M 14: BR 14: 175 BR ,

Use of complexity modeling of physiology signals in real time to predict cardiorespiratory instability Michael R. Pinsky, MD, Dr hc Department of Crit...
Author: Nicholas Blair
2 downloads 4 Views 3MB Size
Use of complexity modeling of physiology signals in real time to predict cardiorespiratory instability Michael R. Pinsky, MD, Dr hc Department of Critical Care Medicine University of Pittsburgh

Monitoring Truth No monitoring device, no matter how accurate or insightful its data will improve outcome outcome, Unless coupled to a treatment, which itself improves outcome Pinsky & Payen. Functional Hemodynamic Monitoring Monitoring,, pp 11-4, 2004 Pinsky & Payen. Crit Care 9:566 9:566--72 2005 Pinsky. Chest 123:2020123:2020-9, 2007

Three Primary Clinical Problems

Medical Issues • Identify circulatory insufficiency before secondary tissue injury occurs • Assess disease severity • Accurately predict response to treatment • Gauge adequacy of specific therapies • Estimate improved predictions of disease severity by additional biological measures (biosensors (biosensors)) • Need to develop metrics to assess these challenges

• Shock represents the expression of inadequate tissue perfusion • Four primary mechanisms result in shock – Hypovolemic, H l i cardiogenic, di i inflammatory, i fl t neurogenic i

• The measured physiological variables reflect the interaction between the primary dysfunction and the host’s adaptive responses

Effect of Hemorrhage on Bladder Mucosal Blood Flow and Mitochondrial Function in the Pig 5 4.5 4

Tissue Flow NA ADH2

• How to identify patients who are becoming hemodynamically unstable before they progress too far? • How to determine the most appropriate therapy to reverse the primary cause for impending circulatory shock? • How to you implement the most appropriate therapy when individual training of care givers and responses of patients vary? vary?

Circulatory Shock in Phase Space

3.5 3

NADH 2 MAP Tissue blood flow Bladder SO2 10 per. Mov. Avg. (Bladder SO2)

140 120 100 80

MAP SO2

2.5 2 1.5

60 40

1 0.5 0

20 0

Calvijo et al. Med Sci Monit 14: BR175BR175-82, 2008

1

Measuring Mean Systemic Pressure at the Bedside

Calculating Pms at the Bedside Effective Circulating Blood Volume Navigator 1 Applied Physiology Pms = (Vs + Vus)/(Cvs + Cas) Vs = (Rvs x CO + Pra) Cv assuming i Cas/Cvs C /C = 1/24 andd R Ras/Rvs /R = 25 25//1 Pms = 0.96Pra 96Pra + 0.04Pa 04Pa + 0.96 x c x CO Eh = PmsPms-Pra Pra Parkin. Crit Care Resusc 1: 311 311--21, 21, 1999

Maas et al. Crit Care Med 37 37:: 912912-8, 2009

Why Blood Pressure Does Not Define Cardiovascular Status

Why Cardiac Output Does Not Define Cardiovascular Status Heart failure

Heart failure Hypovolemia CO = 5 L/min surface MAP = 80 surface

Pms

Pms

Hypovolemia or Sepsis

Eh

Eh

Combining pressure and flow helps

Separating Pms, Eh and SVR for Hypovolemia, Heart Failure and Sepsis

Heart failure

N a va g a to r 3 -D D is p la y

MAP=80 & CO=5 line for Eh

30

25

CO=5 CO= 5 Pms

20

Pms

15 10 20 18 16 14 12 10

0 0 .9

0 .8

8 0 .7

0 .6

Eh

6 0 .5

Eh Eh

TP

Hypovolemia

R

5

0 .4

4 0 .3

0 .2

2

S e p tic S h o ck C a rd io g e n ic S h o c k H e m o rrh a g ic S h o c k

2

Health and Disease Defined as a Time--Space Continuum Time • In a static field of single pointpoint-inin-time data health and disease can be separated in stochastic fashion using Neuronet approach to create a probabilistic equation • In a dynamic field of continuously changing but interinter-related variables, health and disease can be defined by the differences their Lorenz Attractors (ρ) independent of the actual physiological variables raw values.

Natural Non Non--Linear Systems

Chaos Theory and Biology • Non Non--linearity – Chaotic Behavior – Fractal appearance

• Non Non--linear thinking can result in solutions to otherwise unsolvable problems • Benoît Mandelbrot & James A. Yorke

Complexity Theory Self--Organizing Behavior Self

• Fractal Structure – Self similarity at all levels – Defined by minimal organizational rules

• Deterministic • Self Self--organizing • Adapt to external stress

Complexity Theory Self-Organizing Behavior Self-

Thermal scan of Petri dish during E. coli log growth at 37 C

Health and Disease Defined as a Time--Space Continuum Time • In a static field of single pointpoint-inin-time data health and disease can be separated in stochastic fashion using Neuronet approach to create a probabilistic equation • In a dynamic field of continuously changing but interinter-related variables, health and disease can be defined by the differences their Lorenz Attractors (ρ) independent of the actual physiological variables raw values.

With heating to 40 C, the thermal bands driven to randomness

3

Three Primary Clinical Problems • How to identify patients who are becoming hemodynamically unstable before they progress too far? • How to determine the most appropriate therapy to reverse the primary cause for impending circulatory shock? • How to you implement the most appropriate therapy with nonnon-physician when individual responses of patients and care givers vary?

