Critical Care Endotypes

CRITICAL CARE ENDOTYPES:
Collecting and analyzing large datasets to optimize
diagnosis in the ICU
David Maslove, MD, MS, FRCPC
Department of Critical Care Medicine & School of Computing
Director, CONDUIT Lab
Queen’s University, Kingston, Ontario

1
“THE PROBLEM”
AGENDA
4
Physiologic distinctions
5
Lessons learned
6
A Path to Precision
2
DATA = DISTINCTIONS
3
Genomic distinctions

$Billions
0
22.5
45
67.5
90
ICU care NIH
5 million ICU stays
ICU
Death, disability
Health
ICU COSTS (US)

The new england journal of medicine
erraces(mean,18.8percent[95percentconfidence
interval, 17.1 to 20.5 percent]). Mortality rates did
not differ significantly according to sex (men, 22.0
percent; women, 21.8 percent).
The proportion of patients with sepsis who had
any organ failure, a marker of the severity of illness,
increased over time, from 19.1 percent in the first
11yearsto30.2percentinlateryears.Organfailure
occurredin33.6percentofpatientsduringthemost
recent subperiod, resulting in the identification of
184,060 cases of severe sepsis in 1995 and 256,033
in 2000. Organ failure had a cumulative effect on
mortality: approximately 15 percent of patients
without organ failure died, whereas 70 percent of
patients with three or more failing organs (classi-
fied as having severe sepsis and septic shock) died.
The additive effect of organ failure on mortality was
consistentovertime,withimprovementsinsurvival
being most evident among patients with fewer than
three failing organs. The organs that failed most
frequently in patients with sepsis were the lungs (in
18 percent of patients) and the kidneys (in 15 per-
cent of patients); less frequent were cardiovascular
failure (7 percent), hematologic failure (6 percent),
metabolic failure (4 percent), and neurologic fail-
ure (2 percent).
Figure 1. Population-Adjusted Incidence of Sepsis, According to Sex, 1979–2000.
Points represent the annual incidence rate, and I bars the standard error.
Population-AdjustedIncidenceofSepsis
(no./100,000)
300
200
100
0
1979 1981 1983 1997 1999 200119951985 1987 1989 1991 1993
Women
Men
Sepsis(no./100,000)
400
500
Black
White
Other
resourcesforepidemiologicinvestigationsinwhich
the prospective identification of patients is not fea-
sible.8,10 Using ICD-9-CM codes, Angus et al. cre-
ated a composite profile of sepsis from the 1995
hospital discharge records for seven states.2 They
estimated that there were 751,000 cases of severe
sepsisinthatyear,accountingfor2.1to4.3percent
ofhospitalizationsand11percentofalladmissions
to the ICU. These estimates may overstate the in-
cidence of severe sepsis by as much as a factor of
two to four,26 given that the estimated number of
deathsexceedsthecombinednumbersofdeathsre-
portedinassociationwithnosocomialbloodstream
infections27 and septic shock.28
The population-adjusted incidence of sepsis in
the United States has increased significantly over
the past two decades. The relative frequency of spe-
cific causative organisms has shifted over time, as
indicatedbythepublishedliterature,withtheemer-
gence of fungal pathogens29 and the recent pre-
eminence of gram-positive organisms.20,30 The
occurrence of organ failure increased over time
and was an additive contributor to mortality that
remained consistent among patients of different
races and sexes. The decline in mortality is nota-
ble, given the expected increases associated with
increasing age and the increasing severity of ill-
ness, but it is supported by previous analysis of cu-
mulative data from clinical trials.31 Such changes
are most likely attributabletononspecificimprove-
ments in intensive care,32,33 but diagnostic criteria
andcodingpracticesmayinfluencechangesaswell.
The increasing rate of discharge to nonacute care
medical facilities, in combination with the increas-
ing incidence of sepsis and the decrease in over-
all mortality among patients with sepsis, suggests
the fact that the greatest increase in incidence oc-
curred among women, men are consistently more
Figure 3. Numbers of Cases of Sepsis in the United States, According to
the Causative Organism, 1979–2000.
Points represent the number of cases for the given year, and I bars the stand-
ard error.
5,000
0
1979 1981 1983 1997 1999 200119951985 1987 1989 1991 1993
Figure 4. Overall In-Hospital Mortality Rate among Patients Hospitalized
for Sepsis, 1979–2000.
Mortality averaged 27.8 percent during the first six years of the study and 17.9
percent during the last six years. The I bars represent the standard error.
ProportionofPatientswithSepsisWhoDied
0.30
0.20
0.40
0.10
0.00
1979 1981 1983 1997 1999 200119951985 1987 1989 1991 1993
n engl j med 348;16 www.nejm.org april 17, 2003
confidenceinterval,1.24to1.32])andamongnonwhitepersonsthanamongwhiteper-
sons(meanannualrelativerisk,1.90[95percentconfidenceinterval,1.81to2.00]).Be-
tween 1979 and 2000, there was an annualized increase in the incidence of sepsis of
8.7 percent,fromabout164,000cases(82.7per100,000population)tonearly660,000
cases (240.4 per 100,000 population). The rate of sepsis due to fungal organisms in-
creased by 207percent,withgram-positivebacteriabecomingthepredominantpatho-
gens after 1987. The total in-hospital mortality rate fell from 27.8 percent during the
period from 1979 through 1984 to 17.9 percent during the period from 1995 through
2000,yetthetotalnumberofdeathscontinuedtoincrease.Mortalitywashighestamong
blackmen.Organfailurecontributedcumulativelytomortality,withtemporalimprove-
ments in survival among patients with fewer than three failing organs. The average
length of the hospital stay decreased, and the rate of discharge to nonacute care medi-
cal facilities increased.
conclusions
Theincidenceofsepsisandthenumberofsepsis-relateddeathsareincreasing,although
the overall mortality rate among patients with sepsis is declining. There are also dispar-
itiesamongracesandbetweenmenandwomenintheincidenceofsepsis.Gram-positive
bacteria and fungal organisms are increasingly common causes of sepsis.
ICU MORTALITY (US)

100,000 ICU survivors per year
“Mild cognitive impairment” (40%)
“Moderate TBI” (25%)
BEYOND SURVIVAL
(ONTARIO)

