Past, Present, and Future of
Bleeding Management
Research and Quality Improvement and
The Evolution of Bleeding Management Algorithms
Keyvan Karkouti MD, MSc, FRCPC
University of Toronto | UHN, Sinai Health, Women’s College Hospital
Past, Present & Future of Bleeding Management MPOG
Disclosures
I have received research support and/or honoraria from Octapharma, Werfen, and CSL
Behring.
Copilot assist with slide design and refinement.
Past, Present & Future of Bleeding Management MPOG
Overview: the algorithm as a connecting thread in time
FOUNDATIONS
Erber and clinical
judgment
POC ERA
Rapid diagnosis; TACS;
targeted management
NEXT ALGORITHM
TACS II: accessible
precision
MPOG FUTURE
Measure, randomize, scale,
improve outcomes
From algorithm as guidance → algorithm as intervention → algorithm as learning system
3
Bleeding algorithms evolved from expert judgment to rapid phenotype-directed care. The
next advance is to make algorithms measurable, scalable, and randomizable to improve
outcomes.
Past
Present
Future
Past, Present & Future of Bleeding Management MPOG
Algorithms are interventions: they can improve or worsen care
The good
• Standardize care by embedding clinical guidelines
into workflows
• Reduce cognitive biases and diagnostic errors
• Improve efficiency and resource allocation by
prioritizing care on highest-risk patients
• Can improve outcomes
The bad
• Limited by the quality and representativeness of
evidence base
• Can perpetuate embedded biases and worsen
existing inequities
• Can constrain appropriate clinical judgment
• Benefits and harms/costs are rarely evaluated as
rigorously as drugs or devices
The impact of algorithms needs to be formally assessed.
4
Wang et al. SSRN Preprint (https://dx.doi.org/10.2139/ssrn.6149766)
Past, Present & Future of Bleeding Management MPOG
Algorithms may work, but how can we know without measuring?
What this figure captures
• Mortality rates before and after
hospital adoption of risk-stratification
algorithms for ACS
• >300,000 patients; US hospitals; 2013
to 2018
• Heterogenous effect (risk; race)
What the study could not solve
• Annual AHA hospital survey to
capture algorithm adoption
• High risk condition analogous to
major bleeding but better studied and
standardized
5
Wang et al. SSRN Preprint (https://dx.doi.org/10.2139/ssrn.6149766)
Algorithm impact on in
-hospital mortality
Years relative to adoption
PAST
From clinical judgment and delayed
assays to explicit decision pathways
Past, Present & Future of Bleeding Management MPOG
Erber 2002: an early algorithmic worldview
Core principles
• Estimated blood volume
replaced to guide therapy
• Secure hemostasis
• Maintain normovolemia and
oxygen delivery
Decision inputs
• Repeated PT/APTT, fibrinogen,
platelet count
• Clinical judgment remains
essential and priority
• No simple generally applicable
approach
Therapy model
• Correct assumed deficits in
relation to estimated blood
volume replaced
• Treat critical lab levels as
available
• Escalate (step-wise) in
persistent hemorrhage
7
Erber WN. Transfus Apher Sci. 2002;27:83-92
Past, Present & Future of Bleeding Management MPOG
Erber’s algorithm
8
Erber WN. Transfus Apher Sci. 2002;27:83-92
Past, Present & Future of Bleeding Management MPOG
Erber’s algorithm
9
Erber WN. Transfus Apher Sci. 2002;27:83-92
The algorithm, hindered by delayed labs, relied on an inherently unreliable
measure to guide clinical judgment.
