a b s t r a c t

Background: Sepsis is a leading cause of death in hospitals requiring prompt recognition and treatment. The Sepsis Bundle is the cornerstone of sepsis treatment. Studies have evaluated the impact of a sepsis huddle on sepsis bun- dle compliance but not in sepsis identification.

Objective: Measure the effect of a multidisciplinary sepsis bedside huddle in the Emergency Department (ED) on

sepsis identification and sepsis bundle compliance.

Methods: Retrospective, single-center, cohort study. Pre-huddle patients were identified via Best Practice Advi- sory (BPA) alert on the electronic medical record from 11/01/2019-3/31/2020. The post-huddle group were patients for whom a sepsis huddle was activated from 11/01/2020-3/31/2021.

Results: 116 patients met inclusion criteria and 15 were determined to not have sepsis for a total of 21 pre-huddle and 80 post-huddle patients. Comparing pre-post results, sepsis huddle increased code sepsis activation (10% vs 91%, p < 0.001); sepsis bundle compliance (24% vs 80%, p < 0.001); antibiotics within one hour (33% vs 90%, p < 0.001); culture within one hour (67% vs 95%, p < 0.001), order entry <30 min. (29% vs 86%, p < 0.001); and median order entry time (48 vs. 3 min, p < 0.001). Post-huddle, 80% of order entries were <= 20 min. Logistic regression predicting sepsis code found huddle to be the first predictor, (p < 0.0000005). Hour-1 bundle compli- ance was predicted by physician/physician assistant order <=30 min (R² = 0.36, p < 0.0000005).

Conclusion: Sepsis bedside huddle in the ED improves identification and sepsis bundle compliance. Results sug- gest increased order entry speed caused bundle improvement.

(C) 2022

1. Introduction

Sepsis is a leading cause of hospital death and requires prompt rec- ognition and treatment. Every year >1.7 million Americans develop sepsis with almost 270,000 deaths from it [1]. Sepsis can cause life- threatening organ dysfunction due to a dysregulated host response to infection [2]. The sepsis bundle is the cornerstone of sepsis treatment and quality measures. In 2018 the Surviving Sepsis Campaign (SSC) up- date combined the 3-h and 6-h bundles into a single hour-1 bundle [3]. The hour-1 bundle for Initial resuscitation in sepsis and septic shock includes measuring lactate level, obtaining blood cultures before giving antibiotics, administering antibiotics, begin administration of 30 mL/kg crystalloid fluids for hypotension or lactate >=4 mmol/L, and application of vasopressors if hypotensive maintain mean arterial

* Corresponding author.

E-mail addresses: Kathy.Currie@providence.org (K.E. Currie), Hend.Barry@swedish.org (H. Barry), James.Scanlan@swedish.org (J.M. Scanlan), Eric.Harvey@swedish.org (E.M. Harvey).

pressure (MAP) >= 65 mmHg. Increased sepsis bundle compliance has been associated with lower hospital mortality and decreased length of stay [2]. For every hour antibiotics are delayed from hypotension recog- nition there is a 7.6% survival decrease [4]. Although recent studies have shown improvement in sepsis bundle compliance with sepsis huddle use, none have evaluated the impact on code sepsis activation and therefore improved sepsis identification [5-9].

The purpose of this study was to measure the impact of conducting a

multidisciplinary sepsis bedside huddle in the Emergency Department (ED) on improvement in improved sepsis identification and sepsis bundle compliance.

1. Methods
   1. Study design and population

A retrospective cohort study was performed on patients aged >=18 seen in the Swedish Medical Center (SMC) – Ballard Campus Emergency Department for whom the sepsis huddle was activated. SMC Ballard

