a b s t r a c t

Objective: To evaluate the association between Capillary refill time measured by a medical device and sepsis among patients presenting to the Emergency Department (ED).

Methods: This prospective observational study enrolled adult and pediatric patients during ED triage when sepsis was considered a potential diagnosis by the triage nurse. Patients were enrolled at an academic medical center between December 2020 and June 2022. CRT was measured by a research assistant using an investigational med- ical device. The outcomes included sepsis and septic shock defined using sep-3 criteria, septic shock defined as IV antibiotics and a vasopressor requirement, ICU admission, and hospital mortality. Other measures included pa- tient demographics and vital signs at ED triage. We evaluated Univariate associations between CRT and sepsis outcomes.

Results: We enrolled 563 patients in the study, 48 met Sep-3 criteria, 5 met Sep-3 shock criteria, and 11 met prior septic shock criteria (IV antibiotics and vasopressors to maintain mean arterial pressure of 65). Sixteen patients were admitted to the ICU. The mean age was 49.1 years, and 51% of the cohort was female. The device measured CRT was significantly associated with the diagnosis of sepsis by sep-3 criteria (OR 1.23, 95% CI 1.06-1-43), septic shock by sep-3 criteria (OR 1.57, 95% CI 1.02-2.40), and septic shock defined as receipt of IV antibiotics and a va- sopressor requirement (OR 1.37, 95% CI 1.03-1.82). Patients with CRT >3.5 s measured by the DCR device had an odds ratio of 4.67 (95%CI 1.31-16.1) of septic shock (prior definition), and an odds ratio of 3.97 (95% CI 1.99-7.92) of ICU admission, supporting the potential for the 3.5-s cutoff of the DCR measurement.

Conclusions: CRT measured by a medical device at ED triage was associated with the diagnosis of sepsis. Objective CRT measurement using a medical device may be a relatively simple way to improve sepsis diagnosis during ED triage.

(C) 2023

1. Introduction
   1. Background

Sepsis is the most common reason for intensive care unit (ICU) ad- mission in the US, the most common cause of ICU death, and has a 20% 30-day hospital readmission rate [1,2]. Once sepsis has progressed to septic shock, the mortality rates range from 15 to 56% [2]. Timely di- agnosis of sepsis and septic shock is critical as each hour delay of

* Corresponding author at: 3181 SW Sam Jackson Pk. Rd., Portland, OR 97239, United States of America.

E-mail address: [email protected] (M. Hansen).

administration of antibiotics is associated with an 8% increase in mortal- ity for septic shock patients and an 8% increased risk of progression to shock for severe sepsis patients [3,4]. Current sepsis triage scoring tools function poorly [5,6]. Some rely on laboratory tests, limiting their utility in the triage environment [7].

Capillary refill time (CRT) evaluates microcirculatory dysfunction, a hallmark of sepsis physiology [8,9]. However, clinicians traditionally es- timate (without a stopwatch) CRT manually, and it has poor inter- observer and intra-observer agreement [10,11]. Electronic measure- ment of CRT has been shown to improve the accuracy and repeatability of CRT measurement [12]. Incorporating electronic measurement of CRT can potentially improve ED triage for sepsis patients. This study aims to determine the association between CRT and sepsis when measured via an investigational medical device during ED triage.

https://doi.org/10.1016/j.ajem.2023.05.009

0735-6757/(C) 2023

1. Methods
   1. Study design and setting

We conducted a prospective observational study in the adult and pe- diatric EDs of a single Urban academic medical center with a volume of approximately 55,000 annual patient visits of all ages. The study was conducted between December 2020 and June 2022. The local IRB ap- proved the study, and all participants or their caregivers (for minors) consented to participate.

• 1. Selection of participants

Research assistants (RAs) screened for eligible patients in the ED ap- proximately 40 h per week, typically during the hours of 12 pm-8 pm. We searched the electronic health record to identify chief complaints associated with a diagnosis of sepsis and developed a screening list. Some examples include fever, cough, rash, and indwelling device (e.g., catheter) complications. The RAs screened for these complaints and asked the nurses if they felt the patient was at any risk for sepsis. If the nurse had a Clinical concern that the patient was at risk for sepsis, the patient was approached for consent to participate. Patients were ex- cluded if they were less than six months of age or received >500 mL of IV fluids or IV antibiotics before screening. The clinical team was blinded to the CRT results from the device.