Past, Present & Future of Bleeding Management MPOG
The 2002 thresholds: useful structure, important limitations
Fibrinogen
Goal >1.4 g/L; failure <1.0
PT/APTT
Goal <1.5×; failure >1.8×
Platelets
Goal >100×10⁹/L; failure <80
Strengths
• Explicit therapeutic targets
• Repeat testing
• Moving (slowly) away from empiric therapy
• Avoid prophylactic platelets and formula-based
FFP
• Step-wise escalation
Limitations
• Results arrived too late to be clinically
applicable
• Targets based on poor data
• Order of therapies based on historical practice
• TEG was available but required expertise and did
not quantify contributors to bleeding
10
Past, Present & Future of Bleeding Management MPOG
The 2002 thresholds: useful structure, important limitations
Fibrinogen
Goal >1.4 g/L; failure <1.0
PT/APTT
Goal <1.5×; failure >1.8×
Platelets
Goal >100×10⁹/L; failure <80
Strengths
• Explicit therapeutic targets
• Repeat testing
• Moving (slowly) away from empiric therapy
• Avoid prophylactic platelets and formula-based
FFP
• Step-wise escalation
Limitations
• Results arrived too late to be clinically
applicable
• Targets based on poor data
• Order of therapies based on historical practice
• TEG was available but required expertise and did
not quantify contributors to bleeding
11
Refractory bleeding was common, leading to the adoption of rFVIIa into clinical practice despite
high cost and potential (and later confirmed) serious risks.
Past, Present & Future of Bleeding Management MPOG
The algorithm problem in surgery
BLEEDING
Must be recognized
clinically
MECHANISM
Mixed surgical and
coagulopathic
TESTING
Results must arrive
before decisions
THERAPY
Identify and correct
relevant defects
REASSESS
Don’t over or under
treat
An algorithm succeeds only when recognition, testing, treatment, and
reassessment fit the workflow.
Why cardiac surgery is distinctive
• CPB changes platelets, fibrinogen, thrombin, fibrinolysis, and hemoglobin
• The phenotype can change within minutes
12
Clinicians must synthesize numerous inputs and make timely decisions in a highly dynamic setting
Past, Present & Future of Bleeding Management MPOG
Massive hemorrhage protocols for trauma: algorithm as QI
Recognize
Activate early
Mobilize
People + products
Resuscitate
Perfusion + oxygen
Control
Surgery / IR
Reassess
Labs + physiology
Standardization made care reproducible, auditable, and improvable.
But
• MHPs coordinate response
• A logistics algorithm is not the same as a diagnostic-treatment algorithm.
• They do not by themselves identify or target the dominant coagulation defect(s)
• Are designed to avoid primarily the lethal triad
13
Past, Present & Future of Bleeding Management MPOG
Back to surgery: continuum from fixed ratios to defect-directed care
Empiric rounds
Broad replacement
Ratio MHP
Standardized
packages
Lab algorithm
Threshold-directed
POC algorithm
Rapid phenotyping
and targeted Rx
Learning
algorithm
Adopted and
measured
The unresolved questions
• Which test?
• Which threshold?
• Which treatment sequence?
• How should severity modify the response?
• Can the algorithm be implemented reliably across sites?
• Are they cost-effective?
14
Uncontrolled bleeding Controlled bleeding
Scale
Past/Present
Future
PRESENT
Point-of-care algorithms: faster
phenotyping, targeted therapy,
implementation challenges
Past, Present & Future of Bleeding Management MPOG
Raphael 2019: algorithm as a bedside implementation tool
Why it was needed
• Guideline adherence remained
poor
• Practice variation persisted
• POC devices and factor
concentrates were not
universally available
What it added
• A succinct graphical advisory
• Separate VHA and conventional-
test pathways
• Stepwise escalation after
heparin reversal
Key discipline
• Treat excessive microvascular
bleeding
• Assess platelets, fibrinogen, and
factor deficiency
• Reassess after each treatment
round
The algorithm brought guidance to the point of care.