The new england
journal of medicine
established in 1812 may 1, 2014 vol. 370 no. 18
A Randomized Trial of Protocol-Based Care for Early Septic Shock
The ProCESS Investigators*
ABSTR ACT
The members of the writing committee
(Donald M. Yealy, M.D., John A. Kellum,
M.D., David T. Huang, M.D., Amber E.
Barnato, M.D., Lisa A. Weissfeld, Ph.D.,
and Francis Pike, Ph.D., University of Pitts-
burgh, Pittsburgh; Thomas Terndrup, M.D.,
Ohio State University, Columbus; Henry
E. Wang, M.D., University of Alabama at
Birmingham, Birmingham; Peter C. Hou,
M.D., Brigham and Women’s Hospital,
Boston; Frank LoVecchio, D.O., Maricopa
Medical Center, Phoenix; Michael R. Fil-
bin, M.D., Massachusetts General Hos-
pital, and Nathan I. Shapiro, M.D., Beth
Israel Deaconess Medical Center — both
in Boston; and Derek C. Angus, M.D.,
M.P.H., University of Pittsburgh, Pitts-
burgh) assume responsibility for the con-
tent and integrity of the article. Address
reprint requests to Dr. Angus at the De-
partment of Critical Care Medicine, Uni-
versity of Pittsburgh, 3550 Terrace St.,
614 Scaife Hall, Pittsburgh, PA 15261, or
at angusdc@upmc.edu.
*A complete list of investigators in the
Protocolized Care for Early Septic Shock
(ProCESS) study is provided in the Supple-
mentary Appendix, available at NEJM.org.
This article was published on March 18,
2014, at NEJM.org.
N Engl J Med 2014;370:1683-93.
DOI: 10.1056/NEJMoa1401602
Copyright © 2014 Massachusetts Medical Society.
Background
In a single-center study published more than a decade ago involving patients pre-
senting to the emergency department with severe sepsis and septic shock, mortality
was markedly lower among those who were treated according to a 6-hour protocol
of early goal-directed therapy (EGDT), in which intravenous fluids, vasopressors,
inotropes, and blood transfusions were adjusted to reach central hemodynamic
targets, than among those receiving usual care. We conducted a trial to determine
whether these findings were generalizable and whether all aspects of the protocol
were necessary.
Methods
In 31 emergency departments in the United States, we randomly assigned patients
with septic shock to one of three groups for 6 hours of resuscitation: protocol-based
EGDT; protocol-based standard therapy that did not require the placement of a
central venous catheter, administration of inotropes, or blood transfusions; or usu-
al care. The primary end point was 60-day in-hospital mortality. We tested sequen-
tially whether protocol-based care (EGDT and standard-therapy groups combined)
was superior to usual care and whether protocol-based EGDT was superior to pro-
tocol-based standard therapy. Secondary outcomes included longer-term mortality
and the need for organ support.
Results
We enrolled 1341 patients, of whom 439 were randomly assigned to protocol-based
EGDT, 446 to protocol-based standard therapy, and 456 to usual care. Resuscitation
strategies differed significantly with respect to the monitoring of central venous
pressure and oxygen and the use of intravenous fluids, vasopressors, inotropes, and
blood transfusions. By 60 days, there were 92 deaths in the protocol-based EGDT
group (21.0%), 81 in the protocol-based standard-therapy group (18.2%), and 86 in
the usual-care group (18.9%) (relative risk with protocol-based therapy vs. usual
care, 1.04; 95% confidence interval [CI], 0.82 to 1.31; P=0.83; relative risk with
NEGATIVE
original article
The new engl and jour nal o f medicine
Goal-Directed Resuscitation for Patients
with Early Septic Shock
The ARISE Investigators and the ANZICS Clinical Trials Group*
The members of the writing committee
(Sandra L. Peake, M.D., Ph.D., Anthony
Delaney, M.D., Ph.D., Michael Bailey,
Ph.D., Rinaldo Bellomo, M.D., Peter A.
Cameron, M.D., D. James Cooper, M.D.,
Alisa M. Higgins, M.P.H., Anna Hold-
gate, M.D., Belinda D. Howe, M.P.H.,
Steven A.R. Webb, M.D., Ph.D., and Patri-
cia Williams, B.N.) assume responsibility
for the overall content and integrity of the
article. Address reprint requests to Ms.
Belinda Howe at the Australian and New
Zealand Intensive Care Research Centre,
Alfred Centre, Level 6 (Lobby B), 99 Com-
mercial Rd., Melbourne, VIC 3004, Aus-
tralia, or at anzicrc@monash.edu.
Abstr act
Background
Early goal-directed therapy (EGDT) has been endorsed in the guidelines of the Sur-
viving Sepsis Campaign as a key strategy to decrease mortality among patients pre-
senting to the emergency department with septic shock. However, its effectiveness
is uncertain.
Methods
In this trial conducted at 51 centers (mostly in Australia or New Zealand), we randomly
assigned patients presenting to the emergency department with early septic shock
to receive either EGDT or usual care. The primary outcome was all-cause mortality
within 90 days after randomization.
Results
NEGATIVEThe new engl and jour nal of medicine
original article
Trial of Early, Goal-Directed Resuscitation
for Septic Shock
Paul R. Mouncey, M.Sc., Tiffany M. Osborn, M.D., G. Sarah Power, M.Sc.,
David A. Harrison, Ph.D., M. Zia Sadique, Ph.D., Richard D. Grieve, Ph.D.,
Rahi Jahan, B.A., Sheila E. Harvey, Ph.D., Derek Bell, M.D., Julian F. Bion, M.D.,
Timothy J. Coats, M.D., Mervyn Singer, M.D., J. Duncan Young, D.M.,
and Kathryn M. Rowan, Ph.D., for the ProMISe Trial Investigators*
From the Clinical Tr
Care National Audit a
(P.R.M., G.S.P., D.A.H.
Department of Healt
and Policy, London Sc
Tropical Medicine (M
Faculty of Medicine
ABSTR ACT
Background
Early, goal-directed therapy (EGDT) is recommended in international guidelines for
the resuscitation of patients presenting with early septic shock. However, adoption
has been limited, and uncertainty about its effectiveness remains.
Methods
NEGATIVE
original article
The new england journal of medicine
Drotrecogin Alfa (Activated) for Adults
with Severe Sepsis and a Low Risk of Death
Edward Abraham, M.D., Pierre-François Laterre, M.D., Rekha Garg, M.D.,
Howard Levy, M.D., Ph.D., Deepak Talwar, M.D., Benjamin L. Trzaskoma, M.S.,
Bruno François, M.D., Jeffrey S. Guy, M.D., Martina Brückmann, M.D.,
Álvaro Rea-Neto, M.D., Rolf Rossaint, M.D., Dominique Perrotin, M.D.,
Armin Sablotzki, M.D., Ph.D., Nancy Arkins, R.N.,
Barbara G. Utterback, M.S., M.B.A., and William L. Macias, M.D.,
for the Administration of Drotrecogin Alfa (Activated)
in Early Stage Severe Sepsis (ADDRESS) Study Group*
nces
ty of
background
In November 2001, the Food and Drug Administration (FDA) approved drotrecogin
alfa (activated) (DrotAA) for adults who had severe sepsis and a high risk of death. The
abstract
NEGATIVE
NEGATIVE
The new england
journal of medicine
established in 1812 february 28, 2008 vol. 358 no. 9
Vasopressin versus Norepinephrine Infusion
in Patients with Septic Shock
James A. Russell, M.D., Keith R. Walley, M.D., Joel Singer, Ph.D., Anthony C. Gordon, M.B., B.S., M.D.,
Paul C. Hébert, M.D., D. James Cooper, B.M., B.S., M.D., Cheryl L. Holmes, M.D., Sangeeta Mehta, M.D.,
John T. Granton, M.D., Michelle M. Storms, B.Sc.N., Deborah J. Cook, M.D., Jeffrey J. Presneill, M.B., B.S., Ph.D.,
and Dieter Ayers, M.Sc., for the VASST Investigators*
ABSTR ACT
From the iCAPTURE Centre, Vancouver, BC
(J.A.R., K.R.W., A.C.G., M.M.S.); St. Paul’s
Hospital, Vancouver, BC (J.A.R., K.R.W.,
J.S., A.C.G., M.M.S., D.A.); Ottawa Hospi-
tal, University of Ottawa, Ottawa (P.C.H.);
Kelowna General Hospital, Kelowna, BC,
Background
Vasopressin is commonly used as an adjunct to catecholamines to support blood pres-
sure in refractory septic shock, but its effect on mortality is unknown. We hypothe-
sized that low-dose vasopressin as compared with norepinephrine would decrease
mortality among patients with septic shock who were being treated with conventional
(catecholamine) vasopressors.
The new england
journal of medicine
established in 1812 january 10, 2008 vol. 358 no. 2
Hydrocortisone Therapy for Patients with Septic Shock
Charles L. Sprung, M.D., Djillali Annane, M.D., Ph.D., Didier Keh, M.D., Rui Moreno, M.D., Ph.D.,
NEGATIVE
ICU RESEARCH AT A
CROSSROADS

