Top Banner
Introduction Materials Results Conclusions Acknowledgements Severe Sepsis Mortality Prediction with Relevance Vector Machines Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain August 30, 2011 Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain Severe Sepsis Mortality Prediction with Relevance Vector Machines
42

Rvm sepsis mortality_index_beamer

Aug 28, 2014

Download

Health & Medicine

Sepsis is a transversal pathology and one of the main causes of death at the Intensive Care Unit (ICU). It has in fact become the tenth most common cause of death in western societies. Its mortality rates can reach up to 45.7\% for septic shock, its most acute manifestation. For these reasons, the prediction of the mortality caused by sepsis is an open and relevant medical research challenge. This problem requires prediction methods that are robust and accurate, but also readily interpretable. This is paramount if they are to be used in the demanding context of real-time decision making at the ICU. In this brief paper, such a method is presented. It is based on a variant of the well-known support vector machine (SVM) model and provides an automated ranking of relevance of the mortality predictors. The reported results show that it outperforms in terms of accuracy alternative techniques currently in use, while simultaneously assessing the relative impact of individual pathology indicators.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Severe Sepsis Mortality Prediction withRelevance Vector Machines

Vicent J. Ribas

SOCO Research GroupTechical University of Catalonia (UPC)

Barcelona, Spain

August 30, 2011

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 2: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Contents

1 Introduction

2 Materials

3 Results

4 Conclusions

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 3: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Contents

1 Introduction

2 Materials

3 Results

4 Conclusions

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 4: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Introduction

Sepsis is a clinical syndrome defined by the presence of bothinfection and Systemic Inflammatory Response Syndrome(SIRS).

This can lead to severe sepsis (organ dysfunction) or to septicshock (severe sepsis with hypotension refractory to fluidadministration) and multiorgan failure.

In western countries, septic patients account for as much as25% of ICU bed utilization and the pathology occurs in 1% -2% of all hospitalizations.

The mortality rates of sepsis range from 12.8% for sepsis and20.7% for severe sepsis, and up to 45.7% for septic shock.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 5: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Introduction

Sepsis is a clinical syndrome defined by the presence of bothinfection and Systemic Inflammatory Response Syndrome(SIRS).

This can lead to severe sepsis (organ dysfunction) or to septicshock (severe sepsis with hypotension refractory to fluidadministration) and multiorgan failure.

In western countries, septic patients account for as much as25% of ICU bed utilization and the pathology occurs in 1% -2% of all hospitalizations.

The mortality rates of sepsis range from 12.8% for sepsis and20.7% for severe sepsis, and up to 45.7% for septic shock.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 6: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Introduction

Sepsis is a clinical syndrome defined by the presence of bothinfection and Systemic Inflammatory Response Syndrome(SIRS).

This can lead to severe sepsis (organ dysfunction) or to septicshock (severe sepsis with hypotension refractory to fluidadministration) and multiorgan failure.

In western countries, septic patients account for as much as25% of ICU bed utilization and the pathology occurs in 1% -2% of all hospitalizations.

The mortality rates of sepsis range from 12.8% for sepsis and20.7% for severe sepsis, and up to 45.7% for septic shock.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 7: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Introduction

Sepsis is a clinical syndrome defined by the presence of bothinfection and Systemic Inflammatory Response Syndrome(SIRS).

This can lead to severe sepsis (organ dysfunction) or to septicshock (severe sepsis with hypotension refractory to fluidadministration) and multiorgan failure.

In western countries, septic patients account for as much as25% of ICU bed utilization and the pathology occurs in 1% -2% of all hospitalizations.

The mortality rates of sepsis range from 12.8% for sepsis and20.7% for severe sepsis, and up to 45.7% for septic shock.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 8: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The medical management of sepsis and the study of itsprognosis and outcome is a relevant medical researchchallenge.

Provided that such methods are to be used in a clinicalenvironment (ICU), it requires prediction methods that arerobust, accurate and readily interpretable.

Here we present a method based on Relevance VectorMachines (RVM) and compare it with other well establishedshrinkage methods.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 9: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The medical management of sepsis and the study of itsprognosis and outcome is a relevant medical researchchallenge.

Provided that such methods are to be used in a clinicalenvironment (ICU), it requires prediction methods that arerobust, accurate and readily interpretable.