16
Raphael et al. Anesth Analg 2019;129:1209-1221
Past, Present & Future of Bleeding Management MPOG
Raphael: VHA-guided pathway
Design features
• Bleeding is the entry
condition
• Correct heparin effect and
physiology
• Use assay-specific branches
for heparin, platelets,
fibrinogen, factors, and
fibrinolysis
• Reassess after therapy
Implementation tension
• Comprehensive
• Rapid
• But cognitively and
technologically demanding
17
Raphael et al. Anesth Analg 2019;129:1209-1221
Past, Present & Future of Bleeding Management MPOG
Raphael: a conventional-test alternative
Why this matters
• The advisory explicitly
anticipated hospitals without
VHA
• The underlying targets remain
the same: platelets, plasma,
cryoprecipitate, heparin, and
physiology
Persistent limitation
• Conventional assays are
familiar but may arrive too
late unless sampling and
turnaround are engineered
into workflow
18
Raphael et al. Anesth Analg 2019;129:1209-1221
Past, Present & Future of Bleeding Management MPOG
TACS: test the algorithm as a QI intervention
19
Karkouti et al. Circulation 2016;134:1152-1162
Past, Present & Future of Bleeding Management MPOG
TACS: test the algorithm as a QI intervention
12 hospitals
Consecutive CPB
cases
Stepped wedge
2 hospitals cross
monthly
Integrated
testing
ROTEM + platelet
function
Clinical
algorithm
Defect-directed
treatment
Usual workflow
Pragmatic delivery
7,402 88.7% RR 0.83
patients
POC testing in intervention
phase
major bleeding
20
Karkouti et al. Circulation 2016;134:1152-1162
Past, Present & Future of Bleeding Management MPOG
TACS: meaningful effect, incomplete solution
RBC transfusion
RR 0.91
Platelet transfusion
RR 0.77
Major bleeding
RR 0.83
Plasma transfusion
RR 0.98
The next algorithm must preserve targeting while improving usability, accessibility,
and treatment alignment.
21
Karkouti et al. Circulation 2016;134:1152-1162
Despite reproducible benefits, there is limited clinical uptake.
Past, Present & Future of Bleeding Management MPOG
Why POC algorithms have not scaled successfully
Technology
• Capital and cartridge cost
• Specialized equipment
• Quality assurance
Cognition
• Multiparametric outputs
• Device-specific thresholds
• Treating numbers rather than
bleeding
Implementation
• Training and turnover
• Workflow integration
• Reversion to empiricism
Efficacy without usability does not become reliable care.
24
Past, Present & Future of Bleeding Management MPOG
Having an algorithm does not ensure timely, appropriate treatment
What this figure captures
• Subgroup of ITACTIC study patients
who had both CCT and VHA
measured (n=133)
• Only 54% received appropriate goal-
directed therapy and < 25% had
coagulopathy corrected
What it could not solve
25
Lindsay et al. Anesthesiology 2024;141:904-912
• Maybe algorithms do not work
• Unclear if due to delay in treatment
(68 min for VHA; 110 min for CCT) or
ineffectiveness of algorithm or
treatment
Proportion of patients who had corrected coagulopathy
by the 12
th
unit of RBC transfusion
(CCT = conventional; VHA = viscoelastic)
Past, Present & Future of Bleeding Management MPOG
Contemporary practice remains heterogeneous
What this figure captures
• Variable treatment thresholds sheds
light on presence of practice
variability
• 19% do not consider fibrinogen levels
for bleeding management
• 42% used POC testing; 30% standard
labs; 28% no algorithm
What it could not solve
26
Levy et al. JTH 2025;116:3871-3882
• 2023 snapshot of 1030 ASA members
• Low response rate (4% of ASA
members)
Considerations for an adequate level of fibrinogen in
a bleeding patient
Past, Present & Future of Bleeding Management MPOG
Transfusion practice is shifting, but not evenly
What this figure captures
• Annual transfusion rates (2016 – 2023) at
16 Dutch cardiac surgery centres
• Increasing annual usage
• Consistent with increasing adoption of
factor-concentrate-based strategies
What it could not solve
27
Schaap et al. JCVA 2026; In Press
Fibrinogen conc.