-
-
s
,
k
e
-
s
t
r
n
e
e
,
t
m
e
r
a
o
e
.
d
s
s
B
,
Table 1. Clinical trials of biologic response modifiers in sepsis
Target Strategies
Endotoxin (LPS) Monoclonal antibodies
LPS: HA-1A, E5
Enterobacterial common
antigen
Toll-like receptor 4 (TLR4)
antagonists
Eritoran
TAK-242
Anti-CD14
Bactericidal permeability
increasing protein
Taurolidine
Alkaline phosphatase
Polymyxin B
Conjugate
Extracorporeal column
Lipid emulsion
Tumor necrosis factor (TNF) Monoclonal or polyclonal
antibodies
Soluble receptor constructs
Interleukin-1 (IL-1) Recombinant IL-1 receptor
antagonist
Platelet activating factor (PAF) Small molecule inhibitors
PAF acetylhydrolase
Eicosanoids Ibuprofen
Soluble phospholipase
A2 (sPLA2) inhibitor
Nitric oxide L-NMMA
Methylene blue
Hypercoagulability/disseminated
intravascular coagulation (DIC)
APC, Protein C concentrate
TFPI
Antithrombin
Anti-tissue factor antibody
Heparin
Thrombomodulin
Immune suppression Intravenous immunoglobulin
G-CSF, GM-CSF
Interferon g
Endocrinopathy Corticosteroids
Vasopressin
Others Selenium
Lactoferrin
Bradykinin antagonists
Statins
Extracorporeal hemoperfusion
Trends in Molecular Medicine xxx xxxx, Vol. xxx, No. x
Marshall JC. Trends Mol Med; 2014
tients with acute respiratory failure, in-
cluding effects of prone positioning (53–
55), ventilatory strategies (56–58), fluid
management (59), inhaled surfactant
(60), and anti-inflammatory therapies
(61–66). Other RCTs in this category as-
sessed the role of filgrastim in communi-
ty-acquired and nosocomial pneumonia
(67), the effects of selective decontamina-
tion of the digestive tract (68), enteral
nutrition (69), hemofiltration (70), he-
modynamic therapy (71, 72), the poten-
tial benefits of leukocyte-depleted red
blood cell transfusions (73, 74), and the
use of the pulmonary artery catheter
(6–8, 75). Importantly, some studies in
this group were designed to evaluate the
hypothesis that there would be no differ-
ence in mortality among groups, that is,
to show the noninferiority, rather than
the superiority, of a therapeutic strategy:
Chastre et al. (76) showed that adminis-
tration of antibiotics for 8 or 15 days
resulted in similar mortality rates in the
treatment of ventilator-a
monia; in the SAFE stud
administration was show
crystalloid for fluid repl
patients; and Hebert et
that the application of a r
fusion strategy was at le
liberal transfusion strate
Reporting and Metho
ity. We found adequate r
location concealment in
positive studies, six of se
tive studies, and 47 of 5
studies. When we analy
analysis carried out in
study, only five of ten
studies, two of seven
studies, and 37 of 55 (66%
included a precise descri
of analysis in the artic
intention-to-treat analysi
in 66 (90%) of the total
ies. Interestingly, we al
studies that did not clear
a
Cox regression analysis day 28: 1.54 (1.10–2.16), favoring invasive strateg
c
nonresponders to adrenocorticotropic hormone test; d
28-day mortality in all p
regression model: 0.71 (95% confidence interval [CI], 0.53–0.97, p ϭ .03), but sign
rates: ICU mortality relative risk (RR) of 0.89 (95% CI, 0.75–1.05), adjusted odds r
0.78–1.04), adjusted OR of 0.62 (95% CI, 0.36–1.05), p ϭ .08; e
control vs. hemo
Crit Care Med 2008 Vol. 36, No. 4
72 RCTs
“Any effect” (17)
“Positive” (10)
ICU RESEARCH AT A
CROSSROADS
100 studies
0 therapies