Background Early discharge from ICUs to lower acuity monitoring units (step(step-down units or SDUs) is increasing, placing sicker patients in less well staffed units Minimally invasive monitoring devices are used to assess stability outside the ICU MET activation is grouped around morning and afternoon rounds, suggesting instability was missed at other times DeVita et al. Crit Care Med 34 34::2463 2463--78, 78, 2006

An electronic integrated monitoring systems (BioSigns) was developed to identify cardiocardio-respiratory instability using Neuronet analysis of existing ICU patient behavior Tarassenko et al. Br J Anaesth 97: 97:64 64--8, 2006

Efferent Arm: Medical Emergency Teams (MET) Improve Acute Care • System System--wide ICUICU-based MET activation to evaluate and treat patients at risk to develop adverse events • Pre Pre--defined MET activation criteria by non non--MD staff • Reduces d adverse d events Relative l i Risk i k Reduction: d i – Stroke: 78%, 78%, Severe Sepsis: 74%, 74%, Respiratory failure: 79% 79%

• Saves lives: post post--op death decreased 37% 37% • Decreases costs: less ICU transfers, decreased LOS • Bellomo et al. Crit Care Med 32: 32: 916 916--21 21,, 2004

But first one must identify these unstable patients

Explanation BioSign Neural Network

• An early warning system that may alert when all individually measured variables are still in their “normal” ranges • The score is weighted and automated through neural networking data fusion algorithmic processes networking, • Able to recognize changes from normality (defined by a training set) • Alerts for a single parameter deviating by >3 >3 SD from “normal” value in the training set, or 2-3 parameters moving away from normality by a smaller amount • Filtered for noise; requirement for temporal persistence (i.e. 4 out of previous 5 minutes) • Tarassenko et al. Br J Anaesth 97: 97:64 64--8, 2006

Study Design Pre

Phase I 8 weeks

Form Team

13 weeks

Phase III 8 weeks

Implement display

Implement use rules

Evaluate Literature Collect Data, Data Bedside display inactive

Evaluate Data Bedside display active

Implement Clinical Decision Rules for VSI trigger evaluate and Condition C

Develop Plan

Sensitivity and specificity— specificity — determine UPMC specific trigger value

Install Monitors

Train Staff

Phase II

Phase II training

Existing bedside monitor

Integrated Monitor (BioSign)

Phase III training

4

Phase 1 Results

Added Information

333 patient admissions representing all patients, reflecting 18,692 hours continuous monitoring All 7 MET activation events of respiratory and/or cardiac cause were detected by BSI in advance of MET activation Mean advanced detection time prior to MET activation was 6.3h 78% off METfull did not result l in i MET activation i i andd 23 and by t=30 The two trajectories vary by as much as their limits allow

– Pattern recognition of transitions from health to disease • John Hotchkiss, University of Pittsburgh

9

Can Complexity Modeling Define Health and Disease in Phase Space?

Mathematical Solution to Defining Health and Disease in a TimeTime-Physiological Variable Space • Essential elements in defining state are

Hypovolemia

Phase 3 subjects

Normal “Normal”

ρ β Shlomo Ta’asan, CarnegieCarnegie-Mellon University

The Attractor and its Measure The biologically significant quantity is the measure satisfying the equation

div(V (x, α) ρ) − σΔρ = 0 ρ((x)dx ) represent p the fraction of time the trajectory j y spend p in a small volume around x

– Lorenz attractor (ρ (ρ) location – Degree of variance of measured variables ((β β) – Number of variables measured over time

• Of less importance in defining state are – Accuracy of exact measured variables – Absolute value of any variable within “viability” Shlomo Ta’asan, CarnegieCarnegie-Mellon University

Defining Physiological State • By understanding how to characterize ρ one can define the number of determinants and the duration and frequency of observation needed to characterize a state • Health would then be defined as a series of ρ* considered by prior observation to be “good” • Disease would then be defined as any ρ not within the boundaries of any ρ*

Increasing noise to define phase-time planes

Defining Disease as ρ to ρ* Distance Healthy subject

Unhealthy subject

Implications for Diagnosis of Disease • Trend monitoring of interinter-related variables is essential to define ρ • Depending of the specific disease state, nature of the th true t variance i (β ) and (β d error ((σ σ) iin measures, additional specific measures may be needed to increase the accuracy of estimating ρ • The Wasserstein Distance (W) between ρ and ρ*

Shlomo Ta’asan, CarnegieCarnegie-Mellon University

10

Implications for Management of Disease Based on Complexity Modeling • Experience reports that some ρ are “good” in that they characterize healthy people • Some regions in the phase space are “bad” in that the are seen with known shock states • Biological parameters are implicit in V(x, V(x,α α) • Thus treatment reduces to a control problem E(α E( α) = dW2 (ρ(σ), ),ρρ*) Minimization of the distance between ρ(σ) and ρ* Conceptually, this method places within linear thinking nonnon-linear relations

Improving Predictions of Behavior • Increase the duration of measured time to define chaotic behavior – The VSI analysis required >90 >90 minutes to define state accurately enough to be diagnostic under most conditions

• Increase the number of measures of relevant independent variables

Pinsky. Crit Care Med 38 38:S :S649 649--55 55,, 2010

System Identification Driving Questions

Pinsky. Crit Care Med 38 38:S :S649 649--55 55,, 2010

PITT Index Probability of Event

• How complex is the response?