Hallie Prescott, MD, MSc
Professor, University of Michigan
Le a d, Mic higa n Hos pita l Me dic ine Sa fe ty Cons ortium’s Se ps is Initia tive (HMS-Sepsis)
Perioperative Sepsis:
Identification and Treatment
Disclosures
Financial COI: Au r o b a c Therapeutics
Research funding: NIH, AHRQ, VA
Salary support: CDC, BCBSM
Pertinent roles: Lead, HMS-Sepsis (Michigan statewide sepsis CQI)
Co -chair, Surviving Sepsis Campaign Guidelines
This talk represents my views and does not necessarily represent views of US
government, Dept of Veterans Affairs, or any other organization.
Outline
1. Definition
2. Brief epidemiology
3. Diagnosis
4. Antimicrobial therapy
5. Hemodynamic resuscitation
“Sepsis is a life-threatening
complication of infection that arises
when infection or the body’s response
injures tissues and organs.”
What is sepsis?
Septic ShockSepsisIn fe ct io n
Body views microbe
as a problem.
Body accepts
microbe.
Body’s response to microbe is damaging, and damage
may continue even after the microbe is eradicated.
Threshold at which host-microbe interaction causes harm
Colonization
Body benefits
from microbe.
5
How I teach sepsis
Symbiosis
Current operational definitions
Sepsis-3 defines clinically
significant organ dysfunction
as ≥2 new SOFA point s
SOFA Score (range –6)
6 Systems 0 (normal) – 4 (complete failure)
Respiration Pa O2/ FIO2
Coagulation Platelet count
Live r Bilirubin
Cardiovascular MAP, Vasopressors
CNS Glasgow Coma
Re n a l Creatinine, UOP
.
Singer, et al JAMA, 2016.
Seymour, et al. JAMA, 2016.
Current operational definitions
qSOFA identifies clinical signs most predictive of poor outcome
Gla s go w
Coma Scale
<15
Respiratory
Ra t e
≥22
SBP
<100
.
Singer, et al JAMA, 2016.
Seymour, et al. JAMA, 2016.
Outline
1. Definition
2. Brief epidemiology
3. Diagnosis
4. Antimicrobial therapy
5. Hemodynamic resuscitation
Sepsis is common and deadly
1.7 million adults hospitalized annually in U.S. with sepsis
(More than heart attack and stroke combined)
350,000 sepsis-related deaths.
Sepsis contributes to 33%-50% of all in-hospital deaths in the U.S.
.
Rhee, at al. JAMA, 2017.
Liu, et al. JAMA, 2014.
Rhee, et al. JAMA Network Open, 2019.
Hospital-onset sepsis is particularly deadly
Hosp it a l-onset sepsis accounts for 1 in 8 sepsis cases, but 1 in 3 sepsis deaths
1 in 3 patients with hospital-onset sepsis dies in-hospital
2-fold higher mortality than community-onset sepsis
3-fold higher mortality than similar admits not complicated by sepsis
Rhee, at al. CCMed, 2019.
Outline
1. Definition
2. Brief epidemiology
3. Diagnosis
4. Antimicrobial therapy
5. Hemodynamic resuscitation
Always be on the lookout for sepsis
Hospital onset sepsis complicates 1 in 200 hospitalizations
1-4% of general elective surgeries
5-40% of emergency or major abdominal surgeries
Always be on the lookout for sepsis
Septic ShockSepsisIn fe ct io n
Body views microbe as
a problem.
Body accepts
microbe.
(Colonization)
Body’s response to microbe is damaging, and damage
may continue even after the microbe is eradicated.
Colonization
Body benefits
from microbe.
13
Symbiosis
The sooner infection or sepsis is diagnosed, the more treatment-responsive.
Asking daily: is my patient progressing as expected?
Is my patient
progressing as
expected?
Continue to
monitor
Yes, the degree of illness severity is as
expected based on the surgery performed.
Yes, clinical trajectory (vitals, exam, labs) is
improving as expected.
Is the patient progressing as expected?
Is my patient
progressing as
expected?
Continue to
monitor
Yes.
No, worse illness severity than expected post-op.
No, acutely worsening.
No, not improving as expected.
Cohen, et al. Surgery, 2023.
Is the patient progressing as expected?
Is my patient
progressing as
expected?
Continue to
monitor
Yes.
Additional evaluation for infection, guided by type of surgery and symptoms:
•Physical examination, symptom assessment
•Imaging (e.g., chest X-ray, pleural space ultrasound, abdominal ultrasound/CT)
•Labs (e.g., lactate, procalcitonin)
•Micro (e.g., blood culture, peritoneal fluid culture, pleural fluid culture)
No.
Lots of novel diagnostics aids are hitting the market
Diagnost ic Aid
Co m p a n y
Immunoscore Prenosis
Intellisep Cytovale
Me Me d BV Me Me d
Monocyte Distribution Width Beckman Coulter
Pancreatic Stone Protein Albionic
Septicyte ImmuneExpress
Triverity Infammatix
6Ne w
Insufficient evidence
to make a recommendation regarding use of
novel rapid host diagnostics.