Current evidence-based research methods
look at the ways in which patients
resemble each other, rather than the
ways in which they differ.
“On average…”

Met inclusion criteria
Positive response Negative response
Non-responders
NEGATIVE RCT

Met inclusion criteriaNEGATIVE RCT
CASE MIX
aka.
“endotypes”

•Definitions based on diagnostic criteria
•Criteria are vague
•Criteria are arbitrary
•Criteria can be met in numerous ways
SYNDROMES & ENDOTYPES

me
ning an Illness
Box. IOM Diagnostic Criteria for Systemic Exertion
Intolerance Disease
Diagnosis requires that the patient have the following
3 symptoms:
1.Asubstantialreductionorimpairmentintheabilityto
engage in preillness levels of occupational, educa-
tional,social,orpersonalactivitiesthatpersistsformore
than 6 months and is accompanied by fatigue, which is
often profound, is of new or definite onset (not life-
long),isnottheresultofongoingexcessiveexertion,and
is not substantially alleviated by rest AND
2. Postexertional malaisea
AND
3. Unrefreshing sleepa
VAGUE CRITERIA

the ALI non-ARDS category of the
AECC definition; TABLE 4), 50% (95%
CI, 48%-51%) of patients met criteria
for moderate ARDS, and 28% (95% CI,
ARDS at baseline progressed to moder-
ate ARDS and 4% (95% CI, 3%-6%) pro-
gressed to severe ARDS within 7 days;
and 13% (95% CI, 11%-14%) of pa-
Table 3. The Berlin Definition of Acute Respiratory Distress Syndrome
Acute Respiratory Distress Syndrome
Timing Within 1 week of a known clinical insult or new or worsening respiratory
symptoms
Chest imaginga Bilateral opacities—not fully explained by effusions, lobar/lung collapse, or
nodules
Origin of edema Respiratory failure not fully explained by cardiac failure or fluid overload
Need objective assessment (eg, echocardiography) to exclude hydrostatic
edema if no risk factor present
Oxygenationb
Mild 200 mm Hg Ͻ PaO2/FIO2 Յ 300 mm Hg with PEEP or CPAP Ն5 cm H2Oc
Moderate 100 mm Hg Ͻ PaO2/FIO2 Յ 200 mm Hg with PEEP Ն5 cm H2O
Severe PaO2/FIO2 Յ 100 mm Hg with PEEP Ն5 cm H2O
Abbreviations: CPAP, continuous positive airway pressure; FIO2, fraction of inspired oxygen; PaO2, partial pressure of
arterial oxygen; PEEP, positive end-expiratory pressure.
aChest radiograph or computed tomography scan.
bIf altitude is higher than 1000 m, the correction factor should be calculated as follows: [PaO2/FIO2ϫ(barometric pressure/
760)].
cThis may be delivered noninvasively in the mild acute respiratory distress syndrome group.
ARBITRARY CRITERIA

MEETING SYNDROME
CRITERIA
ort, EBM emphasizes the similarities between patients, lea
hich they differ. These differences many of which occur a
en as inextricably linked to a comprehensive understanding
mulation of effective, individualized treatment plans. In th
bstantial physiologic heterogeneity amongst groups of patien
mixed outcomes in the setting of randomized trials, and so
ere is growing appreciation of the limitations of the curre
earch,3
and an increasing appetite to develop new methods
ur current understanding of human illnesses, and of the d
perience them, remains imprecise. This uncertainty is reﬂec
any are syndromic in nature, described by a constellation of
ssic diagnosis for systemic lupus requires that at least 4 o
esent, resulting in 11
4
+ 11
5
+ 11
6
+· · ·+ 11
11
= 1816 way
cumstances arise in the diagnosis of critical illness. Sepsi

Objective: -
Design: -
-
Methods:
-
-
Surviving Sepsis Campaign: International
Guidelines for Management of Severe Sepsis
and Septic Shock: 2012
R. Phillip Dellinger, MD1
; Mitchell M. Levy, MD2
; Andrew Rhodes, MB BS3
; Djillali Annane, MD4
;
Herwig Gerlach, MD, PhD5
; Steven M. Opal, MD6
; Jonathan E. Sevransky, MD7
; Charles L. Sprung, MD8
;
Ivor S. Douglas, MD9
; Roman Jaeschke, MD10
; Tiffany M. Osborn, MD, MPH11
; Mark E. Nunnally, MD12
;
Sean R. Townsend, MD13
; Konrad Reinhart, MD14
; Ruth M. Kleinpell, PhD, RN-CS15
;
Derek C.Angus, MD, MPH16
; Clifford S. Deutschman, MD, MS17
; Flavia R. Machado, MD, PhD18
;
Gordon D. Rubenfeld, MD19
; Steven A.Webb, MB BS, PhD20
; Richard J. Beale, MB BS21
;
Jean-Louis Vincent, MD, PhD22
; Rui Moreno, MD, PhD23
; and the Surviving Sepsis Campaign
Guidelines Committee including the Pediatric Subgroup*
Critical Care Medicine
0090-3493
10.1097/CCM.10.1097/CCM.0b013e31827e83af
LWW
Special Article
Special Article
SEPSIS SYNDROMES

Special Article
TABLE 1. Diagnostic Criteria for Sepsis
Infection, documented or suspected, and some of the following:
General variables
Fever (> 38.3°C)
Hypothermia (core temperature < 36°C)
Heart rate > 90/min–1
or more than two SD above the normal value for age
Tachypnea
Altered mental status
Significant edema or positive fluid balance (> 20mL/kg over 24hr)
Hyperglycemia (plasma glucose > 140mg/dL or 7.7 mmol/L) in the absence of diabetes
Inflammatory variables
Leukocytosis (WBC count > 12,000 µL–1
)
Leukopenia (WBC count < 4000 µL–1
)
Normal WBC count with greater than 10% immature forms
Plasma C-reactive protein more than two SD above the normal value
Plasma procalcitonin more than two SD above the normal value
Hemodynamic variables
Arterial hypotension (SBP < 90mm Hg, MAP < 70mm Hg, or an SBP decrease > 40mm Hg in adults or less than two SD
below normal for age)
Organ dysfunction variables
Arterial hypoxemia (Pao2
/FIO2
< 300)
Acute oliguria (urine output < 0.5mL/kg/hr for at least 2 hrs despite adequate fluid resuscitation)
Creatinine increase > 0.5mg/dL or 44.2 µmol/L
Coagulation abnormalities (INR > 1.5 or aPTT > 60 s)
Ileus (absent bowel sounds)
Thrombocytopenia (platelet count < 100,000 µL–1
)
Hyperbilirubinemia (plasma total bilirubin > 4mg/dL or 70 µmol/L)
Tissue perfusion variables
Hyperlactatemia (> 1 mmol/L)
Decreased capillary refill or mottling
° °
Crit Care Med
Arbitrary
Vague
Met in many ways
SEPSIS SYNDROMES