https://doi.org/10.1016/j.ajem.2022.12.006

0735-6757/(C) 2022

Campus ED is one of seven Emergency Departments in the system with a volume of approximately 22,000 visits per year and a population that includes community, cardiac, neurological, and trauma patients. The sepsis huddle was implemented in October of 2020 as a tool that could be activated at any time expedite recognition and treatment of sepsis patients. Most sepsis huddles were activated upon ED triage based on chief complaint and initial vital signs. However, the sepsis huddle could be activated by any member of the Multidisciplinary team at any time if a patient met two or more Systemic Inflammatory Response Syndrome criteria plus a possible bacterial source of in- fection (Supplemental Fig. 1). Upon sepsis huddle activation a page is sent via Vocera to all ED staff. The multidisciplinary team members (physician or physician assistant, primary nurse, ED pharmacist, and ED charge nurse) then meet at the patient’s bedside for evaluation and activation of code sepsis if indicated (Supplemental Fig. 1). ED phar- macists are available Monday through Friday from 1000 to 1830 and can assist with choice of therapy, order entry, order verification, and antibi- otic delivery. If the ED pharmacist was not available at the time of the sepsis huddle other members of the multidisciplinary team would pro- ceed with assessment. If a code sepsis was activated when ED pharma- cist was not available, then the inpatient hospital pharmacist would receive a page and prioritize order verification, be available for any questions and expedite the delivery of antibiotics. The use of code sepsis was already in place at SMC Ballard Campus ED but had not been opti- mized. Code sepsis was only activated by physicians or physician assis- tants, there was no formal system to empower other members of the multidisciplinary team to bring attention to possibly septic patients, and bundle elements were completed in separate steps often missing the time mark for goal bundle completion. Identification of sepsis pa- tients was delayed due to the nature of the emergency department workflow, in many instances prior to sepsis huddle implementation, pa- tients were waiting for a room in the emergency department waiting room or in an ambulance prior to being seen by the physician or physi- cian assistants. Also, there was no formal and streamlined system set up to notify pharmacy or lab personnel of patients for whom a code sepsis had been activated to ensure prioritization. The duration of the study was limited by the length of residency of the primary author. The pre- huddle implementation period was 11/01/2019-3/31/2020 and pa- tients were identified via the organizations Best Practice Advisory (BPA) which alerted providers via the electronic health records (EHR). Pre-huddle patients were excluded from the data analysis if they were determined to not have sepsis by the ED physician or physi- cian assistant based on the diagnosis documented on progress note via manual chart review. Post-huddle implementation period was defined from 11/01/2020-3/31/2021 and patients included those for whom the sepsis huddle was activated. Post-huddle patients were excluded from the analysis if a sepsis huddle was activated but during the bedside assessment the multidisciplinary team agreed the patient was not septic.

This study was determined to be exempt by our Institutional Review

Board with a waiver of informed consent.

• 1. Data abstraction and variables

Data was collected by the primary author who was not blinded to the study question. List for the pre-huddle group of patients was ob- tained from EHR BPA reports. time zero for the pre-huddle group was defined as the time when patient met SIRS criteria in the EHR and as time of sepsis huddle for post-huddle group. List for post-huddle pa- tients was obtained from sepsis huddle checklist completed by nurse. Our sepsis huddle checklist included: initial time of sepsis huddle acti- vation, SIRS criteria causing activation, the suspected source of infection, time and value of initial and repeat lactate, time of order and adminis- tration of antibiotics, and the time of order and administration of fluids (Supplemental Fig. 2).Type of fluids administered (Supplemental Table 1), and time to antibiotics order entry and pharmacist order

verification was extracted from EHR, as was any data point not recorded on the huddle checklist.

• 1. Outcomes

The primary outcome was the difference in the frequency of code sepsis activation pre-sepsis huddle compared to post-huddle. Second- ary outcomes included: completion of hour-1 sepsis bundle from time zero, blood culture collection, antibiotics start, lactate collection, intra- venous fluids started (IVF), and initiation of vasopressors if required (all within one hour). Completion of hour-1 sepsis bundle was defined as achieved if measurement of lactate level, collection of blood cultures, start of antibiotic administration, start of 30 mL/kg crystalloid fluids for hypotension or lactate >=4 mmol/L, and application of vasopressors only if persistent hypotension after IVF administration were all performed within 60 min of time zero. To determine what step of ordering antibi- otics was causing delays in start additional secondary outcomes were time to antibiotic order entry, time from antibiotics order entry to phar- macist verification, and time from pharmacist verification of order to start of antibiotic administration.

