hospital mortality among adults with sepsis and hypotension: A randomized clinical trial. JAMA 2017; 318:1233–1240
8. Balakumar V, Murugan R, Sileanu FE, et al. Both positive and negative fluid balance may be associated with reduced long-term survival in the critically ill. Crit Care Med 2017; 45:e749–e757
9. Leisman DE, Goldman C, Doerfler ME, et al. Patterns and outcomes associated with timeliness of initial crystalloid resuscitation in a prospective sepsis and septic shock cohort. Crit Care Med 2017; 45:1596–1606
10. Marik PE, Linde-Zwirble WT, Bittner EA, et al. Fluid administration in severe sepsis and septic shock, patterns and outcomes: An analysis of a large national database. Intensive Care Med 2017; 43:625–632
11. Seymour CW, Gesten F, Prescott HC, et al. Time to treatment and mortality during mandated emergency care for sepsis. N Engl J Med 2017; 376:2235–2244
12. Peake SL, Delaney A, Bailey M, et al.; ARISE Investigators; ANZICS Clinical Trials Group: Goal-directed resuscitation for patients with early septic shock. N Engl J Med 2014; 371:1496–1506
13. Yealy DM, Kellum JA, Huang DT, et al.; ProCESS Investigators: A randomized trial of protocol-based care for early septic shock. N Engl J Med 2014; 370:1683–1693
14. Mouncey PR, Osborn TM, Power GS, et al.; ProMISe Trial Investigators: Trial of early, goal-directed resuscitation for septic shock. N Engl J Med 2015; 372:1301–1311
15. Jiang LB, Zhang M, Jiang SY, et al. Early goal-directed resuscitation for patients with severe sepsis and septic shock: A meta-analysis and trial sequential analysis. Scand J Trauma Resusc Emerg Med 2016; 24:23
16. Yunos NM, Bellomo R, Hegarty C, et al. Association between a chloride-liberal vs chloride-restrictive intravenous fluid administration strategy and kidney injury in critically ill adults. JAMA 2012; 308:1566–1572
17. McCluskey SA, Karkouti K, Wijeysundera D, et al. Hyperchloremia after noncardiac surgery is independently associated with increased morbidity and mortality: A propensity-matched cohort study. Anesth Analg 2013; 117:412–421
18. Raghunathan K, Shaw A, Nathanson B, et al. Association between the choice of IV crystalloid and in-hospital mortality among critically ill adults with sepsis*. Crit Care Med 2014; 42:1585–1591
19. Shaw AD, Raghunathan K, Peyerl FW, et al. Association between intravenous chloride load during resuscitation and in-hospital mortality among patients with SIRS. Intensive Care Med 2014; 40:1897–1905
20. Macdonald SPJ, Taylor DM, Keijzers G, et al. REstricted Fluid REsuscitation in Sepsis-associated Hypotension (REFRESH): Study protocol for a pilot randomised controlled trial. Trials 2017; 18:399
21. Rochwerg B, Millen T, Austin P, et al.; Canadian Critical Care Trials Group: Fluids in Sepsis and Septic Shock (FISSH): Protocol for a pilot randomised controlled trial. BMJ Open 2017; 7:e017602
22. Self WH, Semler MW, Wanderer JP, et al. Saline versus balanced crystalloids for intravenous fluid therapy in the emergency department: Study protocol for a cluster-randomized, multiple-crossover trial. Trials 2017; 18:178
23. Semler MW, Self WH, Wang L, et al.; Isotonic Solutions and Major Adverse Renal Events Trial (SMART) Investigators; Pragmatic Critical Care Research Group: Balanced crystalloids versus saline in the intensive care unit: Study protocol for a cluster-randomized, multiple-crossover trial. Trials 2017; 18:129
24. Semler MW, Wanderer JP, Ehrenfeld JM, et al.; SALT Investigators * and the Pragmatic Critical Care Research Group; SALT Investigators: Balanced crystalloids versus saline in the intensive care unit. The SALT randomized trial. Am J Respir Crit Care Med 2017; 195:1362–1372
25. Semler MW, Self WH, Wanderer JP, et al.; SMART Investigators and the Pragmatic Critical Care Research Group: Balanced crystalloids versus saline in critically ill adults. N Engl J Med 2018; 378:829–839
26. Michard F, Teboul JL. Predicting fluid responsiveness in ICU patients: A critical analysis of the evidence. Chest 2002; 121:2000–2008
27. Kumar A, Anel R, Bunnell E, et al. Pulmonary artery occlusion pressure and central venous pressure fail to predict ventricular filling volume, cardiac performance, or the response to volume infusion in normal subjects. Crit Care Med 2004; 32:691–699
28. Connors AF Jr, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. SUPPORT investigators. JAMA 1996; 276:889–897
29. Singer AJ, Williams J, Taylor M, et al. Comprehensive bedside point of care testing in critical ED patients: A before and after study. Am J Emerg Med 2015; 33:776–780
