Article, Cardiology

Presenting hemodynamic phenotypes in ED patients with confirmed sepsis

a b s t r a c t

Objectives: To derive distinct clusters of septic emergency department (ED) patients based on their presenting noninvasive hemodynamic (HD) measurements and to determine if any clinical parameters could identify these groups.

Methods: Prospective, observational, convenience study of individuals with confirmed Systemic Infection. Pre- senting, pretreatment noninvasive HD parameters were compiled using Nexfin (Bmeye/Edwards LifeSciences) from 127 cases. Based on normalized parameters, K-means clustering was performed to identify a set of variables providing the greatest level of intercluster discrimination and intracluster cohesion.

Results: Our best HD clustering model used 2 parameters: the cardiac index (CI [L/min per square meter]) and systemic vascular resistance index (SVRI [dynes.s/cm5 per square meter]). Using this model, 3 different patient clusters were identified.

Cluster 1 had high CI with normal SVRI (CI, 4.03 +- 0.61; SVRI, 1655.20 +- 348.08); cluster 2 low CI with increased vascular tone (CI, 2.50 +- 0.50; SVRI, 2600.83 +- 576.81); and cluster 3 very low CI with markedly elevated SVRI (CI, 1.37 +- 0.81; SVRI, 5951.49 +- 1480.16). Cluster 1 patients had the lowest 30-day overall mortality. Among clinically relevant variables available during the initial Patient evaluation in the ED age, heart rate and tempera- ture were significantly different across the 3 clusters.

Conclusions: Emergency department patients with confirmed sepsis had 3 distinct cluster groupings based on their presenting noninvasively derived CI and SVRI. Further clinical studies evaluating the effect of early cluster-specific therapeutic interventions are needed to determine if there are outcome benefits of ED HD phe- notyping in these patients.

(C) 2016


? Patients were enrolled in the PREMIUM registry in the emergency departments of the following institutions: Henry Ford Hospital, Detroit, Michigan; Detroit Receiving Hospital, Detroit, Michigan; Sant’ Andrea Hospital, Rome, Italy; and VU Medical Center, Amsterdam,


?? Conflicts of interest and sources of funding: The authors have no conflicts of interest to


? The PREMIUM registry was partially funded through an unrestricted research grant

from BMEYE, Edwards LifeSciences, Irvine, California.

* Corresponding author at: Department of Emergency Medicine, Henry Ford, Health System, 2799 West Grand Blvd., Detroit, MI 48202, USA. Tel.: +1 313 850 5930 (Mobile),

+1 313 916 1909 (Office); fax: +1 313 916 7437.

E-mail addresses: [email protected] (R.M. Nowak), [email protected]

(B.P. Reed), [email protected] (P. Nanayakkara), [email protected] (S. DiSomma), [email protected] (M.L. Moyer), [email protected] (S. Millis), [email protected] (P. Levy).

Infection-related conditions prompt over 10 million emergency de- partment (ED) visits in the United States per year with sepsis, the most concerning manifestation, causing an estimated 750 000 deaths annually [1]. Although severe sepsis and septic shock have been assessed and managed with guidance from the Surviving Sepsis Cam- paign (SSC) and Revisions April 2015 by the SSC Executive Committee [2], there are no specific therapeutic guidelines for all sepsis cases. Although severe sepsis and septic shock represent inherent, systemic hemodynamic (HD) derangements caused by infection, the only emer- gency medicine-based recommendations related to HD issues made by these guidelines are to use vasopressors for hypotension that does not respond to initial fluid resuscitation to maintain a mean arterial pres- sure greater than or equal to 65 mm Hg and to reassess volume status

0735-6757/(C) 2016

and tissue perfusion within the ensuing 6 hours for those with a mean arterial pressure less than or equal to 65 mm Hg after initial fluids or with a baseline lactate greater than or equal to 4 mmol/L. In particular, there is limited consideration of other potentially important HD param- eters that reflect forward blood flow or systemic resistance-parameters that may help gauge initial sepsis severity and guide Resuscitative efforts to address subtle perfusion deficits by optimizing fluid and Vasoactive medication administrations.

This may reflect a lack of understanding because it relates to the early HD course of sepsis and its response to therapy, perhaps explaining why approximately 25% of nonventilated, normotensive sep- tic patients with a lactate of between 2.0 and 3.9 mmol/L develop pro- gressive organ dysfunction after hospitalization [3]; 23% of patients with uncomplicated sepsis cases progress to severe sepsis or septic shock within 72 hours of admission [1]; 25% of septic patients admitted to the general medical floor require intensive care unit transfer within 48 hours [4]; and of the 12% of septic patients who develop septic shock within 48 hours of presentation, one-half develop the shock after the first 4 hours of ED arrival [5]. The current approach to evalua- tion of perfusion in sepsis involves physician clinical assessments and vital sign measurements, although the Blood pressure and heart rate (HR) have been shown to be unreliable for estimating the underly- ing HD status in critically ill patients [6]. It has also been shown that the prediction of the underlying HD profiles of chronically or acutely ill pa- tients by cardiologists, intensivists, and emergency physicians (EPs) using clinical evaluation alone is inaccurate and unreliable [6-12]. This then suggests that the initial ED evaluation of patients with sepsis is cur- rently being made based on imperfect/inadequate HD information.