• 1. Measurements

The Digital Capillary Refill (DCR) device was used to measure the CRT for the study. The DCR is a reusable experimental device not yet commercially available. The patient’s finger is placed in a cradle and a soft strap is placed to hold it in place and block ambient light (Fig. 1). The fingertip is depressed and released by the RA, guided by a mobile application that ensures appropriate force (4-9 N) is applied and held for the correct duration (approximately 5 s). Through this process, light reflectance from a green 525 nm LED creates an optical waveform converted by a data-processing algorithm into the CRT. Each patient had 3 repeated measurements, and the mean was used for analysis. The measurement process typically takes 1-2 min. The RA also measured the CRT manually. To ensure that the applied force and duration were consistent between DCR and manual measurements, the RA used the DCR to guide the magnitude and duration of force application, minimiz- ing the effect of external factors that could affect manual CRT results. A stopwatch was started upon force release and stopped when the

Fig. 1. Photograph of the medical device used in the study.

fingertip color returned to baseline. Prior to the study, the RA received hours of one-on-one training by physicians and engineers until the re- search team agreed that the waveform quality and CRT values were con- sistent between the RA and a device inventor.

• 1. Outcomes

The study’s primary outcome was a patient meeting the criteria for sepsis. We evaluated the Sep-3 criteria for sepsis and septic shock [13] and the most recent prior septic shock criteria [14] (IV antibiotics and a vasopressor requirement), ICU admission, and hospital mortality.

• 1. Analysis

We used descriptive statistics to describe the study population. We considered the Sep-3 criteria to be a rise in sequential organ failure (SOFA) score of at least 2 in the setting of a suspected infection. Sep-3 shock was defined as a vasopressor requirement to maintain a mean ar- terial pressure (MAP) of 65 and lactate >2. Septic shock using the old definition, was considered receipt of IV antibiotics and a vasopressor re- quirement to maintain a MAP of 65. We used several definitions of sep- sis since each represents a specific patient group, definitions have changed over time, and our overall goal is to improve outcomes of septic patients, some of whom may not meet the individual regulatory definitions.

We excluded all lactate measurements obtained >6 h from ED ar- rival since these were not likely to be part of initial triage decisions. The 6-h threshold was chosen given periods of marked ED crowding and boarding, resulting in long wait times and potential delays in care. Lactate is measured in this ED using venous blood analyzed on a point-of-care machine operated by nursing staff. We next evaluated the univariate association of the CRT measured using the DCR device, the various sepsis definitions, and ICU admission using univariate logis- tic regression. We performed a similar analysis using the manually- measured CRT. This analysis was repeated for the venous lactate level. elevated lactate is often incorporated into ED sepsis alert protocols, and we felt there would be value in understanding the odds ratios for lactate compared to CRT. We did not conduct multivariable analyses due to the relatively small number of patients who met study defini- tions for septic shock and were admitted to the ICU, which was the group we were primarily interested in. This study followed STARD guidelines for diagnostic accuracy studies [15].

We explored several potential cutpoints for CRT using the LOESS procedure in SAS. LOESS uses localized regression and can be used to find potential cutpoints in data. Following this procedure, the best can- didate cutpoint was evaluated similarly to CRT when used as a continu- ous variable.

1. Results
   1. Characteristics of study subjects

We enrolled 563 patients in the study, 48 met Sep-3 criteria, 5 met Sep-3 shock criteria, and 11 met prior septic shock criteria. Sixteen pa- tients were admitted to the ICU. The mean age was 49.1 years, and 51% of the cohort was female. Table 1 displays the demographics of the study participants by sepsis diagnosis category.

• 1. Results

We noted relatively similar initial heart rates, respiratory rate, age, and sex between groups. We noted differences in the initial MAP, with expectedly lower initial MAP in the groups with shock. The time to the venous lactate result from ED triage vital signs was 60 min (SD 66 min) overall, but it was only 18-19 min in the group with shock. Overall, 49% of the patients received IV antibiotics in the ED, while

Table 1

Characteristics of study participants stratified by various methods of defining sepsis.

Variable All Sep 3 criteria Sep 3 shock Septic Shock, prior??

