You ain t seen nothin yet Critical care informatics Timothy G. Buchman, Ph.D., M.D., FACS, FCCM Founding Director, Emory Center for Critical Care External Faculty, Santa Fe Institute
DISCLAIMER: I DID NOT PICK THE TITLE
"You need educatin' You got to got to school (BTO)
"You need educatin' You got to got to school (BTO) Current generation monito present data that are Not predictive; and Do not offer decision assi capability
None of these is predictive None offers decision-assist One is closed loop TMI
Portrait of Gertrude Stein by Pablo Picasso Metropolitan Museum of Art, New York City. When Gertrude Stein was being wheeled into the operating room for surgery on her stomach, she asked Alice B. Toklas, "What is the answer?" When Toklas did not answer, Stein said, "In that case, what is the question?"
In that case, what is the question?" Healthy well Unhealthy chronic conditions like hypertension Chronically Ill, Acute exacerbations e.g.hemorrhagic stroke Acute illness, e.g. viral illness and Guillain- Barre ICU
In that case, what is the question? Healthy well Given current circumstances (condition & interventions), will the patient recover? ICU
In that case, what is the question? Healthy well Given current circumstances (condition & interventions), will the patient recover? If not, what can/should/must be changed? ICU
System: Neuro
System: Neuro Assumes Complete Description
condition & interventions state
Geometries of Biological Time: Stress Test Indirect reflection of arterial blood flow to heart during exercise 1 st test 1929, now standardized Authentic exercise (or drug, dobutamine) (a way to get from health to stress, at least for study)
State and velocity of change
System: Neuro Assumes Complete Description Data- Based Prediction
prediction
Natural Clinical Experiment : Spontaneous Breathing Trial (SBT) BEFORE SBT AFTER Controlled Ventilation: assisted control (A/C) f=(12-18)min -1 =(0.2-0.3)Hz Tidal Volume=(6-10) ml/kg Spontaneous Breathing continuous positive airway pressure (CPAP) Controlled Ventilation: assisted control (A/C) f=(12-18)min -1 =(0.2-0.3)Hz Tidal Volume=(6-10) ml/kg pressure (t) pressure=5 mmhg pressure (t) time time time 30 min 30 min 30 min
FunBio Clocks/Time
FunBio Clocks/Time BEFORE Perturbation AFTER Heart Rate Spontaneous Breathing Trial (SBT) BEFORE SBT AFTER Controlled Ventilation: assisted control (A/C) f=(12-18)min -1 =(0.2-0.3)Hz Tidal Volume=(6-10) ml/kg Spontaneous Breathing continuous positive airway pressure (CPAP) Controlled Ventilation: assisted control (A/C) f=(12-18)min -1 =(0.2-0.3)Hz Tidal Volume=(6-10) ml/kg pressure (t) pressure=5 mmhg pressure (t) time time time Fluctuation-Dissipation Theorem (FDT) Provides a Simple Analytical Relationship between Post- Perturbation Heart Rate Recovery (HRR) and Heart Rate Variability (HRV) During the Stress Lu Y., Burykin A., Deem M. W., Buchman T. G. Predicting Clinical Physiology: A Markov Chain Model of Heart Rate Recovery after Spontaneous Breathing Trials in Mechanically Ventilated Patients. Journal of Critical Care (2009) 24, 347 361.
FunBio Clocks/Time Fluctuation-Dissipation Theorem formulated by Harry Nyquist, 1928 proved by Herbert B. Callen, 1951 FDT is a powerful tool for predicting the relaxation behavior of a system from its reversible fluctuations in thermal equilibrium. The theorem states a general relationship between the response (response function) to an external disturbance and the internal fluctuation (correlation function ) of the system in equilibrium. Often that linear response takes the form of one or more exponential decays.
f II f ( t0) f II 0 B( t) = B θ ( ( t t0)) = B0θ ( t0 0 t ) < f > = const IV Our first-pass approach: apply equilibrium statistics to the far-from-equilibrium problem Heart Rate Variability (HRV) - microscopic fluctuations of HR at rest (I or IV); Heart Rate Recovery (HRR) macroscopic change of HR after SBT; Stress B (off vent) - perturbation
Heart Rate Prediction under Gaussian Assumption px (, x ) χ(0) 1 1 δx Rx ( x ) = = exp[ ( δx, δx ) M ( δx, δ x ) + ] χ(0) 2 i + 1 i T 1 i i+ 1 i i i+ 1 i i+ 1 px ( ) 2 2 2 i π Μ pt ( ) ( ) n n = Rpt0 HRF( tn) state()* i pit (; n) i Prediction:< >= Patient H1 Heart Rate Frenquency Patient N1 Heart Rate Frenquency Patient P2 Heart Rate Frenquency 0 2000 4000 6000 8000 Patient C Heart Rate Frenquency Patient E Heart Rate Frenquency Patient G Heart Rate Frenquency 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 0 1000 2000 3000 4000 5000 6000 7000 0 2000 4000 6000 8000 0 1000 2000 3000 4000 5000 6000 0 2000 4000 6000 8000 0 1000 2000 3000 4000 5000 6000 7000
Relaxation Time: Prediction vs. Exponential Fit Failure of Prediction? Poorly sedated, altered Neuro status
Developing the capability to do this in real time Digitized at 240 Hz
Summary Prediction tools Tele- ICU
Questions?