We have shown previously that in acute ED illnesses, HD parameters can be easily measured and continuously monitored using noninvasive finger cuff technology [13]. If the HD assessments for patients with con- firmed sepsis are measured and analyzed, they could show different phenotypes. The development of HD-specific phenotypic driven care then might be achievable, perhaps leading to improved outcomes in pa- tients with sepsis. Although there is increasing use of noninvasive mea- sures of the HD status and function in the ED, most of it is in the research domain [14].

Recently, there has been published the third international consensus definitions for sepsis and septic shock [15]. However, these new defini- tions have not been used in previously reported sepsis trials, and it will take time for them to be incorporated into new studies and to be widely used clinically [16]. We studied the presenting HD cluster differences in patients with a final confirmed sepsis diagnosis (which included mostly sepsis but also a few severe sepsis and septic shock cases as defined by the SSC) but, at ED presentation, were thought to have sepsis by the treating ED physician.

Accordingly, the objectives of this study were to use novel finger cuff technology to measure the differences in the presenting underlying HD parameters in patients with confirmed sepsis enrolled in the PREMIUM (Prognostic Hemodynamic Profiling in the Acutely Ill Emergency De- partment Patient) registry, to analyze which of the measured HD vari- ables best distinguished different patient clusters, and to determine if available clinical information obtained during an ED encounter could predict the HD-based clusters identified.

Materials and methods

The PREMIUM International Registry was a large prospective obser- vational study using a novel monitoring device (Nexfin; BMEYE, Edwards LifeSciences, Irvine, CA) to noninvasively and continuously measure beat-to-beat HD measurements in ED adult patients with clin- ically suspected acute heart failure, sepsis, and stroke. Patients were en- rolled in 4 large Urban academic medical centers: Henry Ford Health System, Detroit, Michigan (coordinating center); Detroit Receiving Hos- pital, Wayne State University, Detroit, Michigan; VU University Medical Center, Amsterdam, Netherlands; and Sant’ Andrea Hospital, University

La Sapienza, Rome, Italy. The detailed descriptions and methods of the PREMIUM registry have been previously published [13]. This study was conducted with the ethical principles of the Declaration on Harmo- nization and the International Conference on Harmonization Guidance for Good Clinical Practice. The protocol was approved by each partici- pating center’s institutional review board, and informed consent was obtained from each subject before enrollment. The study was registered before being initiated ( NCT01208077).

For this analysis, we focused on the subset of PREMIUM registry pa- tients with confirmed sepsis. Individuals with suspected sepsis were en- rolled if they had acute symptoms and signs (b3 days), as clinically assessed by the EP thought to be due to sepsis, had blood cultures and/or blood lactate measurements planned, and could be monitored using the Nexfin device before any therapy was given (with the excep- tion of supplemental oxygen and intravenous fluids at b50 mL/h). Patients were deemed to have confirmed sepsis if the initial EP assess- ment (history and physical exam) was sepsis and both the final primary ED (after complete ED workup) and the primary hospital discharge di- agnoses were sepsis.

Patients were excluded from the PREMIUM registry if they were un- able to provide informed consent, could not be enrolled within 4 hours of ED arrival, had end-stage renal disease requiring hemo or Peritoneal dialysis, were suspected or confirmed to be pregnant, had ST-segment elevation acute myocardial infarction, were unavailable for 30-day follow-up, had an active Do Not Resuscitate status, were known to have aortic valvular disease, were transferred from another facility, were excessively agitated, had a left ventricular assist device, or were previously enrolled in the PREMIUM registry or currently enrolled in any therapeutic investigational study. For this specific analysis, patients without a confirmed sepsis diagnosis were also excluded.

Nexfin is a Food and Drug Administration-approved and CE-marked device that derives the digital artery BP noninvasively and continuously through proprietary finger cuff technology, based on the volume clamp method developed by Penaz and Wesseling [17,18]. Cardiac output (CO) and other HD variables are determined from a reconstructed Brachial artery waveform using the Nexfin CO-Trel pulse contour method [19,20], and systemic vascular resistance (SVR) is calculated. The CO, SVR, and other HD variables are displayed beat-to-beat in real time on an accompanying screen and are recorded so that any time averaging and trending over time can be determined using the Nexfin device (Fig. 1). Patient height and weight were entered into the Nexfin device so that indexed values for the measured HD parameters could be calcu- lated. The Nexfin device (now named ClearSight) has been shown to have acceptable measurements (limits of agreement of up to +-30% for CO [21]) when compared with invasive pulmonary artery catheter- derived measurements [20,22,23].