Pneumonia
Acute Kidney Injury
ARDS
Mechanical Ventilation
Shock
Vasopressors Dialysis
Cephalosporin Carbapenem
CT US
Therapies
Diagnoses
Blood
counts
Vitals
Biochem
Reports
BP
ECG

Data = Distinctions
46 KB59 KB

BIAS
AVAILABILITY
EMR notes
brain biopsy
Hepatitis
serology
Lung tumor
EGFR status
Heart rate
CBC
eye colour
gene expression
Fitbit data

RELIABLE
INFORMATIVE
EMR notes
brain biopsy
Hepatitis
serology
Lung tumor
EGFR status
gene expression
Fitbit data
CBC
Heart rate
eye colour

SEPSIS•Life-threatening organ dysfunction caused by dysregulated
host response to infection.
•Heterogeneous
•Complex molecular pathophysiology

Clinical Investigations
Toward a clinically feasible gene expression-based subclassification
strategy for septic shock: Proof of concept
Hector R. Wong, MD; Derek S. Wheeler, MD; Ken Tegtmeyer, MD; Sue E. Poynter, MD;
Jennifer M. Kaplan, MD, MS; Ranjit S. Chima, MD; Erika Stalets, MD; Rajit K. Basu, MD;
Lesley A. Doughty, MD
H
uman septic shock is a com-
plex syndrome displaying a
high degree of heterogene-
ity (1–3). Our inability to
manage and account for this heterogene-
ity is a major barrier to the conduct of
proposed that these repeated failures re-
flect our relative inability to address the
profound heterogeneity of septic shock,
rather than intrinsic flaws of the biolog-
ical principles being tested (1, 4). Robust
illness severity scores are currently avail-
hierarchical clustering. Importantly, the
gene expression patterns represented the
first 24 hrs of meeting criteria for septic
shock, which is an ideal time frame for
clinically relevant stratification. Post hoc
phenotype analyses revealed that patients
Objective: To develop a clinically feasible stratification strat-
egy for pediatric septic shock, using gene expression mosaics
and a 100-gene signature representing the first 24 hrs of admis-
sion to the pediatric intensive care unit.
Design: Prospective, observational study involving microarray-
based bioinformatics.
Setting: Multiple pediatric intensive care units in the United
States.
Patients: Ninety-eight children with septic shock.
Interventions: None other than standard care.
Measurements and Main Results: Patients were classified into
three previously published, genome-wide, expression-based sub-
classes (subclasses A, B, and C) having clinically relevant phe-
notypic differences. The class-defining 100-gene signature was
depicted for each individual patient, using mosaics generated by
the Gene Expression Dynamics Inspector (GEDI). Composite mo-
saics were generated representing the average expression pat-
terns for each of the three subclasses. Nine individual clinicians
served as blinded evaluators. Each evaluator was shown the 98
individual patient mosaics and asked to classify each patient into
one of the three subclasses, using the composite mosaics as the
reference point. The respective sensitivities, specificities, positive
predictive values, and negative predictive values of the subclas-
sification strategy were >84% across the three subclasses. The
classification strategy also generated positive likelihood ratios of
>16.8 and negative likelihood ratios of <0.2 across the three
subclasses. The ␬ coefficient across all possible interevaluator
comparisons was 0.81.
Conclusions: We have provided initial evidence (proof of con-
cept) for a clinically feasible and robust stratification strategy for
pediatric septic shock based on a 100-gene signature and gene
expression mosaics. (Crit Care Med 2010; 38:1955–1961)
KEY WORDS: microarray; gene expression; stratification; staging;
septic shock; pediatrics
ogous manner to what has been achieved in
the oncology field (7–11).
We (12) recently reported the identifi-
cation of three subclasses of children
with septic shock as a means to address
more effectively the intrinsic heterogene-
ity of this syndrome. The three subclasses
were initially identified based exclusively
on genome-wide expression patterns, us-
ing microarray analysis and unsupervised
METHODS
Patients and Mi
patients in the sept
their respective mic
previously reported
and confirmation
Board approval (12
wide expression stud
nome U133 Plus 2
Cincinnati, OH.
This study was supported, in part, by Grants
R01GM064619 and RC1HL100474 from the National
Institutes of Health (HRW).
The authors have not disclosed any potential con-
flicts of interest.
For information regarding this article, E-mail:
hector.wong@cchmc.org
Copyright © 2010 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/CCM.0b013e3181eb924f
Crit Care Med 2010 Vol. 38, No. 10