	N?	Mean(SD)		N?	Mean(SD)		N?	Mean(SD)		N?	Mean(SD)
Age	558	49.1(23.2)		48	56.5(20.2)		5	59.5		11	50.9(22.0)
Female Sex	563	51%		48	48%		5	60%		11	63%
white race	563	78%		48	77%		5	80%		11	82%
Black race	563	5%		48	4%		5	0%		11	9%
Latino	563	7%		48	8%		5	0%		11	0%
Other race	563	13%		48	10%		5	20%		11	9%
Diabetes	563	7.4%		48	8.3%		5	20%		11	9.1%
Chronic kidney disease	563	9.4%		48	12.5%		5	60%		11	36%
Chronic liver disease	563	2.0%		48	2.1%		5	0		11	0
Active malignancy	563	17.1%		48	2.3%		5	20%		11	9.1%
Immune compromising- medication	563	17.1%		48	2.3%		5	0		11	18.2%
Initial heart rate	550	103.7(25.7)		48	105.7(24.1)		5	95.2(26.4)		11	108.1(22.6)
initial systolic blood pressure	547	127.1(23.5)		48	120.2(25.8)			97.2(21.1)		11	99.7(18.0)
Initial MAP	547	88.9(18.1)		48	81.5(23.3)		5	65.0(11.9)		11	64.3(18.8)
Initial respiratory rate	535	20.4(6.2)		48	23.5(10.6)		5	20.2(2.5)		11	22.6(4.8)
Max temperature	545	37.5(0.93)		48	37.8(1.1)		5	37.1(0.79)		11	37.8(1.21)
Lowest systolic blood pressure	549	110.2(23.4)		48	98.6(26.0)		5	57.8(13.7)		11	65.9(16.0)
Lowest ED MAP	549	77.7(17.2)		48	68.6(19.3)		5	41.8(9.2)		11	42.5(10.9)
Venous lactate	338	1.71(1.39)		48	2.5(2.6)		5	3.2(1.3)		11	3.1(3.4)
Time from triage vitals to lactate in minutes	317	60 (66)		38	39(35)		5	19 (9.6)		10	18(12)
Total ED IVF	544	794(888)		46	1334(1082)		5	1500(1000)		11	2045(1214)
ED IV antibiotics	546	49%		48	100%		5	100%		11	100%
Discharged home	563	27.2%		48	4.1%		5	0		0	0
Admit OBS	563	15.8%		48	16.7%		5	0		11	9.1%
Admit Ward	563	48%		48	54.2%		5	0		11	9.1%
Admit ICU	563	6.6%		48	23.4%		5	100%		11	82%
Survive to discharge	545	98%		46	93%		4	50%		10	80%
Hospital LOS (hrs)	542	111(176)		48	195(193)		2	97.1(80.3)		11	169(219)
ICU LOS	548	14(77)		48	65(164)		2	71.7(73.8)		11	129(162)
Device CRT	543	3.5 (1.9)		47	4.2 (2.1)		5	5.37(1.61)		11	4.8(2.2)
Visual CRT	520	2.4 (0.89)		47	2.88(0.96)		5	2.49 (0.50)		11	3.20(1.43)

* Reflects the sample with that data point available of the total 563, 48, 5, and 11 for each category respectively.

?? Septic shock using the second international consensus definition which includes receipt of IV antibiotics and vasopressors to maintain a mean arterial pressure >65,

100% received IV antibiotics in the sep-3, sep-3 shock, and prior septic shock definitions. We noted that the DCR CRT was higher among pa- tients with shock. We did not conduct statistical significance tests for these comparisons as they were exploratory.

We next conducted univariate analyses of the association between the CRT measured with the DCR device (Table 2a), visually by an RA using a stopwatch with the patient’s finger in the DCR finger cradle, to guide the force magnitude and duration (Table 2b), and various defini- tions of sepsis among other outcomes. We similarly evaluated the asso- ciation between the initial venous lactate and the same outcomes (Table 2c). We noted the DCR measured CRT was significantly associ- ated with the diagnosis of sepsis by sep-3 criteria (OR 1.23, 95% CI 1.06-1-43), septic shock by sep-3 criteria (OR 1.57, 95% CI 1.02-2.40), and the prior definition of septic shock (OR 1.37, 95% CI 1.03-1.82). DCR measured CRT was associated with significantly increased odds of ICU admission (OR 1.30, 95% CI 1.10-1.54) and decreased odds of survival to hospital discharge (OR 0.63, 95% CI 0.43-0.85). These associations were similar to those obtained by the initial venous lactate level. Visually measured CRT, where the applied force was measured and guided by DCR, shared similar associations though it was neither associated with sep-3 shock diagnosis nor survival to hospital discharge.