A trained clinical research associate applied the Nexfin device to consented/enrolled patients using a standardized previously reported methodology [13]. All treating EPs and ED nurses were blinded to the HD data (Nexfin screen covered) for the duration of the monitoring pe- riod, which was up to 4 hours for each study patient. length of hospital stay , inpatient mortality, and unscheduled medical visits were re- corded by phone and/or medical record review through 30 days postdischarge.

Patient baseline characteristics, comorbidities, ED vital signs, labora- tory results, LOS and mortality, and unscheduled medical visits through 30 days postdischarge were collected for confirmed sepsis cases. In nonintubated and nonsedated acutely ill patients, HD noninvasive con- tinuous monitoring shows short-term variances in parameters that are reflective of changes in the patient’s environment and their disease state/progression. We chose to use 15-minute averages of the measured untreated HD parameters to smooth out these variations, although in the future, shorter time periods may be shown to be equally adequate. For this analysis, we have included only the initial 15-minute averaged measurements and not the changes in these measurements over time or posttreatment. To determine which of the numerous available

Fig. 1. Photographs showing the various aspects of the Nexfin device used in the PREMIUM registry.

Table 1

Presenting mean hemodynamic parameters

Hemodynamic parameters (15-minute averaged)

Overall (N = 127)

Cluster 1 (N = 72)

Cluster 2 (N = 50)

Cluster 3 (N = 5)


Brachial SBP (mm Hg)

112.1 (30.04)

112.3 (24.74)

110.3 (31.40)

126.6 (71.9)


Brachial DBP (mm Hg)

63.7 (13.76)

64.0 (12.40)

62.2 (13.70)

75.0 (27.31)


Brachial mean BP (mm Hg)

81.3 (19.48)

81.9 (16.8)

79.5 (19.31)

92.2 (46.84)


Finger SBP (mm Hg)

104.4 (31.42)

106.3 (26.51)

100.5 (33.20)

118.5 (68.09)


Finger DBP (mm Hg)

53.5 (16.15)

54.4 (14.79)

51.4 (15.48)

61.8 (35.20)


Finger mean BP (mm Hg)

68.8 (19.64)

69.8 (17.4)

66.3 (19.30)

80.5 (43.72)


Aortic SBP (mm Hg)

99.4 (27.20)

99.5 (22.11)

98.4 (27.14)

109.5 (73.95)


Aortic DBP (mm Hg)

64.4 (14.98)

65.4 (13.12)

62.6 (14.22)

67.9 (38.22)


Aortic mean BP (mm Hg)

79.6 (19.87)

80.5 (17.04)

77.6 (19.10)

85.0 (52.17)


Maximum dp/dt (mm Hg/s)

1167.7 (614.91)

1220.3 (507.00)

1109.0 (666.56)

998.9 (1322.30)


Minimum dp/dt (mm Hg/s)

578.8 (379.77)

646.1 (299.04)

467.8 (363.43)

719.3 (1038.01)


dp/dt (mm Hg/s)

898.3 (498.96)

955.7 (393.75)

818.3 (533.39)

873.1 (1210.46)


HR (beat/min)

96 (19.73)


89.0 (19.23)

70.9 (18.0)


CO (L/min)

6.3 (2.09)

7.6 (1.47)

4.8 (1.28)

2.62 (1.59)


CI (L/min per square meter)

3.3 (1.01)


2.50 (0.50)



Left ventricular ejection time (s)

0.3 (0.04)

0.26 (0.03)

0.28 (0.08)

0.33 (0.08)


SV (mL)

66.7 (22.75)

75.7 (18.24)

56.8 (22.40)

37.6 (22.70)


SVI (mL/m2)

35.1 (10.26)

40.1 (8.33)

29.2 (8.09)

20.0 (10.43)


SVR (dyn*s/cm5)

1196.8 (669.73)

888.4 (198.49)

1414.9 (434.40)

3455.5 (1560.14)


SVRI (dyn*s/cm5 per square meter)

2196.6 (1029.1)

1655.2 (348.08)

2600.8 (576.81)

5651.5 (1480.16)


Presenting 15-minute averaged hemodynamic parameters with SDs in parentheses.

P refers to differences between the 3 clusters.

presenting HD parameters best differentiated HD subgroups within the confirmed sepsis cohort, cluster analysis was used.