Here we present a method based on Relevance VectorMachines (RVM) and compare it with other well establishedshrinkage methods.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 10: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The medical management of sepsis and the study of itsprognosis and outcome is a relevant medical researchchallenge.

Provided that such methods are to be used in a clinicalenvironment (ICU), it requires prediction methods that arerobust, accurate and readily interpretable.

Here we present a method based on Relevance VectorMachines (RVM) and compare it with other well establishedshrinkage methods.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 11: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Contents

1 Introduction

2 Materials

3 Results

4 Conclusions

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 12: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Materials

A prospective observational cohort study of adult patients withsevere sepsis was conducted at the Critical Care Departmentof the Vall d’ Hebron University Hospital (Barcelona, Spain).

Data from 354 patients with severe sepsis was collected atthis ICU between June, 2007 and December, 2010.

The mean age of the patients in the database was 57.08 (withstandard deviation ±16.65) years.

40% of patients were female.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 13: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Materials

A prospective observational cohort study of adult patients withsevere sepsis was conducted at the Critical Care Departmentof the Vall d’ Hebron University Hospital (Barcelona, Spain).

Data from 354 patients with severe sepsis was collected atthis ICU between June, 2007 and December, 2010.

The mean age of the patients in the database was 57.08 (withstandard deviation ±16.65) years.

40% of patients were female.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 14: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Materials

A prospective observational cohort study of adult patients withsevere sepsis was conducted at the Critical Care Departmentof the Vall d’ Hebron University Hospital (Barcelona, Spain).

Data from 354 patients with severe sepsis was collected atthis ICU between June, 2007 and December, 2010.

The mean age of the patients in the database was 57.08 (withstandard deviation ±16.65) years.

40% of patients were female.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 15: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Materials

A prospective observational cohort study of adult patients withsevere sepsis was conducted at the Critical Care Departmentof the Vall d’ Hebron University Hospital (Barcelona, Spain).

Data from 354 patients with severe sepsis was collected atthis ICU between June, 2007 and December, 2010.

The mean age of the patients in the database was 57.08 (withstandard deviation ±16.65) years.

40% of patients were female.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 16: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The collected data show the worst values for all variablesduring the first 24 hours of evolution of Severe Sepsis.

Organ dysfunction was evaluated by means of the SOFA scoresystem, which objectively measures organ dysfunction for 6organs/systems.

Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14)Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86)

Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs (SOFA 1-2) 1.68 (1.09)Hepatic (HEPA) 0.48 (0.92) Failure Organs (SOFA 3-4) 1.51 (1.02)Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32)

Severity was evaluated by means of the APACHE II score,which was 23.03 ± 9.62 for the population under study.

The mortality rate intra-ICU for our study population was29.44%.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 17: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The collected data show the worst values for all variablesduring the first 24 hours of evolution of Severe Sepsis.

Organ dysfunction was evaluated by means of the SOFA scoresystem, which objectively measures organ dysfunction for 6organs/systems.

Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14)Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86)

Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs (SOFA 1-2) 1.68 (1.09)Hepatic (HEPA) 0.48 (0.92) Failure Organs (SOFA 3-4) 1.51 (1.02)Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32)

Severity was evaluated by means of the APACHE II score,which was 23.03 ± 9.62 for the population under study.

The mortality rate intra-ICU for our study population was29.44%.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 18: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The collected data show the worst values for all variablesduring the first 24 hours of evolution of Severe Sepsis.

Organ dysfunction was evaluated by means of the SOFA scoresystem, which objectively measures organ dysfunction for 6organs/systems.

Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14)Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86)

Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs (SOFA 1-2) 1.68 (1.09)Hepatic (HEPA) 0.48 (0.92) Failure Organs (SOFA 3-4) 1.51 (1.02)Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32)

Severity was evaluated by means of the APACHE II score,which was 23.03 ± 9.62 for the population under study.

The mortality rate intra-ICU for our study population was29.44%.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 19: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The collected data show the worst values for all variablesduring the first 24 hours of evolution of Severe Sepsis.

Organ dysfunction was evaluated by means of the SOFA scoresystem, which objectively measures organ dysfunction for 6organs/systems.