• 1. Statistical analysis

Baseline and demographic characteristics were summarized using descriptive statistics (median and interquartile range [IQR] for continu- ous variable, and frequency and percentages for categorical variables). Median and IQR was used for continuous data that was not normally distributed. Mann-Whitney U test, Chi-Square test, and Fisher’s Exact test were used to determine if there were relationships between sepsis huddle and primary and secondary outcomes. Logistic regression was performed with code sepsis and the hour-1 bundle compliance as de- pendent outcomes. Stepwise forward conditional entry was used (p = 0.05 was entry criteria). Potential predictor variables for both equations included huddle, age and sex, physician order time (<10 min, <20 min,

<25 min, and <30 min); time to pharmacist order verification, time be- tween order verification and antibiotic start, sepsis indicators (tachycar- dia, high or Low Temperatures, lactates >2, WBC, systolic blood pressure, repeat lactate) general infection measures, (abdominal, pneumonia, py- elonephritis, nephritis, urinary tract infection, skin and soft tissue infec- tion, C. difficile, meningitis, febrile neutropenia, osteomyelitis and joint infection). All regression equations included the complete data set of pre-and post-huddle patients and were tested with Hosmer and Lemeshow goodness of fit statistics. Because of the retrospective nature of the study, a sample size Power calculation was not conducted. Addi- tionally, the time period of the study was limited by the length of resi- dency training of the primary author. Statistics were performed with IBM SPSS 22, Armonk, New York.

1. Results
   1. Study patients and baseline demographics

A total of 116 patients met inclusion criteria (30 pre-huddle patients and 86 post-huddle patients). Of the 30 patients in the pre-huddle group, 9 were excluded, and of the 86 patients in the post-huddle group, 6 were excluded (Fig. 1). Baseline and demo- graphic characteristics are presented in Table 1 for patients included in the analyses. The most common possible source of infection was pneumonia in the pre-huddle group and unknown source in the post-huddle group (Table 1). Of note, there was one patient in the post-huddle group with known COVID-19 treated for superimposed pneumonia, and there was another patient that was tested for COVID-19 with results pending (later positive) and was also treated for pneumonia.

Fig. 1. Patient selection and analysis.

Table 1

Baseline and demographic characteristics.

Characteristic Pre-huddle

(n = 21)

Post-huddle (n = 80)

• 1. Outcomes
     1. Primary outcomes

Between the two groups, more code sepsis were activated in the post-huddle group compared to the pre-huddle group (91.3% vs 9.5%; p < 0.001).

n (%) or median [IQR] p-value

Age (years) 68 [54-78] 55 [41.8-69] 0.037

Female gender 11 (52.4) 30 (37.5) 0.217

Temperature (>38.0 ?C or <36.0 ?C) 10 (47.6) 29 (36.3) 0.34

HR (>90 bpm) 18 (85.7) 73 (91.3) 0.43

RR (>20 bpm) 14 (66.7) 40 (50) 0.17

WBC (>12 x 10³ or <4 x 10³ cells) 15 (71.4) 51 (63.8) 0.51

SBP (<90 mmHg) 0 6 (7.5) 0.34

Lactate (>=2 mmol/L) 16 (76.2) 48 (60) 0.17

Septic shock 1 (4.8) 5 (6.3) 0.79

Vasopressors required 1 (4.8) 4 (5) 1.0

Possible source of infection: 0.199

Abdominal 0 9 (11.2)

Respiratory 8 (38.1) 13 (16.3)

Genitourinary 3 (14.3) 15 (18.7)

skin and soft tissue infection 5 (23.8) 14 (17.5)

Other 1 (4.8) 8 (10)

Unknown source 4 (19) 21 (26.3)

Abbreviations: IQR = interquartile range, HR = heart rate, RR = respiratory rate, WBC = white blood count, SBP = systolic blood pressure. Abdominal infection includes appendi- citis, peritonitis, diverticulitis, infectious colitis and Clostridioides difficile. Genitourinary in- cludes pyelonephritis and cystitis. Other includes meningitis, febrile neutropenia, and osteomyelitis and joint infection.