The presenting averaged 15-minute normalized HD parameters se- lected (based on projected clinical relevance) for inclusion in cluster analysis were stroke volume (SV); stroke volume index (SVI); SVR; sys- temic vascular resistance index (SVRI); CO; cardiac index (CI); left ventricular ejection time; average, maximum, and minimum delta pressure/delta time (dp/dt); brachial systolic BP (SBP), Diastolic BP (DBP), and mean BP; finger SBP, DBP, and mean BP; aortic SBP, DBP, and mean BP; and HR.

K-means clustering was performed to identify a set of variables that provided the greatest level of intercluster discrimination and intracluster cohesion. The parameter k, representing the number of clusters, was identified iteratively by maximizing the ratio of inter- (dis- crimination) and intracluster (cohesion) dispersion. Among models with comparable ratios, those with fewer clusters (smaller k) were pre- ferred to minimize overfitting. Principal components analysis validates the choice of small k as appropriate for the data.

Cluster group comparisons were then performed using a multiway analysis of variance to determine if the 3 HD cluster groups were differ- entiated by patient baseline characteristics, comorbidities, ED vital signs, or laboratory testing results. Fifteen-minute averages, normaliza- tion, and cluster analysis were performed using R 3.0.2. Statistical com- parisons and summaries were performed using SAS version 9.4 (Cary, NC).


There were 127 patients with confirmed sepsis of 194 (66%) suspected cases enrolled in the PREMIUM registry. Enrollment per site was as follows: Henry Ford Health System, 41 (32.1%); Wayne State University, 1 (0.8%); VU University Medical Center, 66 (52.0%); and Uni- versity La Sapienza, 19 (15.1%). All enrolled study patients were admit- ted to the hospital for continued care. The 2-parameter model including CI (L/min/m2) and SVRI (dynes.s/cm5/m2) provided the best discrimi- nation between clusters, as measured by the ratio of inter- to intracluster dispersion (2.51). The presenting averaged HD parameters of clinical interest for all patients with sepsis and for each of the individ- ual clusters are shown in Table 1. Cluster 1 (n = 72) had high CI (4.03 +- 0.61) and normal SVRI (1655.20 +- 325). Cluster 2 (n = 50) showed low CI (2.50 +- 0.050) with increased SVRI (2600.83 +- 576.81). Cluster 3 (n = 5) had very decreased CI (1.37 +- 0.81) and markedly elevated SVRI (5951.49 +- 1480.16). Values for CI and SVRI between clusters 1,

2, and 3 were significantly different (all P b .001). The normal HD pa- rameter ranges referenced by the Nexfin manufacturer are CI 2.5 to

4.0 and SVRI 1200 to 2400. Fig. 2 shows the plotted values for the CI and SVRI for the 3 clusters and indicates the high and low values for these 2 HD parameters.

Overall and cluster-specific baseline patient characteristics, comor- bidities, and Home medications are shown in Table 2. Patient age was the only characteristic variable that differed significantly between the clusters. Although comorbidities and prescribed medications for sepsis were common in the overall cohort, the only significant differences across the 3 clusters were a history of heart failure and the home use of diuretics/vasodilators. The only initial ED clinical parameters (Table 3) that were significantly different among the 3 HD subgroups were the presenting HR, temperature, and creatinine values. Notably, the SBP and DBP, respiratory rate, pulse oximetry values, glucose, B- type natriuretic peptide (BNP), and lactate measurements were not sig- nificantly different among the 3 sepsis-clustered groups.

Fig. 2. The 3 sepsis cluster plots using the cardiac and SVR indices. The lines indicate the normal reference ranges for these parameters, as recommended by the Nexfin manufacturer.

Table 2

Patient characteristics, comorbidities,

and home medications

Overall (N = 127)

Cluster 1 (N = 72)

Cluster 2 (N = 50)

Cluster 3 (N = 5)


Age (y)

66 +- 16

57 +- 15

77 +- 10

80 +- 8


Weight (kg)

77 +- 21

75 +- 17

81 +- 27

68 +- 16



70 (55)

41 (57)

25 (25)

4 (80)



35 (28)

13 (26)

13 (26)

3 (60)



81 (64)

42 (58)

37 (74)

2 (40)



63 (50)

29 (40)

31 (62)

3 (60)


Myocardial infarction

15 (12)

8 (11)

7 (14)

0 (0)



18 (14)

10 (14)

7 (14)

1 (20)


Congestive heart failure

25 (20)

8 (11)

16 (32)

1 (20)



27 (21)

11 (15)

14 (28)

2 (40)



37 (29)

17 (24)

17 (34)

3 (60)


Liver disease

9 (7)

6 (8)

3 (6)

0 (0)


Renal disease

32 (25)

15 (21)

15 (40)

2 (40)


cocaine use

4 (3)

4 (6)

0 (0)

0 (0)


Alcohol use

15 (12)

11 (15)

3 (6)




19 (15)

14 (19)

5 (10)

0 (0)



45 (35)

17 (24)

27 (54)

1 (20)



7 (6)

2 (3)

3 (6)



COPD, chronic obstructive pulmonary disease.