1.15*
PCC
1.10*
Platelets
1.09*
Plasma
1.08*
RBCs
1.02*
1.00 1.05 1.10 1.15 1.20
Adjusted OR per year · * all p < 0.001
• May not be applicable to NA practice
• Why usage is increasing for all
products (except rFVIIa) despite PBM
*All ORs adjusted for age, sex, procedure type, and year.
Past, Present & Future of Bleeding Management MPOG
Same guidelines. Vastly different practices. Heterogeneity persists.
What this figure captures
• Variability lower than previous
reports, but remains high
• E.g., RBC use varied by >2-fold across
centres sharing the same national
guidelines
What it could not solve
28
• Cause of the persistent variability
• How much of it represents
inappropriate transfusions
BLOOD PRODUCT NATIONAL
RATE
CENTRE VARIABILITY
Red Blood Cells 30.6%
17.6 – 39.4%
Platelets 20.3%
6.1 – 34.8%
Fibrinogen Concentrate 13.5%
1.8 – 28.6%
Plasma 11.0%
0.5 – 29.4%
PCC 7.4%
0.3 – 25.5%
rFVIIa 0.5%
0.0 – 2.9%
Schaap et al. JCVA 2026; In Press
VARIABILITY
28% T0 31%
14% TO 22%
11% TO 15%
4% TO 11%
4% TO 8%
0.3% TO 0.1%
TEMPORAL
CHANGE
Past, Present & Future of Bleeding Management MPOG
Unnecessary transfusions have consequences
What this figure captures
• RBC carries greater risk than non-RBC
• Combined RBC and non-RBC
exposure identifies a high-risk
population and supports testing
strategies that reduce both
29
Ad et al. ATS 2022;113:748-756
Past, Present & Future of Bleeding Management MPOG
Algorithms can reduce variability. Yes, even surgeons.
What this figure captures
• 29 hospitals; 80 surgeons; >35,000
elective cardiac surgeries
• Pre/post study; 2009 to 2015; national
PBM guidelines published in 2012
• 24% reduction in RBC transfusions
• 37% reduction in variation
What it could not solve
30
Irving et al. Vox Sang 2025;120:47-54
• Applicability to NA
• Causes of variability not explored
• Did not account for other providers,
reasons for transfusion, time trends
Past, Present & Future of Bleeding Management MPOG
The evidence changed the treatment side of the algorithm
FIBRES
Concentrate noninferior to
cryoprecipitate
FARES-II
PCC superior to plasma
strategy
CHIPS
Cold-stored platelets
noninferior
TXA trials
Indication and timing
matter
A modern algorithm should not pair new diagnostics with old product defaults.
Design implications
• Testing sequence and treatment sequence must evolve together
• Clinically important outcomes, logistics, and cost all matter
31
Past, Present & Future of Bleeding Management MPOG
Callum et al. JAMA 2019;322:1966-1976
FIBRES: fibrinogen concentrate vs cryoprecipitate
Patient recruitment completed in ~1.75 years (Feb 17 to Nov 18)
Fibrinogen concentrate is noninferior (and superior in a priori subgroups) to cryoprecipitate and offers meaningful logistical advantages: rapid preparation,
standardized dosing, no ABO compatibility required, and reduced infectious risk. Demonstrated that treatment logistics can be improved without increasing transfusion.
Past, Present & Future of Bleeding Management MPOG
Karkouti et al. JAMA 2025;333:1781-1792
Patient recruitment completed in ~1.5 years (Nov 22 to May 24)
FARES II: PCC vs FP
PCC demonstrated SUPERIOR hemostatic efficacy AND safety versus frozen plasma — a paradigm-changing result for factor replacement in cardiac surgery bleeding.
The result supports redesigning the treatment side of cardiac-surgery bleeding algorithms.