• • 1. Secondary outcomes

Results for secondary outcomes are reported in Table 2. Hour-1 bun- dle completion was achieved in more patients post-huddle than in the pre-huddle group (80% vs 23.8%; p < 0.001). Blood culture collection within one hour and start of antibiotic administration within one hour were also increased post-huddle relative to pre-huddle (95% vs 66.7%; p = 0.001 and 90% vs 33.3%; p < 0.001 respectively). Although collec- tion of lactate and start of IVF within one hour was numerically higher in the post-huddle group this was not statistically significant (97.5% vs 90.5%; p = 0.14 and 87.5% vs 81%; p = 0.44 respectively). Vasopressor requirement was uncommon in both groups (4.8% in pre-huddle group and 5% in post-huddle group), but when required all patients were started within one hour in both groups. Antibiotic order entry within 30 min was significantly improved in the post-huddle group compared to the pre-huddle group (86.3% vs 28.6%; p < 0.001). In the post-huddle group 69% of the orders were entered within 10 min or less while this only occurred in 19% of pre-huddle. Time from antibiotic order entry to pharmacist order verification and time from antibiotic

Table 2

Primary and secondary outcomes.

Outcome Pre-huddle (n = 21) Post-huddle (n = 80) p-value n (%) or median [IQR]

Code sepsis activation 2 (9.5) 73 (91.3) <0.001

Sepsis bundle completed <=1 h 5 (23.8) 64 (80) <0.001

Blood culture collected <=1 h 14 (66.7) 76 (95) 0.001

Antibiotics started <=1 h 7 (33.3) 72 (90) <0.001

Lactate collected <=1 h 19 (90.5) 78 (97.5) 0.14

IVF started <=1 h 17 (81) 70 (87.5) 0.44

When vasopressors required, started <=1 h 1 (100) 4 (100) N/A

Antibiotic order entry <=30 min 6 (28.6) 69 (86.3) <0.001

Time to antibiotic order entry (minutes) 48 [20-84] 3 [0-14.3] <0.001

Time from antibiotic order entry to pharmacist order verification (minutes) 1 [1-2] 2 [1-3] 0.17

Time from antibiotic order verification by pharmacist to administration (minutes) 19 [11-33] 16 [7.6-22.3] 0.16

Abbreviations: IQR = interquartile range, IVF = intravenous fluids.

Table 3

Logistic regression outcomes: Final eqs. (N = 101).

Dependent variable	Predictors	Order entry1	R2 at entry2	B	Wald	df	p-value3
Sepsis code	Huddle (pre vs. post)	1	0.36	5.23	27.0	1	<0.0000005
	Initial lactate (<2/>=2)	2	0.1	-1.55	3.4	1	0.067
Hour-1 bundle	MD order <=30 min	1	0.36	-3.60	23.6	1	<0.0000005
	Initial lactate (<2/>=2)	2	0.1	1.90	6.9	1	0.009
	Administration time	3	0.06	-0.04	5.8	1	0.016

1 = Variable order entry is sequential (1 = first, 2 = second, 3 = third).

2 = Nagelkerke R² at point of variable equation entry.

3 = p-value of individual in final equation, with all variables entered.

order verification to administration did not differ between groups (p = 0.17 and p = 0.16 respectively).

Logistic regression predicting sepsis code found huddle to be the first predictor, (p < 0.0000005, Table 3) and initial lactate measure to be a borderline second predictor (p = 0.067). Hour-1 bundle compli- ance was predicted by physician/physician assistant order <30 min (R² = 0.36, p < 0.0000005) followed by initial lactate (R² = 0.10, p = 0.009) and administration time (R² = 0.06, p = 0.016, see Table 3). It is notable that zero-order examination of bundle predictors would find huddle to be highly predictive of hour-1 bundle compliance, but it is less so than physician order entry <30 min (huddle Chi-square = 24.3, <30 min entry Chi-square = 33.1), and therefor does not enter the equation. Hosmer and Lemeshow tests found no significant good- ness of fit violations in either regression. Post hoc analysis of the hour- 1 bundle results indicated that if the physician order entry was <30 min, then times between order entry and antibiotic administration of

<=10 min, >10 and <=20 min, and >20 min all resulted in >50% bundle compliance. However, initial order times were between 30 and 60 min, only <10 min administration times achieved >50% bundle compli- ance. If the physician order entry was 30-60 min and administration time was >10 and <=20 min, bundle compliance was 20%. If administra- tion time was >20 min bundle compliance was 0%.