Means with SDs are shown. All parentheses are percent values.

P refers to differences between the 3 clusters.

The clinical outcomes documented for all patients with sepsis and for each of the 3 clusters are also shown in Table 3.A significant difference seen between the clusters was the overall 30-day mortality, with most deaths (20%) occurring in clusters 2 and 3 (low CI) and only in 5.6% of cluster 1 patients (high CI). Overall, the patients with confirmed sepsis had an overall 30-day mortality of 11.8%, a mean LOS of 8.7 days, and an unscheduled outpatient visit rate within 30 days of discharge of 10.2%.


This is the first report, to our knowledge, of the monitoring and re- cording of the presenting HD parameters of not yet treated ED patients with a confirmed primary diagnosis of sepsis. On presentation, septic

patients resemble each other based on the usual clinical parameters available but have different underlying HD profiles that can be clustered into 3 separate phenotypes. Importantly, these unique clusters cannot reliably be identified using current ED assessments and laboratory test- ing. Although HR was different between groups, an inverse relationship with mortality was noted, challenging common current perspectives on risk modification in sepsis. Perhaps, the lack of knowledge of these un- derlying HD profiles explains in part why so many patients with uncom- plicated sepsis deteriorate once admitted to the hospital. It could be that HD-guided therapy might result in optimal fluid resuscitation/forward blood flow/systemic perfusion and thus result in improved outcomes.

Existing approaches to HD assessment have inherent limitations, particularly as factors to guide therapy. Central venous pressure, if

Table 3

Baseline vital signs, laboratory values, and outcomes

Overall (N = 127)

Cluster 1 N = 72)

Cluster 2 (N = 50)

Cluster 3 (N = 5)


SBP (mm Hg)

123.8 +- 28.9

123.6 +- 27.8

124.0 +- 28.4

126.0 +- 29.9


DBP (mm Hg)

69.8 +- 16.0

69.3 +- 14.4

70.2 +- 17.9

71.4 +- 20.9



99.4 +- 21.1

106.0 +- 18.8

92.1 +- 19.9

75.6 +- 27.7


Respiratory rate

21.7 +- 6.6

20.9 +- 5.1

22.9 +- 8.2

21.5 +- 6.2


Temperature (C)

37.8 +- 1.4

38.2 +- 1.1

37.3 +- 1.5

36.5 +- 1.7


Pulse oximetry (%)

95.9 +- 4.1

96.4 +- 3.5

95.6 +- 3.9

93.0 +- 10.7


Recent EF (%)

57.3 +- 11.8

61.5 +- 6.2

53.0 +- 15.3

Not done


BUN (mg/dL)

27.4 +- 39.2

18.7 +- 27.4

38.9 +- 50.9

30.1 +- 24.5


BUN (mmol/L)

76.9 +- 109.9

52.5 +- 76.8

109.0 +- 142.6

84.3 +- 68.5


Creatinine (mg/dL)

1.5 +- 1.0

1.29 +- 0.8

1.7 +- 1.1

2.38 +- 1.4


Creatinine (mmol)

113.8 +- 73.9

98.5 +- 59.7

129.1 +- 82.9

180.5 +- 180.6


Glucose (mg/L)

151 +- 227

167.7 +- 295.0

124.6 +- 57.6

151 +- 84.5


Glucose (mmol)

8.8 +- 12.6

9.3 +- 16.4

6.9 +- 3.2

8.4 +- 4.7


BNP (pg/mL)

540 +- 1329

277.1 +- 230.7

775.3 +- 1779.6

146.5 +- 195.9


NT pro-BNP (ng/mL)

2898 +- 2806

2092.3 +- 3192.3

3502.8 +- 2793.9

Not done


Lactate (mg/dL)

2.6 +- 2.7

2.2 +- 2.5

3.0 +- 3.0

2.9 +- 1.3


Procalcitonin (ng/mL)

15.1 +- 34.3

34.7 +- 57.4

4.8 +- 2.3

11.4 (1)


WBC (x103)

14.0 +- 6.9

14.2 +- 6.5

14.2 +- 7.5

9.5 +- 3.9


Neutrophils (%)

77.9 +- 16.9

77.5 +- 18.6

78.1 +- 16.3

79.5 +- 9.6


Inpatient deaths (%)

12 (9.5)

4 (5.6)

7 (14.0)

1 (20.0)


30-day postdischarge deaths (%)

3 (2.4)

0 (0)

3 (6.0)

0 (0.0)


Overall 30-day deaths (%)

15 (11.8)

4 (5.6)

10 (20.0)

1 (20.0)


>=1 30-day visit (%)

13 (10.2)

8 (11.0)

4 (8.0))

1 (20.0)


LOS (d)


8.7 +- 9.3

8.2 +- 8.5

9.7 +- 10.8

7.2 +- 4.3












Mean values with SDs are shown. P refers to difference between the 3 clusters.