Cardiovascular (CV) 2.86 (1.62) Haematologic (HAEMATO) 0.78 (1.14)Respiratory (RESP) 2.31 (1.15) Global SOFA score 7.94 (3.86)

Central Nerv. Sys. (CNS) 0.48 (1.00) Dysf. Organs (SOFA 1-2) 1.68 (1.09)Hepatic (HEPA) 0.48 (0.92) Failure Organs (SOFA 3-4) 1.51 (1.02)Renal (REN) 1.06 (1.20) Total Dysf. Organs 3.18 (1.32)

Severity was evaluated by means of the APACHE II score,which was 23.03 ± 9.62 for the population under study.

The mortality rate intra-ICU for our study population was29.44%.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 20: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

List of attributes used in this study:Age

Gender

Sepsis Focus

Germ Class

Polimicrobial Infection

Base Pathology

Cardiovascular SOFA Score

Respiratory SOFA Score

CNS SOFA score

Hepatic SOFA Score

Renal SOFA Score

Haematologic SOFA Score

Total SOFA Score

Dysfunctioning Organs forSOFA 1-2

Dysfunctioning Organs forSOFA 3-4

Total Number ofDysfunctioning Organs

Mechanical Ventilation

Oxygenation IndexPaO2/FiO2

Vasoactive Drugs

Platelet Count

APACHE II Score

Surviving Sepsis CampaignBundles 6h

Haemocultures 6h

Antibiotics 6h

Volume 6h

O2 Central Venous Saturation 6h

Haematocrit 6h

Transfusions 6h

Dobutamine 6h

Surviving Sepsis CampaignBundles 24h

Glycaemia 24h

PPlateau

Worst Lactate

O2 Central Venous Saturation

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 21: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Contents

1 Introduction

2 Materials

3 Results

4 Conclusions

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 22: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Mortality Prediction with RVM

The model performance was evaluated by means of 10-FoldCross-Validation.

The RVM yielded an accuracy of mortality prediction of 0.80;a prediction error of 0.24; a sensitivity of 0.66; and aspecificity of 0.80.

RVM selected the following attributes (corresponding toweights):

Number of dysfunctioning organs (w1 = −0.039)Mechanical Ventilation (w2 = −0.101)APACHE II (w3 = −0.337)Resuscitation Bundles (6h) (w4 = 0.037)

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 23: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Mortality Prediction with RVM

The model performance was evaluated by means of 10-FoldCross-Validation.

The RVM yielded an accuracy of mortality prediction of 0.80;a prediction error of 0.24; a sensitivity of 0.66; and aspecificity of 0.80.

RVM selected the following attributes (corresponding toweights):

Number of dysfunctioning organs (w1 = −0.039)Mechanical Ventilation (w2 = −0.101)APACHE II (w3 = −0.337)Resuscitation Bundles (6h) (w4 = 0.037)

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 24: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Mortality Prediction with RVM

The model performance was evaluated by means of 10-FoldCross-Validation.

The RVM yielded an accuracy of mortality prediction of 0.80;a prediction error of 0.24; a sensitivity of 0.66; and aspecificity of 0.80.

RVM selected the following attributes (corresponding toweights):

Number of dysfunctioning organs (w1 = −0.039)Mechanical Ventilation (w2 = −0.101)APACHE II (w3 = −0.337)Resuscitation Bundles (6h) (w4 = 0.037)

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 25: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The coefficients corresponding to the rest of attributes wereset to values close to zero as part of the training process. Thisreduces the number of attributes (34 to just 4) and improvesits interpretability.

The negative weights (number of dysfunctioning organs,mechanical ventilation, APACHE II) are related to a highermortality risk.

The SSC bundles (resuscitation bundles) are associated to aprotective effect (i.e. antibiotics administration, performanceof haemocultures, administration of volume and vasoactivedrugs and so on).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 26: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The coefficients corresponding to the rest of attributes wereset to values close to zero as part of the training process. Thisreduces the number of attributes (34 to just 4) and improvesits interpretability.

The negative weights (number of dysfunctioning organs,mechanical ventilation, APACHE II) are related to a highermortality risk.

The SSC bundles (resuscitation bundles) are associated to aprotective effect (i.e. antibiotics administration, performanceof haemocultures, administration of volume and vasoactivedrugs and so on).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 27: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The coefficients corresponding to the rest of attributes wereset to values close to zero as part of the training process. Thisreduces the number of attributes (34 to just 4) and improvesits interpretability.