30. Jones AE, Tayal VS, Sullivan DM, et al. Randomized, controlled trial of immediate versus delayed goal-directed ultrasound
to identify the cause of nontraumatic hypotension in emergency department patients. Crit Care Med 2004; 32:1703–1708
31. Sasmaz MI, Gungor F, Guven R, et al. Effect of focused bedside ultrasonography in hypotensive patients on the clinical decision of emergency physicians. Emerg Med Int 2017; 2017:6248687
32. Volpicelli G, Lamorte A, Tullio M, et al. Point-of-care multiorgan ultrasonography for the evaluation of undifferentiated hypotension in the emergency department. Intensive Care Med 2013; 39:1290–1298
33. Boyd JH, Sirounis D, Maizel J, et al. Echocardiography as a guide for fluid management. Crit Care 2016; 20:274
34. Shokoohi H, Boniface KS, Pourmand A, et al. Bedside ultrasound
reduces diagnostic uncertainty and guides resuscitation in patients with undifferentiated hypotension. Crit Care Med 2015; 43:2562–2569
35. Sekiguchi H, Harada Y, Villarraga HR, et al. Focused cardiac ultrasound
in the early resuscitation of severe sepsis and septic shock: A prospective pilot study. J Anesth 2017; 31:487–493
36. Berger T, Green J, Horeczko T, et al. Shock index and early recognition of sepsis in the emergency department: Pilot study. West J Emerg Med 2013; 14:168–174
37. Gliklich RE, Dreyer NA, Leavy MB, et al. Standards in the Conduct of Registry Studies for Patient-Centered Outcomes Research: A guidance document for the Patient-Centered Outcomes Research Institute2012Washington DC, Patient-Centered Outcomes Research Institute.
38. Wooldridge JM. Econometric Analysis of Cross Section and Panel Data. 2010Second Edition. Cambridge, MA, The MIT Press.
39. Royston P, White IR. Multiple Imputation by Chained Equations (MICE): Implementation in Stata. J Stat Softw 2011; 45:1–20
40. Rubin DB, Schenker N. Multiple imputation in health-care databases: An overview and some applications. Stat Med 1991; 10:585–598
41. Nattino G, Lemeshow S, Phillips G, et al. Assessing the calibration of dichotomous outcome models with the calibration belt. Stata J 2017; 17:1003–1014
42. Wyss R, Ellis AR, Brookhart MA, et al. The role of prediction modeling in propensity score estimation: An evaluation of logistic regression, bCART, and the covariate-balancing propensity score. Am J Epidemiol 2014; 180:645–655
43. Manasia AR, Nagaraj HM, Kodali RB, et al. Feasibility and potential clinical utility of goal-directed transthoracic echocardiography performed by noncardiologist intensivists using a small hand-carried device (SonoHeart) in critically ill patients. J Cardiothorac Vasc Anesth 2005; 19:155–159
44. Prekker ME, Scott NL, Hart D, et al. Point-of-care ultrasound
to estimate central venous pressure: A comparison of three techniques. Crit Care Med 2013; 41:833–841
45. Lee CW, Kory PD, Arntfield RT. Development of a fluid resuscitation protocol using inferior vena cava and lung ultrasound
. J Crit Care 2016; 31:96–100
46. Bednarczyk JM, Fridfinnson JA, Kumar A, et al. Incorporating dynamic assessment of fluid responsiveness into goal-directed therapy: A systematic review and meta-analysis. Crit Care Med 2017; 45:1538–1545
47. Labovitz AJ, Noble VE, Bierig M, et al. Focused cardiac ultrasound
in the emergent setting: A consensus statement of the American Society of Echocardiography and American College of Emergency Physicians. J Am Soc Echocardiogr 2010; 23:1225–1230
48. Hu K, Gupta N, Teran F, et al. Variability in interpretation of cardiac standstill among physician sonographers. Ann Emerg Med 2018; 71:193–198
49. Randazzo MR, Snoey ER, Levitt MA, et al. Accuracy of emergency physician assessment of left ventricular ejection fraction and central venous pressure using echocardiography. Acad Emerg Med 2003; 10:973–977
50. Lederer DJ, Bell SC, Branson RD, et al. Control of confounding and reporting of results in causal inference studies. Guidance for authors from editors of respiratory, sleep, and critical care journals. Ann Am Thorac Soc 2019; 16:22–28
51. Raita Y, Goto T, Faridi MK, et al. Emergency department triage prediction of clinical outcomes using machine learning models. Crit Care 2019; 23:64
52. Atkinson PR, Milne J, Diegelmann L, et al. Does point-of-care ultrasonography improve clinical outcomes in emergency department patients with undifferentiated hypotension? An international randomized controlled trial from the SHoC-ED investigators. Ann Emerg Med 2018; 72:478–489