BUN, blood urea nitrogen; EF, ejection fraction; IQR, interquartile ranges; WBC, white blood cell count. LOS, data missing for 3 patients, all in cluster 1.

a P for LOS is derived nonparametrically using the Kruskal-Wallis ?2 statistic.

maintained at 8 to 12 mm Hg, may indicate that a septic person is not volume deficient, but it does not predict the lack of continued fluid re- sponsiveness [24-26]. The gold standard for continued fluid responsive- ness is an improvement of SV or CO by greater than or equal to15% after a 500-mL bolus fluid challenge [27]. A ScvO2 greater than or equal to 70% is an indicator of adequate oxygen delivery but not oxygen con- sumption, and its normalization alone has been not shown to improve clinical outcomes [28-30]. The use of bedside ultrasonography to deter- mine the venaCaval index as guidance for fluid administration may identify volume deficiency in patients but does not reliably predict fur- ther fluid responsiveness [31]. The use of dynamic HD changes with passive leg raising (causing a reversible fluid challenge) requires im- provements in SV, radial Pulse pressure, or femoral artery flow to deter- mine fluid responsiveness in nonMechanically ventilated patients [32]. However, a recent single-center study reported that using passive leg raising to determine fluid responsiveness with bioreactance-initiated HD monitoring (fluid responsive if SV increase was N10%) to guide fluid therapy in septic patients with a lactate greater than 3.0 mmol/L did not show improvement in Lactate clearance at 3 hours [33].

The use of easily applied finger cuff monitoring allows the continu- ous noninvasive monitoring of forward blood flow (SVI and CI) and so can be used to optimize fluid therapy based on forward blood flow in- creases. It can be applied to uncomplicatED septic patients because those admitted to the floor currently have significant rates of deteriora- tion that may be preventable with fluid and resulting CI/SVRI optimiza- tions. Those with severe sepsis or septic shock receiving early goal- directed therapy (EGDT) may also benefit from continuous finger cuff-derived monitoring because Napoli [34] has demonstrated that EGDT-treated patients who died (25.9%) had similar central venous pressure and lactate measurements but significantly lower SV and CI values (determined by impedance cardiography). The EGDT guidelines used would not have been able to determine fluid responsiveness and thus optimization of the SV and CI and hence perfusion. Improving for- ward flow in these patients may have resulted in less mortality in the EGDT patient group.

Is the use of noninvasive HD monitoring with phenotypic determi- nation in all sepsis cases clinically valuable? Our cluster I patients had high CI and normal SVRI values. This is a normal initial response to sepsis: increased forward blood flow with associated normal or lower SVRI in an attempt to meet increased metabolic demands. Cluster 2 demonstrated a low CI and increased SVRI, and so, these patients might have benefitted from CI and perfusion improvement by fluid responsiveness-directed therapies. Cluster 3 with its very low CI and markedly increased SVRI would definitely require HD-guided fluid ad- ministration and possible vasodilator therapy to increase the CI and sys- temic perfusion. Patients with a history of congestive heart failure and those treated with diuretics or vasodilators require HD monitoring because these patients were seen most often in clusters 2 and 3 (with decreased CI). This suggests that, at the least, a complex interaction be- tween underlying Cardiovascular function and presenting HD profile in septic patients exists. How specifically the sepsis cluster phenotypes should be treated is the subject of further, ongoing investigations.

Although there are published scoring systems that risk stratify in- fected patients, there is little information available that links differential treatment and responses to therapy with patient outcomes based on these scoring patterns. Available Emergency Medicine (EM) scoring systems include the Mortality in Emergency Department Sepsis score for those ED patients with suspected infection; the Confusion, Urea ni- trogen, Respiratory rate, BP, 65 years and older score for those with pneumonia; and the Rapid Emergency Medicine score for nonsurgical ED patients. All of these scores effectively predict 28-day in-hospital mortality [35]. In addition, the recently described Sepsis Patient Evalua- tion in the Emergency Department score has also been shown to predict 28-day mortality in septic patients [36]. Very recently, it has been re- ported that patients with sepsis who have a bedside quick sequential (sepsis-related) organ failure assessment score with at least 2 of 3

criteria (respiratory rate >=22/min, altered mentation, and SBP

<=100 mm Hg) are more likely to have poorer outcomes [15]. All these scoring systEMS use varying combinations of patient historical informa- tion, physiologic data, and selected laboratory values, but none provide any specific guidance for individualizED patient care, whereas the defin- ing of HD phenotypes may give direction to optimal initial ED therapy and changes to care based on additional monitoring.