The negative weights (number of dysfunctioning organs,mechanical ventilation, APACHE II) are related to a highermortality risk.

The SSC bundles (resuscitation bundles) are associated to aprotective effect (i.e. antibiotics administration, performanceof haemocultures, administration of volume and vasoactivedrugs and so on).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 28: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Comparison with Shrinkage Feature Selection Methods forLogistic Regression

The predictive ability of the RVM has also been compared tothat of other well established shrinkage methods for regression.

The selected attributes and coefficients for each method were:

Ridge Regression:

Number of dysfunctioningorgans for SOFA 3-4(w1 = −0.021)

APACHE II (w2 = −0.127)

Worst Lactate(w3 = −0.126).

Lasso:

Age (w1 = 0.007)

Germ Class (w2 = 0.005)

PaO2/FiO2 (w3 = 0.001)

APACHE II (w4 = −0.006)

SvcO2 6h (w5 = −0.001)

Haematocrit 6h(w6 = 0.009)

Worst Lactate(w7 = −0.023)

SvcO2 (w8 = −0.006).

Logistic Regression with backwardfeature selection:

Intercept (w1 = 4.16)

Number of DysfunctioningOrgans (w1 = −0.57)

APACHE II (w2 = −0.09)

Worst Lactate (w3 = −0.30)

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 29: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Comparison with Shrinkage Feature Selection Methods forLogistic Regression

The predictive ability of the RVM has also been compared tothat of other well established shrinkage methods for regression.

The selected attributes and coefficients for each method were:

Ridge Regression:

Number of dysfunctioningorgans for SOFA 3-4(w1 = −0.021)

APACHE II (w2 = −0.127)

Worst Lactate(w3 = −0.126).

Lasso:

Age (w1 = 0.007)

Germ Class (w2 = 0.005)

PaO2/FiO2 (w3 = 0.001)

APACHE II (w4 = −0.006)

SvcO2 6h (w5 = −0.001)

Haematocrit 6h(w6 = 0.009)

Worst Lactate(w7 = −0.023)

SvcO2 (w8 = −0.006).

Logistic Regression with backwardfeature selection:

Intercept (w1 = 4.16)

Number of DysfunctioningOrgans (w1 = −0.57)

APACHE II (w2 = −0.09)

Worst Lactate (w3 = −0.30)

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 30: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The three shrinkage methods evaluated in this section agreedin detecting as prognostic factors the Severity measured bythe APACHE II score and acidosis.

Organ dysfunction and mechanical ventilation or otherparameters related to it like PaO2/FiO2 also play a role in theprognosis of Sepsis.

The accuracy of each method was the following:

Method AUC Error Rate Sens. Spec.RVM 0.80 0.24 0.66 0.80

Logistic 0.77 0.27 0.66 0.76Ridge 0.69 0.28 0.67 0.73Lasso 0.70 0.32 0.67 0.68

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 31: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The three shrinkage methods evaluated in this section agreedin detecting as prognostic factors the Severity measured bythe APACHE II score and acidosis.

Organ dysfunction and mechanical ventilation or otherparameters related to it like PaO2/FiO2 also play a role in theprognosis of Sepsis.

The accuracy of each method was the following:

Method AUC Error Rate Sens. Spec.RVM 0.80 0.24 0.66 0.80

Logistic 0.77 0.27 0.66 0.76Ridge 0.69 0.28 0.67 0.73Lasso 0.70 0.32 0.67 0.68

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 32: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

The three shrinkage methods evaluated in this section agreedin detecting as prognostic factors the Severity measured bythe APACHE II score and acidosis.

Organ dysfunction and mechanical ventilation or otherparameters related to it like PaO2/FiO2 also play a role in theprognosis of Sepsis.

The accuracy of each method was the following:

Method AUC Error Rate Sens. Spec.RVM 0.80 0.24 0.66 0.80

Logistic 0.77 0.27 0.66 0.76Ridge 0.69 0.28 0.67 0.73Lasso 0.70 0.32 0.67 0.68

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 33: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Comparison with the APACHE II Mortality Score

The Risk-of-Death (ROD) formula based on the APACHE IIscore can be expressed as:

ln

(ROD

1 − ROD

)= −3.517 + 0.146 · A + ε

where A is the APACHE II score and ε is a correction factordepending on clinical traits at admission in the ICU.