Focusing on the physician order entry-bundle compliance relation- ship independent of administration times, there are apparent limits for time of order entry in both the pre and post huddle cohorts. If phy- sician orders were entered in 30 min or less, both the pre and post hud- dle groups achieved majority bundle compliance (57% and 86%, respectively). However, if physician entries took >35 min, hour-1 bun- dle compliance was very poor for both pre and post huddle groups (7% and 14%, respectively). It should also be noted that the huddle imple- mentation greatly reduced extreme physician entry times (43% pre-huddle times >1 h, 5% post-huddle times >1 h).

1. Discussion

In this retrospective single center cohort study, implementation of a multidisciplinary sepsis bedside huddle in the ED improved identifica- tion of sepsis patients based on the increase in code sepsis activation. The sepsis huddle significantly improved administration of antibiotics and collection of blood culture within one hour. Improvements in com- pletion of individual sepsis bundle elements led to significantly im- proved completion of sepsis bundle within one hour. Our results showing use of a sepsis huddle improves sepsis bundle compliance is consistent with previous studies [5-9]. Despite this improvement, 20% of patients in the post-huddle group did not have their sepsis bundle completed within one hour. Initiation of IVF within one hour remains below 90% in the post-huddle group, a target goal for future improve- ment. The implementation of sepsis huddle significantly improved phy- sicians and physician assistants entering in antibiotic orders in <=30 min. Our results suggest that the delay in starting antibiotics within one hour in the pre-huddle group typically comes from delayed antibiotic order entry. Specifically, order entry within 10 min or less allowed for bundle completion within one hour even if the time to start antibiotics was

delayed. Additionally, time from order entry to order verification and time from order verification to administration did not significantly differ pre-post huddle implementation.

Our results strongly suggest that the huddle results in greater sepsis code activation. Additionally, our results show the sepsis huddle facili- tates prompt order entry, which in turn makes hour-1 bundle compli- ance is very achievable. Post-huddle 69% of physician orders were in 10 min or less, and 80% were in 20 min or less. These predominantly rapid order entries demonstrate that timely orders are very possible with a huddle in place. We believe that if a huddle is implemented order entry <=10 min is very possible and should be the target goal. If such orders are delayed, compliance becomes difficult or impossible, and patient lives may be put at risk given that bundle compliance has been associated with lower hospital mortality [2].

Our finding that initial lactate levels are associated with both sepsis code activation and hour-1 bundle compliance makes intuitive sense. Initial lactate levels are likely to be viewed as a crude index of sepsis se- verity. It is very plausible that the ED staff (physicians and pharmacists) will be more likely to achieve the hour-1 bundle if they think that the sepsis case is more severe or progressing more rapidly.

Our study took place prior to the most recent update of the SSC guideline update in 2021. The 2021 SSC guidelines recommend a rapid assessment of the likelihood of infection. If patients have possible septic shock or high likelihood of sepsis administration of antibiotics should be started within 1 h, otherwise for patients with possible sepsis but with- out shock the timeframe for antibiotics start is extended to 3 hours [10]. In our study by bringing attention to possibly septic patients earlier with the use of the sepsis huddle in the ED the multidisciplinary team has more time to determine whether symptoms are likely due to a viral or bacterial infection. Antibiotics were only initiated when the multidisci- plinary team determined it was highly probable that the patient was septic due to a bacterial infection therefore our practices were in agree- ment with the 2021 SSC guideline updates.

Other studies have used huddles or similar multidisciplinary inter- ventions to improve sepsis care. Flynn et al. reported the use of a multi- disciplinary team bedside assessment to improve antibiotic start within one hour of sepsis onset [5]. Sonis et al. reported the use of a bedside sepsis huddle to improve time to antibiotic administration [7]. This is the first study to utilize a sepsis huddle as a tool for improved identifica- tion of sepsis patients, which has been utilized to improve sepsis bundle compliance. However, this study had some limitations. First, this was a single center observational study with a small sample size so generaliz- ability may be limited. Second, our study used SIRS to trigger the activa- tion of the huddle which can be positive in patients without organ dysfunction. Future studies should use the sequential (sepsis-related) organ failure assessment (SOFA) score for diagnosis of sepsis [11]. Addi- tionally, we did not formally assess physicians’ possible perceptions of additional burden as a result of sepsis huddle implementation. How- ever, some of the members of the multidisciplinary team that partici- pated in the huddles (physicians and pharmacists) stated their impression was that overall ED workflow was not substantially increased. These limitations may be addressed in future larger, random- ized studies.