In the SAS LOESS analysis, 3.5 s emerged as a potentially viable cutpoint for CRT. Patients with CRT >3.5 s measured by the DCR device had an odds ratio of 4.67 (95%CI 1.31-16.1) of septic shock (prior defi- nition), and an odds ratio of 3.97 (95% CI 1.99-7.92) of ICU admission, supporting the potential for the 3.5-s cutoff of the DCR measurement (Table 3). Sensitivities and specificities from this analysis are also in- cluded in Table 3.

Table 2 Univariate associations between CRT, lactate, and patient outcomes among those at ED tri- age suspected to have sepsis.

Table 2a: Associations between DCR CRT (continuous variable) and sepsis outcomes

Outcome Odds Ratio 95% CI

Sep 3 sepsis 1.23 1.064-1.432

Sep 3 shock 1.567 1.021-2.404

Septic shock (prior definition) 1.37 1.029-1.824

ICU admission 1.299 1.096-1.539

Acute care admission 1.076 0.984-1.177

Survival to discharge 0.634 0.473-0.849

Table 2b: Associations between CRT (continuous variable) measured visually by RA using a stopwatch with pressure and release guided by DCR device and sepsis outcomes

	Outcome	Odds Ratio	95% CI
	Sep 3 sepsis	1.602	1.204-2.133
	Sep 3 shock	1.075	0.431-2.862
	Septic shock (prior definition)	1.66	1.127-2.462
	ICU admission	1.56	1.15-2.11
	Acute care admission	1.32	1.071-1.625
	Survival to discharge	0.73	0.47-1-12
Table 2c: Associations between sepsis outcomes and venous lactate
	Outcome	Odds Ratio	95% CI
	Sep 3 sepsis	1.369	1.125-1.667
	Sep 3 shock	1.339	1.017-1.763
	Septic shock (prior definition)	1.353	1.087-1.684
	ICU admission	1.375	1.127-1.677
	Acute care admission	0.896	0.764-1.050
	Survival to discharge	0.691	0.556-0.860

Table 3

Test Characteristics of DCR CRT >3.5 s and sepsis outcomes.

Outcome	Odds Ratio	95% CI	Sensitivity	Specificity
Sep 3 sepsis	1.77	0.96-3.27	40%	73%
Sep 3 shock	*	N/A	64%	73%
Septic shock (prior definition)	4.67	1.31-16.1	100%	73%
ICU admission	3.97	1.99-7.92	58%	74%
Acute care admission	1.626	1.12-2.36	33%	77%
Survival to discharge	0.375	0.12-1.18	73%	50%

* all patients in this category had CRT >3.5 s so the model did not run.

1. Discussion
   1. Overall interpretation

This study enrolled 563 ED patients during ED triage who were at risk for sepsis. CRT measured by the DCR device was associated with sepsis defined using several criteria, ICU admission, and hospital mortal- ity. The DCR-measured CRT cutpoint of 3.5 s was also associated with several sepsis outcomes. We noted the cutpoint of 3.5 s was relatively specific for sepsis using several different diagnostic criteria and, there- fore, may be most useful to rule in the disease. CRT measured by the DCR could potentially trigger initiation of a sepsis care pathway. These test characteristics relate to the CRT value when used alone. Incorporat- ing CRT into a screening protocol that includes other variables is a prom- ising area for future research. The DCR device has the potential to improve the timeliness of care for sepsis patients in the ED.

One challenge of ED triage for potentially septic patients is identify- ing the group at a high enough risk to benefit from prompt IV antibiotics and other important treatments without too many false positives. We asked the Triage nurses to identify patients with any risk of sepsis, and 8.6% of patients from this larger group met Sepsis criteria. If a clinical de- cision rule was applied to this group and identified patients with a like- lihood of sepsis above 70%, immediately initiating a sepsis care protocol that includes IV antibiotics could improve outcomes with an acceptable rate of false positives. Approximately 0.4% of all ED visits are for sepsis, and screening all ED patients for sepsis regardless of triage nurse input would be lower yield and result in excessive false positives [16]. We also found that when the DCR device guided the finger force application, duration of hold, and release, and RAs used a stopwatch, visually mea- sured CRT was also associated with sepsis. This provides additional evi- dence of validity to the function of the DCR device.

Early recognition of sepsis is the critical first step to meeting the “1-h

bundle” goals set by the “Surviving Sepsis Campaign,” which include ad- ministering IV fluids and antibiotics [17]. Recognizing patients at high risk for sepsis at ED triage may offer the best chance to meet the 1-h bundle and maximize patient outcomes.