Prescott, et al. CCMed, 2026.
Infection Acute Organ
Dysfunction
On presentation
Could organ dysfunction be
due to infection?
Is infection-related acute
organ dysfunction present?
Ongoing
New clinical
deterioration
(especially new
q SOFA criteria)
Could my patient’s deterioration be due
to a new infection?
How I approach sepsis diagnosis in my practice
Are there signs/symptoms of new infection, e.g., redness,
swelling, purulent drainage, cough, sputum?
Is there evidence of an alternate, competing diagnosis?
Outline
1. Definition
2. Brief epidemiology
3. Diagnosis
4. Antimicrobial therapy
5. Hemodynamic resuscitation
2026 SSC Recs on antibiotic timing, same as 2021
Prescott, et al. CCMed, 2026.
Rationale for stratified timing recommendation
Tim e -to-antibiotics matters,
particularly for patients with shock
Diagnostic uncertainty is
common in practice
(38% in ICU; 9% in ED)
Seymour,
e t a l. NEJ M
. 2017
Liu ,
e t a l. AJ RCCM
. 2017
Peltan,
et al. Chest.
2019
Pak,
e t a l. CID
. 2023
He ch t m a n ,
e t a l. AJ RCCM
. 2024
Kl e i n Louwenberg,
e t a l. Cr it Ca r e
, 2015.
Rhee,
e t a l. Cr it Ca r e
, 2016.
Ta ylo r,
et a l. An n a l s ATS
, 2020.
Shappell,
et al. Crit Care Med
, 2021.
Hooper,
et a l. Clinica l infectious Dis
., 2023.
Tim e -to-antibiotics matters,
particularly for patients with shock
Seymour,
e t a l. NEJ M
. 2017
Liu ,
e t a l. AJ RCCM
. 2017
Peltan,
et al. Chest.
2019
Pak,
e t a l. CID
. 2023
He ch t m a n ,
e t a l. AJ RCCM
. 2024
Shock
He ch t m a n ,
e t a l. AJ RCCM
. 2024
Urgency of treatment differs by illness severity
No shock
In practice: I stratify urgency by hypotension
Miller, et al. AJRCCM, 2013.
Whiles, et al. Crit. Care Med, 2017.
Bisa rya , et al. CHEST, 2022.
Evans, et al. Intensive Care Med and Critical Care Med. 2021.
Hypotension No Hypotension
Progression to shock NOT
knowable in real-time.
Timely treatment avoids
progression to septic shock.
Antibiotic selection should be tailored to patient
Treat likely pathogens based on:
•site of infection
•local epidemiology
•patient-specific risk factors for MDR or atypical pathogens
Avoid unnecessary coverage where feasible
Lung and urinary tract are common sites of
infection, almost never anaerobic.
Is empiric anti-anaerobic coverage for
undifferentiated sepsis associated with harm?
Chanderraj, et al. JAMA IM, 2024
Real-world experiment: pip-tazo drug shortage
Chanderraj, et al. JAMA IM, 2024
4523 treated with vancomycin and pip-tazo
3046 treated with vancomycin and cefepime
Chanderraj, et al. JAMA IM, 2024
5% abs. mortality reduction with Cefepime over Pip/Tazo
Chanderraj, et al. JAMA IM, 2024.
Finding is supported by preclinical and translational data
2026 SSC Guidelines emphasize stewardship
Ne w
Suggest
against use of empiric anti-anaerobic coverage in patients at
low risk of anaerobic infection (i.e., lung, urine source)
Ne w
Suggest use
of empiric anti-anaerobic coverage in patients at high risk
of anaerobic infection (e.g., intra-abdominal, obstetric, necrotizing)
Revised
Suggest
against use of empiric anti-fungal coverage
implication: case-by-case use, only in highest risk patients
Upgraded
Recommend
antibiotic de-escalation when cultures return
Prescott, et al. CCMed, 2026.
In sum, antimicrobial therapy should be
Tim e ly
Delivered promptly, especially if hypotensive
Effective
Cover the likely pathogens
Appropriate
Avoid coverage of unlikely pathogens (wherever feasible)
Prescott, et al. CCMed, 2026.
Outline
1. Definition
2. Brief epidemiology
3. Diagnosis
4. Antimicrobial therapy
5. Hemodynamic resuscitation
Fluid resuscitation guidelines
Up to
30 ml/kg
bolus
Additional fluid guided by
dynamic assessment of fluid-responsiveness
30 ml/kg
bolus
Additional fluid guided by
dynamic assessment of fluid-responsiveness
2021, 2026 SSC Guidelines suggestion:
2025 ESICM Guidelines on fluid therapy:
2025 ESICM Guidelines on shock:
Up to
30 ml/kg
bolus
Additional fluid guided by dynamic assessment of fluid
responsiveness AND assessment of fluid tolerance
Initial pragmatic fluid bolus
Prescott, et al. CCMed, 2026.