Past, Present & Future of Bleeding Management MPOG
The next algorithm must solve four implementation problems
TIMING
Results ready when
bleeding starts
INTERPRETATION PRIORITY
Identify and treat the
dominant defect(s) first
SCALABILITY
Low capital and
operating costs
And preserve clinical nuance
• Treat only clinically significant bleeding
• Treat multiple abnormalities concurrently when bleeding is severe or massive
• Prioritizes factor concentrates over allogeneic products
• Incorporates VHA if available for persistent bleeding
34
Few familiar inputs aligned
with clinical severity
NEXT
TACS II: an accessible precision algorithm
designed for pragmatic evaluation
Past, Present & Future of Bleeding Management MPOG
TACS II: familiar assays, engineered timing
Near end of CPB
CBC + Clauss
fibrinogen at 36°C
After reversal
POC INR ≥10
minutes after
protamine
Clinical trigger
Visual bleeding
scale ≥2 after 5 min
packing
Initial treatment
Factor concentrates
prioritized
If bleeding
continues
Repeat assays +
VHA if available
The innovation is not the assays. It is how to better use the tools we already have.
It is timing + severity + priority + reassessment + escalation.
Status
• Proposed algorithm undergoing stepwise evaluation
• Not yet a validated guideline
36
Karkouti et al. BJA 2026; Accepted for publication
Past, Present & Future of Bleeding Management MPOG
TACS II algorithm
37
Past, Present & Future of Bleeding Management MPOG
How to read TACS II: three decisions, one stop rule
1. Should I treat?
• Clinically significant bleeding
• Persists after packing
• Do not treat an abnormal test
alone
2. What first?
• Most severe category first
• Within a row: fibrinogen → INR →
platelets
• Concurrent correction for
severe/massive bleeding
3. What next?
• Reassess after each therapy
• Repeat assays if bleeding
continues
• Use VHA if available if bleeding
continues after first-line therapy
Stop treating when bleeding stops.
38
Past, Present & Future of Bleeding Management MPOG
TACS II evaluation pathway
VALIDATE
Assay timing +
bleeding trigger
IMPLEMENT
Single-centre
adherence and
usability
MEASURE
Transfusion,
bleeding, cost,
safety
RANDOMIZE
Multicentre cluster
trial
SCALE
Feedback into QI
Why MPOG fits
• Cardiac procedures and physiologic data
• Medication and blood-product exposure data
• Site-level implementation and fidelity
• Clinical outcomes and reoperation data
39
FUTURE
From a single-centre algorithm to a registry-
embedded policy trial designed to reduce
unnecessary allogeneic transfusion and improve
outcomes
Past, Present & Future of Bleeding Management MPOG
Advances in diagnostics, interventions, and clinical care
Diagnostics Interventions Clinical management
Arguably, the latter will have the greatest positive impact, and MPOG is uniquely
positioned to make it possible.
41
POC, whole-blood thrombography
Synthetic platelets
Learning-health-system loop
Past, Present & Future of Bleeding Management MPOG
VEGA-2 provides the operating model
Policy
• Phenylephrine vs
norepinephrine as first-line
vasopressor
• Accepted treatments become
site-period defaults
Design
• Pragmatic multiple-period
cluster-randomized crossover
• Clinical override preserved
Data
• Eligibility, exposure, fidelity,
covariates, and outcomes from
routine EHR data
TACS II could use the same logic: randomize the algorithm as a site-period policy.
42
Past, Present & Future of Bleeding Management MPOG
A VEGA-like PBM algorithm trial: the core design
UNIT
Hospital-period
policy
INTERVENTION
PBM intervention
COMPARATOR
Current site practice
/ accepted standard
CROSSOVER
Alternating periods
ANALYSIS
Cluster-period +
patient-level
outcomes
Pragmatic safeguards
• Clinical override always available
• Protocol fidelity measured, not assumed
• Implementation toolkit standardized
• Contamination and carryover assessed
Candidate outcomes
• Clinically significant bleeding
• Blood-component exposure
• Factor-concentrate use
• Re-exploration, AKI, serious adverse events
• Cost-effectiveness
43
Past, Present & Future of Bleeding Management MPOG
What MPOG would need to capture
ELIGIBILITY
Cardiac surgery + CPB + site-period policy
TRIGGER
Bleeding severity or a validated proxy
EXPOSURE
Assays, timing, treatment sequence
FIDELITY
Did care follow the assigned pathway?