The application of this ROD formula to the population understudy yields an error rate of 0.28 (higher than RVM), asensitivity of 0.55 (very low) and a specificity of 0.80. TheAUC was 0.70 (lower than RVM).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 34: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Comparison with the APACHE II Mortality Score

The Risk-of-Death (ROD) formula based on the APACHE IIscore can be expressed as:

ln

(ROD

1 − ROD

)= −3.517 + 0.146 · A + ε

where A is the APACHE II score and ε is a correction factordepending on clinical traits at admission in the ICU.

The application of this ROD formula to the population understudy yields an error rate of 0.28 (higher than RVM), asensitivity of 0.55 (very low) and a specificity of 0.80. TheAUC was 0.70 (lower than RVM).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 35: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Contents

1 Introduction

2 Materials

3 Results

4 Conclusions

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 36: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Conclusions

In the assessment of ROD for critically ill patients, sensitivityis important due to the fact that more aggressive treatmentand therapeutic actions may result in better outcomes for highrisk patients.

We have put forward an RVM-based method for theprediction of ROD in septic patients, which has been shown toproduce accurate, specific results, while improving theinterpretability and actionability of the results.

This method has proven to be superior in terms of accuracy(error rate, specificity and AUC) than other well establishedshrinkage methods (Lasso and Ridge).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 37: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Conclusions

In the assessment of ROD for critically ill patients, sensitivityis important due to the fact that more aggressive treatmentand therapeutic actions may result in better outcomes for highrisk patients.

We have put forward an RVM-based method for theprediction of ROD in septic patients, which has been shown toproduce accurate, specific results, while improving theinterpretability and actionability of the results.

This method has proven to be superior in terms of accuracy(error rate, specificity and AUC) than other well establishedshrinkage methods (Lasso and Ridge).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 38: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

Conclusions

In the assessment of ROD for critically ill patients, sensitivityis important due to the fact that more aggressive treatmentand therapeutic actions may result in better outcomes for highrisk patients.

We have put forward an RVM-based method for theprediction of ROD in septic patients, which has been shown toproduce accurate, specific results, while improving theinterpretability and actionability of the results.

This method has proven to be superior in terms of accuracy(error rate, specificity and AUC) than other well establishedshrinkage methods (Lasso and Ridge).

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 39: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

From a medical viewpoint, this study shows that it is possibleto derive a reliable prognostic score from a parsimonious set ofphysiopathologic and therapeutic variables, which are availableat the onset of severe sepsis for medical experts at the ICU.

The method may be understood as a generalization of theROD formula which takes not only the contribution of theAPACHE II score into consideration, but also other importantlife-threatening clinical traits

The performance of the proposed method has been evaluatedin a single ICU and a limited population sample. Future workshould lead towards a multi-centric prospective study, in orderto validate its generalizability.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 40: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

From a medical viewpoint, this study shows that it is possibleto derive a reliable prognostic score from a parsimonious set ofphysiopathologic and therapeutic variables, which are availableat the onset of severe sepsis for medical experts at the ICU.

The method may be understood as a generalization of theROD formula which takes not only the contribution of theAPACHE II score into consideration, but also other importantlife-threatening clinical traits

The performance of the proposed method has been evaluatedin a single ICU and a limited population sample. Future workshould lead towards a multi-centric prospective study, in orderto validate its generalizability.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 41: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

From a medical viewpoint, this study shows that it is possibleto derive a reliable prognostic score from a parsimonious set ofphysiopathologic and therapeutic variables, which are availableat the onset of severe sepsis for medical experts at the ICU.

The method may be understood as a generalization of theROD formula which takes not only the contribution of theAPACHE II score into consideration, but also other importantlife-threatening clinical traits

The performance of the proposed method has been evaluatedin a single ICU and a limited population sample. Future workshould lead towards a multi-centric prospective study, in orderto validate its generalizability.

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines

Page 42: Rvm sepsis mortality_index_beamer

Introduction Materials Results Conclusions Acknowledgements

THANK YOU

Vicent J. Ribas SOCO Research Group Techical University of Catalonia (UPC) Barcelona, Spain

Severe Sepsis Mortality Prediction with Relevance Vector Machines