RESEARCH Open Access
Identification of sepsis subtypes in critically ill
adults using gene expression profiling
David M Maslove1,2*
, Benjamin M Tang3,4
and Anthony S McLean3
Abstract
Introduction: Sepsis is a syndromic illness that has traditionally been defined by a set of broad, highly sensitive
clinical parameters. As a result, numerous distinct pathophysiologic states may meet diagnostic criteria for sepsis,
leading to syndrome heterogeneity. The existence of biologically distinct sepsis subtypes may in part explain the
lack of actionable evidence from clinical trials of sepsis therapies. We used microarray-based gene expression data
from adult patients with sepsis in order to identify molecularly distinct sepsis subtypes.
Methods: We used partitioning around medoids (PAM) and hierarchical clustering of gene expression profiles
from neutrophils taken from a cohort of septic patients in order to identify distinct subtypes. Using the medoids
learned from this cohort, we then clustered a second independent cohort of septic patients, and used the
resulting class labels to evaluate differences in clinical parameters, as well as the expression of relevant
pharmacogenes.
Results: We identified two sepsis subtypes based on gene expression patterns. Subtype 1 was characterized by
increased expression of genes involved in inflammatory and Toll receptor mediated signaling pathways, as well as
a higher prevalence of severe sepsis. There were differences between subtypes in the expression of
pharmacogenes related to hydrocortisone, vasopressin, norepinephrine, and drotrecogin alpha.
Conclusions: Sepsis subtypes can be identified based on different gene expression patterns. These patterns may
generate hypotheses about the underlying pathophysiology of sepsis and suggest new ways of classifying septic
patients both in clinical practice, and in the design of clinical trials.
Keywords: Sepsis, severe sepsis, septic shock, gene expression profiling, microarray analysis, biomedical informatics,
critical care, intensive care
Maslove et al. Critical Care 2012, 16:R183
http://ccforum.com/content/16/5/R183
10
ering
−40 −20 0 20 40
−20−1001020
Final PAM clustering
Component 1
Component2
each plot, the patients are plotted within a two-dimensional
s in the first plot are colored according to the cluster
nbank. The colors in the second and third plots reflect the
results of the clustering at that stage. (A) Initial clustering
Page 4 of 11
between patients with sepsis and non-sepsis controls.
This latter approach has the potential to exclude genes
that may be important in differentiating sepsis subtypes,
rather than differentiating sepsis from controls.
Our results highlight the complexity and heterogeneity
of sepsis at the molecular level, a finding in keeping with
were obtained from neutrophils collected within
24 hours of admission to the ICU. While it has been sug-
gested that the tissue used and timing of microarray ana-
lysis could have a significant impact on gene expression
studies in sepsis [33], the experimental conditions were
similar for all patients and for both cohorts, so that dif-
−40 −20 0 20 40
−20−10010
Validation cohort clustering
Component 1
Component2
Figure 4 Validation cohort clustering. Clusters resulting from assignment of the validation cohort samples to the closest derivation medoid.
Maslove et al. Critical Care 2012, 16:R183
http://ccforum.com/content/16/5/R183
Page 7 of 11
Derivation Validation

Pathway p-value Bonferonni
Inflammation mediated by chemokine and
cytokine signaling pathway
2.7E-08 0.0000048
Toll receptor signaling pathway 7.66E-05 0.0135
T cell activation 4.75E-03 0.835
p38 MAPK pathway 5.38E-03 0.946
JAK/STAT signaling pathway 7.59E-03 1
DIFFERENTIATING PATHWAYS

Mortality (%) Male (%) Severe sepsis (%) Ventilated (%) Dialysis (%) Pressors (%)
0.00.10.20.30.40.50.60.7
Subtype1
Subtype 2
Length of stay (d) Age Apache II Apache III SAPS II
020406080
Subtype1
Subtype 2
*
CLINICAL
FEATURES

Pharmacogenes
Gene ID Fold Change
drotrecogin alpha
TFPI 1.74
SERPINB2 1.61
CP 1.52
GGCX 1.49
SERPIND1 1.58
SERPINB6 1.82
SERPINE1 1.43
THBD 0.53
F5 0.48
vasopressin
GNG11 1.73
GNG5 1.43
GNAQ 0.58
hydrocortisone
5-LOX 0.34
ANXA1 0.64
norepinephrine
NNMT 1.32
MOXD1 1.42
PharmGKB
Reduction of the multiple organ injury and dysfunction caused
by endotoxemia in 5-lipoxygenase knockout mice and by
the 5-lipoxygenase inhibitor zileuton
Marika Collin,* Antonietta Rossi,†
Salvatore Cuzzocrea,‡
Nimesh S. A. Patel,* Rosanna Di Paola,‡
Julia Hadley,* Massimo Collino,* Lidia Sautebin,†
and Christoph Thiemermann*,1
*Centre for Experimental Medicine, Nephrology & Critical Care, The William Harvey Research Institute,
Queen Mary, University of London, United Kingdom; †
Department of Experimental Pharmacology, Universita`
‘Federico II’, Naples, Italy; and ‡
Department of Clinical and Experimental Medicine and Pharmacology,
University of Messina, Italy
Abstract: The role of 5-lipoxygenase (5-LOX) in
the pathophysiology of the organ injury/dysfunc-
tion caused by endotoxin is not known. Here, we
investigate the effects of treatment with 5-LOX
INTRODUCTION
Inflammation is a complex set of interactions among soluble
mediators, circulating cells, and vessel walls and may arise in
lipopolysaccharide (LPS; Escherichia coli, 6 mg/kg
i.v.) or vehicle (saline). 5-LOX؊/؊
mice and wild-
type littermate controls were treated with LPS (E.
coli, 20 mg/kg intraperitoneally) or vehicle (sa-
line). Endotoxemia for 6 h in rats or 16 h in mice
resulted in liver injury/dysfunction (increase in the
serum levels of aspartate aminotransferase, alanine
aminotransferase, ␥-glutamyl transferase, alkaline
phosphatase, bilirubin), renal dysfunction (creati-
nine), and pancreatic injury (lipase, amylase). Ab-
sence of functional 5-LOX (zileuton treatment or
targeted disruption of the 5-LOX gene) reduced
the multiple organ injury/dysfunction caused by
endotoxemia. Polymorphonuclear leukocyte infil-
tration (myeloperoxidase activity) in the lung and
ileum as well as pulmonary injury (histology) were
markedly reduced in 5-LOX؊/؊
mice. Zileuton also
reduced the LPS-induced expression of CD11b/
CD18 on rat leukocytes. We propose that endog-
enous 5-LOX metabolites enhance the degree of
multiple organ injury/dysfunction caused by severe
endotoxemia by promoting the expression of the
adhesion molecule CD11b/CD18 and that inhibi-
tors of 5-LOX may be useful in the therapy of the
organ injury/dysfunction associated with endotoxic
shock. J. Leukoc. Biol. 76: 961–970; 2004.
Key Words: shock ⅐ ␤2-integrins ⅐ CD11a/CD18 ⅐ CD11b/CD18
⅐ leukotrienes
5-LOX is predominantly expressed by cells of myeloid origin,
particularly neutrophils, eosinophils, macrophages/monocytes,
and mast cells [4, 5]. LTs are involved in the genesis of
inflammation and edema, because of their effects on vascular
permeability, plasma extravasation, and diapedesis of white
blood cells [6, 7], and they may also play an important role in
adaptive immune responses [8]. There is now good evidence
that LTs play a pivotal role in the pathophysiology of asthma [9,
10] and psoriasis [11, 12], as well as in conditions associated
with ischemia-reperfusion (I/R) of skin [13, 14], brain [15], and
kidney [13, 16, 17]. LTs also play a physiological role in the
host defense against microbial infections [18].
LTB4 is a proinflammatory mediator that activates polymor-
phonuclear leukocytes (PMN), thus changing their shape and
promoting their binding to endothelium by inducing the ex-
pression of cell-adhesion molecules. The localization of leuko-
cytes to the site of inflammation requires several families of
adhesion molecules. Firm adhesion of leukocytes to the micro-
vascular endothelium is dependent on the function of the class
of adhesion molecules called ␤2-integrins, which are expressed
on neutrophil surface and interact with members of the immu-
noglobulin (Ig) supergene family expressed on endothelial cells
such as intercellular adhesion molecule-1 (ICAM-1) [19–22].
1
Correspondence: Centre for Experimental Medicine, Nephrology & Criti-
cal Care, William Harvey Research Institute, Queen Mary, University of
London, Charterhouse Square, London, EC1M 6BQ, UK. E-mail:
c.thiemermann@qmul.ac.uk
Received June 14, 2004; accepted August 3, 2004; doi: 10.1189/
jlb.0604338.
Journal of Leukocyte Biology Volume 76, November 2004 961
So what about those trials?