Transition to personalized resuscitation approach
Upgraded
certainty
Suggest
using dynamic assessment of fluid responsiveness to guide ongoing
resuscitation. (
low certainty
)
Remark: Dynamic assessment refers to change in CO or SV in response to passive leg raise or fluid
challenge (e.g., 500 ml in 5-10 min).
Ne w Insufficient evidence to rec. a specific device (e.g. Starling, PiCCO)
Prescott, et al. CCMed, 2026.
Why dynamic assessment? 1- predicts response to fluid
Dynamic assessment using PLR predicts whether CO
will inc re a s e with IV fluid bolus .
•Se ns itivity: 0.85
•Spe c ific ity: 0.91
•Pos itive LR: 11
•Negative LR: 0.13
Exa m findings (e.g., dry mucus membranes, peripheral
edema) do not predict responsiveness.
Heterogeneity of patient populations, criteria for fluid-responsiveness (10-
15% increase in SV or CO), and monitoring device (TTE, pulse contour,
bioreactance).
Monnet, et al. Intensive Care Medicine, 2016. (21 Studies, ~1000 patients, dynamic measures)
Bentzer, et al. JAMA, 2016. (50 studies, ~2500 patients, static and dynamic measures)
Why dynamic assessment? 2- improves outcomes
In systematic review of 18 RCTs, fluid management by dynamic
measures is associated with:
•Lower mortality: RR 0.91 (moderate certainty)
•Less renal replacement therapy (RRT): RR 0.75 (moderate certainty)
•More ventilator-free days: 0.8 more (very low certainty)
He te roge ne ity of pa tie nt popula tions (s urgic a l ICU, m e dic a l ICU, CV s urge ry ICU), c rite ria for
fluid-responsiveness (variable % increase in SVV, SV, PPV or CO), and monitoring device.
Bedna rczyk, et al. C C Me d , 2017.
Ehrmann, et al. Crit Care Explorer. 2019.
Basmaji, et al. J Intensive Care Med. 2024.
FRESH Trial
Patients: 124 patients with septic shock in 13 ICUs in US and UK, after
mean 2.3L fluid resuscitation (~25 ml/kg)
•23% with comorbid heart failure
Intervention: protocol-guided fluid and vasopressor titration based on
“fluid responsiveness” assessment with PLR
•42% fluid-responsive on 1st assessment; 82% ever fluid-responsive; 70% with change
Comparator: usual care
Outcome: fluid balance (primary)
Dougla s , et al. CHEST, 2020.
FRESH trial protocol
Dougla s , et al. CHEST, 2020.
FRESH trial outcomes
Outcomes
Intervention
(N=83)
Control
(N=41) p
Fluid balance at 72 hours or day 3, mean 0.7 L 2.0 L 0.02
Receipt of renal replacement therapy 5% 18% 0.04
Receipt of invasive mechanical ventilation 18% 34% 0.04
Length of ICU stay, mean 3.3 day 6.2 day 0.11
30-day mortality 15.7% 22.0% NR
Douglas, et al. CHEST, 2020.
What we see happening in clinical practice…
500
ml
Recheck
static
measures
500
ml
Recheck
static
measures
500
ml
Recheck
static
measures
500
ml
Recheck
static
measures
“Stuttering resuscitation”:
1L No more fluid due to clinical history, worry about fluid-related harms
“Aborted resuscitation”:
≥30 ml/kg resuscitation lower mortality in HMS
Munroe, et al, JNO, 2026.
Top-performing hospitals are pragmatic and proactive
Prescott, et al. CHEST, 2024.
Proactive and pragmatic
Give 30 ml/kg to most
patients with severe sepsis
Unified message from ICU
and emergency department
Slow and cautious
Give 500ml, recheck
Don’t want to cause
volu m e ove rlo a d
Lo we s t
adjusted
morta lity
Highe s t
adjusted
morta lity
In conclusion
Sepsis is a time-sensitive medical emergency.
Prompt eval/diagnosis, timely/effective/appropriate antimicrobial therapy,
and tailored resuscitation save lives.
Closing quote
“History will judge us not by how well we treated sepsis in the ICU,
but by whether fewer people ever needed to get there”
–Tex Kissoon
2026 Critical Care Congress
Keynote Address
Questions
Slide Appendix
Target populations in JNO study
Is NICOM reliable in my patient?
Study NPopulation Gold Standard Correlation
coefficient Sensitivity Specificity
Squara et al.
2007 110 Post-cardiac surgery Thermodilution
(PAC) 0.82 93% 93%
Raval et al.
2008
111
(5 centers) CVC ICU, cath lab
Thermodilution
(PAC) 0.78 NR NR
Rich et al. 2012 50 Pulmonary hypertension
in cath lab Fick during RHC 0.83 88.9%
(directional
change)
100%
(directional
change)
Rali, et al. 2020 50 Cardiogenic shock Thermodilution
Indirect Fick
0.132
0.275
Pandhita et al.
2021 18 heart transplant eval
LVEF 19% ± 7%
Thermodilution
(PAC)
0.88 rest
0.92 PLR NR NR
NICOM not feasible with: Pulseless LVAD; HR > 150-160
PLR false neg with: LE amputation, intra-abdominal HTN