• 1. Conclusion

In conclusion, sepsis carries a high risk for morbidity and mortality requiring prompt recognition and treatment. The implementation of a multidisciplinary bedside sepsis huddle in the emergency department improves sepsis identification as evidenced by the increase in code sep- sis activation, and sepsis bundle compliance. The results are very prom- ising, and we believe sepsis huddles could be effectively implemented in most Emergency Departments.

CRediT authorship contribution statement Kathy E. Currie: Writing – review & editing, Writing – original draft,

Resources, Project administration, Methodology, Investigation, Formal

analysis, Data curation, Conceptualization. Hend Barry: Writing – review & editing, Writing – original draft, Supervision, Methodology, Conceptualization. James Scanlan: Writing – review & editing, Writing – original draft, Visualization, Validation, Formal analysis. Eric M. Harvey: Writing – review & editing, Writing – original draft, Supervi- sion, Resources.

Declaration of Competing Interest

This research did not receive any specific grant from funding agen- cies in the public, commercial, or not-for-profit sectors, and the authors of this manuscript declare they have no financial disclosures or conflicts of interest.

Acknowledgements

We would like to acknowledge and thank Nancy Driesner, PharmD, Marc Burdic, DO, Meghan Schoot, MD, Wendy Connors, RN, Andrew Dauner, PharmD, Kristine Eng, RPh, Barbara Brenneman, PharmD, and the Swedish Medical Center – Ballard Campus Emergency Department Staff for all their support in helping to complete this study and providing exceptional care to patients every day.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.ajem.2022.12.006.

References

1. Clinical Information [Internet]. Centers for Disease Control and Prevention. Centers for Disease Control and Prevention; 2020. [cited 2020Oct9]. Available from: https://www.cdc.gov/sepsis/clinicaltools/index.html?CDC_AA_refVal=https%3A%2F

%2Fwww.cdc.gov%2Fsepsis%2Fdatareports%2Findex.html.

1. Rhodes A, Evans LE, Alhazzani W, Levy MM, Antonelli M, Ferrer R, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock: 2016. Intensive Care Med. 2017 Jan 18;43(3):304-77.
2. Levy MM, Evans LE, Rhodes A. The surviving sepsis campaign bundle. Crit Care Med. 2018 Jun;46(6):997-1000.
3. Kumar A, Roberts D, Wood KE, Light B, Parrillo JE, Sharma S, et al. Duration of hypo- tension before initiation of effective antimicrobial therapy is the critical determinant of survival in human septic shock. Crit Care Med. 2006 Jun;34(6):1589-96.
4. Flynn JD, Mcconeghy KW, Flannery AH, Nestor M, Branson P, Hatton KW. Utilization of pharmacist responders as a component of a multidisciplinary sepsis bundle. Ann Pharmacother. 2014 Jun 5;48(9):1145-51.
5. Laine ME, Flynn JD, Flannery AH. Impact of pharmacist intervention on selection and timing of appropriate antimicrobial therapy in septic shock. J Pharm Pract. 2017 May 15;31(1):46-51.
6. Sonis JD, Benzer TI, Black L, Filbin MR, Hayes BD, Hibbert KA, et al. Utilization of a multidisciplinary emergency department sepsis huddle to reduce time to antibiotics and improve SEP-1 compliance. Am J Emerg Med. 2020 Nov;38(11):2400-4.
7. Walsh D, Gekle R, Bramante R, Decena E, Raio C, Levy D. Emergency department sep- sis huddles: achieving excellence for sepsis benchmarks in New York State. Am J Emerg Med. 2020 Feb;38(2):222-4.
8. Yarbrough N, Bloxam M, Priano J, Lynch PL, Hunt LN, Elfman J. Pharmacist impact on sepsis bundle compliance through participation on an emergency department sepsis alert team. Am J Emerg Med. 2019 Apr;37(4):762-3.
9. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Crit Care Med. 2021 Nov 14;49(11) e1063-e1143.
10. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. The third international consensus definitionsfor sepsis and septic shock (sepsis-3). JAMA. 2016 Feb 23;315(8):801-10.