Pharmacogenes
Gene ID Fold Change
drotrecogin alpha
TFPI 1.74
SERPINB2 1.61
CP 1.52
GGCX 1.49
SERPIND1 1.58
SERPINB6 1.82
SERPINE1 1.43
THBD 0.53
F5 0.48
vasopressin
GNG11 1.73
GNG5 1.43
GNAQ 0.58
hydrocortisone
5-LOX 0.34
ANXA1 0.64
norepinephrine
NNMT 1.32
MOXD1 1.42
PharmGKB
This Provisional PDF corresponds to the article as it appeared upon acceptance. Copyedited and
fully formatted PDF and full text (HTML) versions will be made available soon.
Flavocoxid, a dual inhibitor of COX-2 and 5-LOX of natural origin, attenuates the
inflammatory response and protects mice from sepsis
Critical Care 2012, 16:R32 doi:10.1186/1364-8535-16-R32
Alessandra Bitto (abitto@unime.it)
Letteria Minutoli (lminutoli@unime.it)
Antonio David (adavid@unime.it)
Natasha Irrera (nirrera@unime.it)
Mariagrazia Rinaldi (mariagrazia.rinaldi@yahoo.it)
Francesco S Venuti (fsvenuti@unime.it)
Critical Care
This Provisional PDF corresponds to the article as it appeared upon acceptance. Copyedited and
fully formatted PDF and full text (HTML) versions will be made available soon.
Flavocoxid, a dual inhibitor of COX-2 and 5-LOX of natural origin, attenuates the
inflammatory response and protects mice from sepsis
Critical Care 2012, 16:R32 doi:10.1186/1364-8535-16-R32
Alessandra Bitto (abitto@unime.it)
Letteria Minutoli (lminutoli@unime.it)
Critical Care
So what about those trials?

MOLECULAR
COMPLEXITY
X
DYNAMIC
CHANGE
(GRO-b), CCL2 (MCP-1), CXCL8 (IL-8) and CXCL10)
maximum 2–4 h after endotoxin administration, consis
early activation of innate immunity. Subsequently, the e
of several members of the nuclear factor kappa/relA
transcription factors (NFKB1, NFKB2, RELA and RELB
their zenith.
The time period 4–6 h after endotoxin injection seemed
the expression of a number of transcription factors was
including both those that initiate and those that limit
immune response. In the former group, these included
transducer and activators of transcription (STAT genes)
cAMP-response element-binding protein (CREB) and
enhancer binding protein (CEBP) gene families. Transcri
tors limiting the innate immune response included sup
cytokine signalling 3 (SOCS3) and IKBK genes. There was a
(4–6 h) in increased mRNA abundance of secreted and m
associated proteins that limit the inflammatory response,
IL1RAP, IL1R2, IL10 and TNFRSF1A. Together, these dat
hensively document the temporal modulation of genes c
the innate immune response in a human model that
from an acute proinflammatory phase to unencumbered
regulation, concluding with full recovery and a normal ph
To further elucidate the global changes during inflamm
subsequent return to homeostasis, we sought to compu
decipher the principal networks involved. The specificity o
tions for each gene was calculated, as defined by the percen
direct connections to other genes showing significant trans
changes. A network pathway was initiated by the gene
highest specificity of connections, and was propagated ac
the descent of the specificity. Individual significant pathw
ified by a statistical likelihood calculation (P , 0.0001) we
to represent the biological processes.
Our global representation of the inflammatory respons
toxin, shown in Fig. 3a, comprises a network of 1,556 gene
interactions. This network consists of a subset of 1,214 ge
responsive to in vivo endotoxin administration, and 342 a
Vol 437|13 October 2005|doi:10.1038/nature03985 Vol 437|13 October 2005|doi:10.1038/nature0398

Geneexpression
Time
Sepsis Genomics

Sample Tumor ?
Timing Once ?
Analysis Static ?
Cancer Care Critical Care

ster cohesiveness between whole blood and leukocyte iso-
s, with whole blood-derived data yielding significantly
her average silhouette widths (median values 0.28 and
9 respectively, P ¼ 0.002). In the specific task of forming
(median values 0.21 for whole blood vs 0.16 for isola
P ¼ 0.13). The lack of statistical significance in these analy
may in part be due to a loss of statistical power in this sma
subset of studies.
IG. 3. Principal components analysis (PCA) representations of gene expression data yielding clusters with differing average silhouette wid
shapes of the points reflect the cluster assignment as determined by the partitioning around medoids (PAM) clustering algorithm used in the analysis. F
pes represent sepsis cases while open shapes are control samples, which in the case of the two studies shown were non-septic ICU patients. The left p
ws gene expression data derived from neutrophil RNA (GSE6535) that results in overlapping clusters, while the right panel shows gene expression
ved from whole blood RNA (GSE32707) that results in more cohesive clusters. These differences are reflected in the average silhouette widths (0.16 fo
rophil-derived data vs 0.48 for whole blood-derived data). Clustering of whole blood data revealed a distinct subset of controls, whereas sepsis and coMaslove & Marshall (2016), Shock, 293(3)
STANDARDS & METHODS
Sp 94% vs 78%
Avg. Sil (k=2) 0.41 vs 0.29

• Phase II RCT examining the effects of bovine
lactoferrin in preventing nosocomial infection
in the ICU
• Over 100 patients enrolled at 4 centres in
Canada
• Time series gene expression data from a
subset of ~ 80 patients
PREVAIL
Day 28
Baseline
Day 3
Day 7
Day 14
Day 21