OUTCOMES
Products, bleeding, reoperation, organ injury, costs
44
Past, Present & Future of Bleeding Management MPOG
The enabling asset: a validated perioperative bleeding phenotype
Direct signal
credible EBL • reoperation
• procedural hemostasis
Physiology
hemoglobin trajectory •
instability • shock
Response
TXA • RBC • platelets •
fibrinogen • factors
Resources
cell salvage • ICU •
complications
A transfusion count alone is not a bleeding phenotype.
45
Past, Present & Future of Bleeding Management MPOG
Phenotype validation: the method must be transportable
Validation
• Procedure-specific chart review
• Transportability across sites
• Sensitivity to practice culture
• Ability to distinguish bleeding from treatment
preference
46
An et al. RPTH 2026;10:106616
What this figure captures
• Chart review versus HER dentification of
ISTH defined major bleeding and CRNM in
433 patients on OAC
• Major bleeding: Sn 91%, Sp 92%, PPV 91%,
NPV 92% for major bleeding
• CRNM: Sn 66%, Sp 86%, PPV 52%, NPV
92%
CRNM = clinically relevant nonmajor bleeding
What MPOG measures today: transfusion decisions, not bleeding
Past, Present & Future of Bleeding Management MPOG
TRAN-01 / TRAN-02
• Audit whether hemoglobin was
measured before transfusion
• Assess potential over-
transfusion
• Focus on the hemoglobin as the
driver of decision to transfuse
Important gap
• Transfusion is influenced by
clinician and institutional
decisions
• These measures do not directly
quantify bleeding
• They capture a moment in time
rather the dynamic nature of
surgery that influence treatment
decisions
Future State
• Capture and validate PBM-
related variables
• Include therapies and
interventions (cell salvage)
• Strengthen routine linkage with
STS and NSQIP where available
• Use registry-linked outcomes to
validate phenotype performance
across sites
The present MPOG framework audits transfusion. The next framework should
measure the bleeding pathway.
A discussion point
Past, Present & Future of Bleeding Management MPOG
Can bleeding management become a flagship use case for the
next phase of MPOG?
Observational data identify uncertainty.
QI shows where variation persists.
Pragmatic trials determine which strategy should be used.
What methodological work is required to enable more patient-centred
outcomes that can advance both QI and research in bleeding management
to become a learning-health-system loop?
Past, Present & Future of Bleeding Management MPOG
From trial back to QI: the learning-health-system loop
CARE
Algorithm used in
routine workflow
DATA
Phenotype + exposure +
fidelity
QI
Benchmark and
feedback
TRIAL
Randomize residual
uncertainty
EVIDENCE
Update the algorithm
The algorithm becomes a versioned, testable clinical product.
49
Past, Present & Future of Bleeding Management MPOG
A proposed MPOG bleeding program
1. DEFINE
Validate bleeding severity and outcome phenotypes
2. BENCHMARK
Map algorithms, assays, products, and variation
3. IMPLEMENT
Deploy bleeding management algorithm toolkit at study sites
4. RANDOMIZE
Cluster-crossover effectiveness trial
5. SCALE
Convert results into automated QI
50
Past, Present & Future of Bleeding Management MPOG
Takeaways
1
Algorithms are the thread in the evolution of bleeding management.
2
POC algorithms improved targeting, but implementation and usability remain limiting.
3
TACS II redesigns timing, severity, treatment priority, and stop rules.
4
MPOG could test TACS II-like algorithms as a VEGA-like cluster-crossover policy trial.
5
A validated bleeding phenotype and measurable fidelity are prerequisites.
51
Thank you
Past, Present & Future of Bleeding Management MPOG