The PREMIUM registry was strictly an observational study in 4 urban academic centers that used convenience sampling and may not be a Representative sample of all ED sepsis cases. Patients had to be stable enough to provide informed consent and not require immediate clinical therapy, and so, our results may not be applicable to all critically ill pa- tients with sepsis. Although we have described the HD clustering of pa- tients with sepsis on ED arrival, we have not shown that targeting unique therapy for these specific groups improves clinical symptoms or outcomes (hospitalizations, LOS, 30-day deaths, unscheduled medi- cal visits, or Hospital readmissions) nor have we described what hap- pens to clustering over time with different therapies. In addition, as shown in Fig. 2, within each of the 3 clusters described, there are some intragroup HD profile differences that may be further relevant to the tailoring of specific therapy to the individual patient. Lastly, the di- agnosis of sepsis was not determined by independent adjudicators but was instead based on the treating clinicians. Although inherently prag- matic, such an approach to diagnosis may not be entirely accurate.


We have described the presenting HD parameters for ED patients with a diagnosis of confirmed sepsis. There is much variation in these underlying measurements, but through the use of cluster analysis, we have shown that 3 distinct HD-based sepsis phenotypes based on the CI and SVRI exist and that these cannot be reliably identified using more traditional clinical features such as demographics, comorbidities, home medications, ED vital signs, and results of laboratory testing. Fur- ther clinical studies are needed to determine if the individualization of acute therapy based on the different HD clusters for patients with sepsis will improve outcomes.