Objective:
Design:
Setting:
Patients:
Interventions:
1
2
-
-
Multicenter Comparison of Machine Learning
Methods and Conventional Regression for
Predicting Clinical Deterioration on the Wards
Matthew M. Churpek, MD, MPH, PhD1
; Trevor C. Yuen, MS1
; Christopher Winslow, MD2
;
David O. Meltzer, MD, PhD1
; Michael W. Kattan, MBA, PhD3
; Dana P. Edelson, MD, MS1
Copyright © 201ght © 2016 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.
www.ccmjournal.org
Patients:
Interventions:
Measurements And Main Results: -
-
-
-
-
-
p
Conclusions:
-
Crit Care Med
Key Words:
/CCM.0000000000001571
-
-
-
-
-
-
-
HEART RATE
MONITORING

HEART RATE
VARIABILITY
HEALTHY SICK

DFA Alpha 0.30 20.30 (0.47) (0.51) 20.28 20.25 0.13 20.20 20.23 0.07
Wavelet AUC (0.77) (0.72) 20.17 (0.51) (0.69) (0.68) 20.08 (0.72) (0.58) (0.99) 0.01
doi:10.1371/journal.pone.0006642.t005
Figure 1. Individual smoothed percentage change. ((HRV) in wavelet variability. Individual wavelet (HRV is studied every 1 h in infected (n = 14)
Ahmad et al. PLoS ONE, 2009
Variability analysis

●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
0
1000
2000
3000
0 5 10 15
SOFA score
SDNN
factor(value1num)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
HRV in
ICU
“Healthy” vs. ICU
(P = 0.015)

●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●●
●
●
●
●
●
0
1000
2000
3000
0 5 10 15
SOFA score
SDNN
factor(value1num)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
HRV in
ICU
“Within ICU”
(P = 0.04)

HRV in
ICU
intervention
HRV
RESPONSE

artefact artefact missing
DATA QUALITY

HR
RR
SpO2
ABPm
ABPd
ABPs
NBPm
NBPd
NBPs
0 5000 10000 15000 20000
Number of measurements
Vitalsign
Non−terminal zeros
Terminal zeros, not assoc. with death
Terminal zeros assoc. with death
HR
RR
SpO2
ABP
NBP
0 50000 100000 150000
Number of VSDs
Vitalsign
Insufficient data
Sufficient data with gaps
Sufficient data with no gaps
0
50000
100000
150000
ABP NBP
vital sign
count
Errors
No errors
numberofmeasurements
Maslove et al. Critical Care Medicine 2016, (In press)

ABPm ABPd ABPs NBPm NBPd NBPs HR RR SpO2
0.0
0.5
1.0
1.5
2.0
Time of day
Percentoutliers
ABPm ABPd ABPs NBPm NBPd NBPs HR RR SpO2
0
1000
2000
3000
4000
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
03:00
09:00
15:00
21:00
Time of day
Outliercount

0
100
200
300
1980 1990 2000 2010
year
count
0
100
200
300
1980 1990 2000 2010
year
count
“wearable technology”
clinical trial
“Enthusiasm gap”

Interrogation of Patient Smartphone Activity Tracker
to Assist Arrhythmia Management
Joshua Rudner, DO; Carol McDougall, MSN, APN-C; Vivek Sailam, MD; Monika Smith, DO; Alfred Sacchetti, MD*
*Corresponding Author. E-mail: sacchetti1011@gmail.com, Twitter: @SacchettiAlfred.
A 42-year-old man presented to the emergency department (ED) with newly diagnosed atrial ﬁbrillation of unknown
duration. Interrogation of the patient’s wrist-worn activity tracker and smartphone application identiﬁed the onset of
the arrhythmia as within the previous 3 hours, permitting electrocardioversion and discharge of the patient from the ED.
[Ann Emerg Med. 2016;-:1-3.]
0196-0644/$-see front matter
Copyright © 2016 by the American College of Emergency Physicians.
http://dx.doi.org/10.1016/j.annemergmed.2016.02.039
INTRODUCTION
Activity trackers are becoming more commonplace as
examination results. Auscultation of the heart demonstrated
a rapid irregularly irregular rhythm with no abnormal heart
CARDIOLOGY/CASE REPORT
Annals of Emergency Medicine, 2016

Kroll, Boyd, Maslove. JMIR 2016, (In submission)
FITBIT HEART RATES

ICU DISEASES ARE SYNDROMIC.
SUBTYPES ABOUND.
DATA = DISTINCTION.
DATA ABOUND.
Precision
diagnosis

similar patients. The solid and dashed lines are the mean and 95% confidence
en training data homogeneity and size is apparent; as the number of similar patients in
t a rapid rate thanks to increasing training data size but starts to degrade gradually due
er the receiver operating characteristic curve; AUPRC: area under the precision-
Personalized Mortality Prediction
Lee, Maslove, Dubin. PLOS One (2015)
PATIENT
SIMILARITY
Fig 1. Mortality prediction performance of death counting among similar patients. The solid and dashed lines are the mean and 95% confidence
intervals, respectively, from 10-fold cross-validation. A trade-off between training data homogeneity and size is apparent; as the number of similar patients in
the training data increases, predictive performance improves initially at a rapid rate thanks to increasing training data size but starts to degrade gradually due
to decreasing homogeneity within the training data. AUROC: area under the receiver operating characteristic curve; AUPRC: area under the precision-
recall curve.
doi:10.1371/journal.pone.0127428.g001
Personalized Mortality Prediction

All High SOFA Low SOFA
0.00
0.25
0.50
0.75
0.00
0.25
0.50
0.75
0.00
0.25
0.50
0.75
0.00
0.25
0.50
0.75
CCUCSRUMICUSICU
AUPRC AUROC AUPRC AUROC AUPRC AUROC
Areaundercurve
Score
Custom
SAPS
Lee, Maslove. J Intensive Care Med (2015)
LOCAL
DATA

ACKNOWLEDGEMENTS
Dr. J. Gordon Boyd
Dr. John Muscedere
Dr. Joon Lee
Michael Wood
Victoria Tolls
Usman Raza
Miranda Hunt
Nicole O’Callahagn
Ilinca Georgescu
david.maslove@queensu.ca
www.conduitlab.org

Critical Care Endotypes

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Critical Care Endotypes

Similar to Critical Care Endotypes (20)

Recently uploaded

Recently uploaded (20)

Critical Care Endotypes