One unique element of this study was the focus on the ED triage en- vironment. Triage nurses are responsible for distributing the limited ED resources and determining the timing and pace of initial patient care. Ef- fective Screening tools are available for other time-sensitive emergen- cies, such as stroke and STEMI, though similar tools for sepsis remain elusive. CRT measured objectively may be one potential method to im- prove rapid sepsis screening and may be particularly useful when incor- porated into a decision rule that includes other predictive factors.

One previous study found that CRT measured electronically using a Pulse oximeter had similar sensitivity and specificity for detecting sep- sis using sep-3 criteria compared to the qSOFA score and serum lactate concentrations [18]. However, this previous study was retrospective, and over 60% of patients screened had sepsis, reflecting a high pre-test probability and not consistent with patients in a triage or real-world practice scenario. Our sensitivity and specificity were somewhat lower, though the disease prevalence in our study was nearly 8 times lower, which likely has a notable impact [19]. Further, the definition of sepsis was limited to sep-3, and other sepsis-related outcomes were not evaluated.

Future research should focus on larger patient samples and multiple institutions. With larger samples, it may be possible to develop a practical sepsis triage algorithm using objective data, including CRT, that accurately identifies patients most likely to benefit from early intervention.

• 1. Limitations

This study was conducted in an Urban academic ED using a non- consecutive sample, potentially limiting generalizability. The lactate value was obtained by a point-of-care technology that may not be avail- able in some EDs. Therefore, the 60 min from ED arrival to lactate result in our study may be significantly shorter than centers that need to send this test to the lab. We did not evaluate CRT as potentially related to other causes of shock. This study was also limited to mostly those self- identifying as White race and future studies should address the perfor- mance of the DCR among a wide range of skin tones. There are also con- flicts of interest since some study authors are co-founders and co- owners of the company developing this device. However, these individ- uals were not involved in the screening or consenting of any study par- ticipants, and the data cleaning and analyses were primarily conducted by a statistician independent from the company (HX). The study was also conducted under a conflict-of-interest plan set in place by the university.

• 1. Conclusions

In this study of ED triage patients with concern for sepsis, CRT mea- sured by the DCR device was associated with confirmed sepsis, septic shock, ICU admission, and hospital mortality. This non-invasive and in- tuitive test appears to provide valuable physiologic data to supplement early sepsis risk stratification.

Funding

The Oregon Health & Science University Biomedical Innovations Pro- gram provided funding for the study. The funder had no role in the de- sign and conduct of the study; collection, management, analysis or interpretation of the data; preparation review, or approval of the man- uscript; or decision to submit for publication.

Author contributions

-MH, DS, RS, SB, SF made substantial contributions to the design of the work.

-MH, DS, JG, TR, RS, and SB made substantial contributions to the ac- quisition of the data.

-MH and HX made substantial contributions to the analysis.

-All authors made substantial contributions to the interpretation of the data, drafting or revising for critically important intellectual content, and provide final approval of the manuscript.

CRediT authorship contribution statement

Matthew Hansen: Writing – original draft, Supervision, Project ad- ministration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Jordan Gillespie: Writing – review & editing, Resources, Project administration, Investigation, Data curation. Tyne Riddick: Writing – review & editing, Project administration, Meth- odology, Investigation, Data curation. Ravi Samatham: Writing – review & editing, Visualization, Validation, Software, Project adminis- tration, Methodology, Investigation, Formal analysis, Data curation. Steven Baker: Writing – review & editing, Supervision, Resources, Methodology, Investigation, Formal analysis, Conceptualization. Scott Filer: Writing – review & editing, Resources, Project administration, Funding acquisition, Conceptualization. Haichang Xin: Writing –

review & editing, Visualization, Methodology, Formal analysis, Data curation. David Sheridan: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Funding acquisition, Conceptualization.

Declaration of Competing Interest

MH, DS, SF, and SB are equity holders in Promedix Inc. and have con- flicts of interest. Promedix Inc. has an exclusive license to or owns the intellectual property relevant to this research and is currently develop- ing this technology. This research was conducted under a conflict-of- interest management plan at Oregon Health & Science University.