  1. Glickman SW, Cairns CB, Otero RM, Woods CW, Tsalik EL, Langley RJ, et al. Disease progression in Hemodynamically stable patients presenting to the emergency de- partment with sepsis. Acad Emerg Med 2010;17:383-90.
  2. Dellinger RP, Levy MM, Rhodes A, Annane D, Gerlach H, Opal SM, et al. Surviving Sepsis Campaign: international guidelines for the management of severe sepsis and septic shock: 2012. Crit Care Med 2013;41:580-635.
  3. Arnold RC, Sherwin R, Shapiro NI, O’Connor JL, Galspey L, Singh S, et al. Multicenter observational study of the development of progressive organ dysfunction and ther- apeutic interventions in normotensive sepsis patients in the emergency department. Acad Emerg Med 2013;20:433-40.
  4. Caterino JM, Jalbuena T, Bogucki B. Predictors of acute decompensation after admis- sion in ED patients with sepsis. Am J Emerg Med 2010;28:631-6.
  5. Capp R, Horton CL, Takhar SS, Ginde AA, Peak DA, Zane R, et al. Predictors of patients who present to the emergency department with sepsis and progress to septic shock between 4 and 48 hours of emergency department arrival. Crit Care Med 2015;43: 983-8.
  6. Wo CCJ, Shoemaker WC, Appel PL, Bishop MH, Kram HB, Hardin E. Unreliability of Blood pressure and heart rate to evaluate cardiac output in emergency resuscitation and critical illness. Crit Care Med 1993;21:218-23.
  7. Stevenson LW, Perloff JK. The limited reliability of physical signs for estimating he- modynamics in chronic heart failure. JAMA 1989;261:884-8.
  8. Eisenberg PR, Jaffe AS, Schuster DP. Clinical evaluation compared to pulmonary ar- tery catheterization in the hemodynamic assessment of critically ill patients. Crit Care Med 1984;12:549-53.
  9. Veale WN, Morgan JH, Beatty JS, Sheppard SW, Dalton ML, Van De Water JM. Hemo- dynamic fluid status in the trauma patient: are we slipping? Am Surg 2005;71: 621-6.
  10. Neath SX, Lazio L, Guss DA. Utility of impedance cardiography to improve physician estimation of hemodynamic parameters in the emergency department. Congest Heart Fail 2005;2005:17-20.
  11. Rodriguez RM, Lum-Lung M, Dixon K, Nothmann A. A prospective study on esophageal Doppler hemodynamic assessment in the ED. Am J Emerg Med 2006;24:658-63.
  12. Nowak RM, Sen A, Garcia AJ, Wilke H, Yang JJ, Nowak MR, et al. The inability of emer- gency physicians to adequately clinically estimate the underlying hemodynamic profiles of acutely ill patients. Am J Emerg Med 2012;30:954-60.
  13. Nowak RM, Nanayakkara P, DiSomma S, Levy P, Schrijver E, Huyghe R, et al. Noninvasive hemodynamic monitoring in emergency department patients with suspected heart fail- ure, sepsis and stroke: the PREMIUM registry. West J Emerg Med 2014;15:786-94.
  14. Middleton PM, Davies SR. noninvasive hemodynamic monitoring in the emergency department. Curr Opin Crit Care 2011;17:342-50.
  15. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitions for sepsis and shock (sepsis-3). JAMA 2016;337(15):801-10.
  16. Jacob JA. New sepsis diagnostic guidelines shift focus to organ dysfunction. JAMA 2016;315:739-40.
  17. Wesseling KH. A century of noninvasive arterial pressure measurement from Marey to Penaz and Finapres. Homeostasis 1995;36:2-3.
  18. Wesseling KH, De Wit B, Vander Hoeven GMA. Physical, calibrating finger vascular physiology for Finapres. Homeostasis 1995;36:67-82.
  19. Girdulich P, Prentza A, Wesseling KH. Models of brachial to finger pulse wave distor- tion and pressure decrement. Cardiovasc Res 1997;33:698-705.
  20. Bogert LW, Wesseling KH, Schera O, van Lieshout EJ, de Mol BAJM, van Goudoever J, et al. Pulse contour cardiac output derived from non-invasive arterial pressure in cardiovascular disease. Anesth 2010;65:119-25.
  21. Critchley LA, Critchley JA. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput 1999;15:85-91.
  22. Sokolski M, Rydlewska A, Krakowiak B, Biegus J, Zymlinski R, Waldemar B, et al. Comparison of invasive and non-invasive measurements of haemodynamic param- eters in patients with advanced heart failure. J Cardiovasc Med 2011;12:773-8.
  23. Stover JF, Stocker R, Lenherr R, Neff TA, Cottini SR, Zoller B, et al. Noninvasive cardiac output and Blood pressure monitoring cannot replace an invasive monitoring sys- tem in critically ill patients. BMC Anesthesiol 2009;9:6. 147-2253-9-6.
  24. Vincent JL, Weil MH. Fluid challenge revisited. Crit Care Med 2006;34:1333-7.
  25. De Witt B, Joshi R, Meislin H, Mosier JM. Optimizing oxygen delivery in the critically ill: as- sessment of volume responsiveness in the septic patient. J Emerg Med 2014;47:608-15.
  26. Marik PE, Baram M, Vahid B. Does the central venous pressure predict fluid respon- siveness? A systematic review of the literature and the tale of seven mares. Chest 2008;134:172-8.
  27. Cecconi M, Arulkumaran N, Kilic J, Ebm C, Fhodes A. Update on hemodynamic mon- itoring and management in septic patients. Minerva Anestesiol 2014;80:701-11.
  28. The ProCESS Investigators, Yealy DM, Kellum JA, Juang DT, et al. A randomized trial of protocol-based care for early septic shock. N Engl J Med 2014;370:1683-93.
  29. The ARISE Investigators and the ANZICS Clinical trials Group. Goal-directed resusci- tation for patient with early septic shock. N Engl J Med 2014;371:1496-506.
  30. Mouncey PR, Osborn TM, Power GS, for the ProMISE trial investigators, et al. Trail of early, goal-directed resuscitation for septic shock. N Engl J Med 2015;372:1301-10.
  31. De Valk S, Olgers TJ, Holman M, Ismael F, Ligtenberg JJM, ter Maaten JC. The caval index: an adequate non-invasive ultrasound parameter to predict fluid responsive- ness in the emergency department? BMC Anesthesiol 2014;14:114-9.
  32. Preau S, Saulnier F, Dewavrin F, Durocher A, Chagnon JL. Passive leg raising is predic- tive of fluid responsiveness in spontaneously breathing patients with severe sepsis or acute pancreatitis. Crit Care Med 2010;38:819-25.
  33. Kuan WS, Ibrahim I, Leong BSH, Jain S, Lu Q, Cheung YB, et al. Emergency depart- ment management of sepsis patients: a randomized, goal-oriented, noninvasive sep- sis trial. Ann Emerg Med 2016;67:367-78.
  34. Napoli AM, Machan JT, Corl K, Forcada A. The use of impedance cardiography in predicting mortality in emergency department patients with severe sepsis and sep- tic shock. Acad Emerg Med 2010;17:452-5.
  35. Howell MD, Donnino MW, Talmor D, Clardy P, Ngo L, Shapiro NL. Performance of se- verity of illness scoring systems in emergency department patients with infection. Acad Emerg Med 2007;14:709-14.
  36. Bewersdorf JP, Hautmann O, Kofink D, Khalil AA, Abidin IZ, Loch A. The SPEED (Sep- sis Patient Evaluation in the Emergency Department) score: a risk stratification and outcome prediction tool. Eur J Emerg Med 2015. 000000000000344.

Leave a Reply

Your email address will not be published. Required fields are marked *