References

1. Sun A, Netzer G, Small DS, et al. Association between index hospitalization and hos- pital readmission in Sepsis survivors*. Crit Care Med. 2016;44(3):478-87. https:// doi.org/10.1097/CCM.0000000000001464.
2. Bauer M, Gerlach H, Vogelmann T, Preissing F, Stiefel J, Adam D. Mortality in sepsis and septic shock in Europe, North America and Australia between 2009 and 2019- results from a systematic review and meta-analysis. Crit Care. 2020;24(1):239. https://doi.org/10.1186/s13054-020-02950-2.
3. Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of ef- fective antimicrobial therapy is the critical determinant of survival in human septic shock*. Crit Care Med. 2006;34(6):1589. https://doi.org/10.1097/01.CCM. 0000217961.75225.E9.
4. Whiles BB, Deis AS, Simpson SQ. Increased time to initial antimicrobial administra- tion is associated with progression to septic shock in severe sepsis patients. Crit Care Med. 2017;45(4):623-9. https://doi.org/10.1097/CCM.0000000000002262.
5. Singer AJ, Taylor M, Domingo A, et al. Diagnostic characteristics of a clinical screen- ing tool in combination with measuring bedside lactate level in emergency depart- ment patients with suspected Sepsis. Acad Emerg Med. 2014;21(8):853-7. https:// doi.org/10.1111/acem.12444.
6. Filbin MR, Thorsen JE, Lynch J, et al. Challenges and opportunities for emergency de- partment Sepsis screening at triage. Sci Rep. 2018;8(1):11059. https://doi.org/10. 1038/s41598-018-29427-1.
7. Pierrakos C, Velissaris D, Bisdorff M, Marshall JC, Vincent JL. Biomarkers of sepsis: time for a reappraisal. Crit Care. 2020;24(1):287. https://doi.org/10.1186/s13054- 020-02993-5.
8. Charlton M, Sims M, Coats T, Thompson JP. The microcirculation and its measure- ment in sepsis. J Intensive Care Soc. 2017;18(3):221-7. https://doi.org/10.1177/ 1751143716678638.
9. Hernandez G, Castro R, Bakker J. Capillary refill time: the missing link between macrocirculation and microcirculation in septic shock? J Thorac Dis. 2020;12(3): 1127-9. https://doi.org/10.21037/jtd.2019.12.102.
10. Sheridan DC, Baker SD, Kayser SA, Jones D, Hansen ML. Variability of capillary refill time among physician measurements. J Emerg Med. 2017;53(5):e51-7. https:// doi.org/10.1016/j.jemermed.2017.06.035.
11. Brabrand M, Hosbond S, Folkestad L. Capillary refill time: a study of interobserver re- liability among nurses and nurse assistants. Eur J Emerg Med. 2011;18(1):46-9. https://doi.org/10.1097/MEJ.0b013e32833b4fba.
12. Gillespie J, Hansen M, Samatham R, Baker SD, Filer S, Sheridan DC. Capillary refill technology to enhance the accuracy of Peripheral perfusion evaluation in Sepsis. J Intensive Care Med. 2022. https://doi.org/10.1177/08850666221087685. Published online March 21. 08850666221087685.
13. Singer M, Deutschman CS, Seymour CW, et al. The third international consensus def- initions for sepsis and septic shock (Sepsis-3). JAMA. 2016;315(8):801-10. https:// doi.org/10.1001/jama.2016.0287.
14. Levy MM, Fink MP, Marshall JC, et al. 2001 SCCM/ESICM/ACCP/ATS/SIS international sepsis definitions conference. Intensive Care Med. 2003;29(4):530-8. https://doi. org/10.1007/s00134-003-1662-x.
15. Korevaar DA, Cohen JF, Reitsma JB, et al. Updating standards for reporting diagnostic accuracy: the development of STARD 2015. Res Integr Peer Rev. 2016;1(1):7. https://doi.org/10.1186/s41073-016-0014-7.
16. Strehlow MC, Emond SD, Shapiro NI, Pelletier AJ, Camargo CA. National Study of emergency department visits for sepsis, 1992 to 2001. Ann Emerg Med. 2006;48 (3):326-331.e3. https://doi.org/10.1016/j.annemergmed.2006.05.003.
17. Levy MM, Evans LE, Rhodes A. The surviving Sepsis campaign bundle: 2018 up- date. Intensive Care Med. 2018;44(6):925-8. https://doi.org/10.1007/s00134- 018-5085-0.
18. Yasufumi O, Morimura N, Shirasawa A, et al. Quantitative capillary refill time pre- dicts sepsis in patients with suspected infection in the emergency department: an observational study. J Intensive Care. 2019;7(1):29. https://doi.org/10.1186/ s40560-019-0382-4.
19. Leeflang MMG, Rutjes AWS, Reitsma JB, Hooft L, Bossuyt PMM. Variation of a test’s sensitivity and specificity with disease prevalence. CMAJ. 2013;185(11):E537-44. https://doi.org/10.1503/cmaj.121286.