Variability Analysis and Machine Learning are useful for clinical decision support at the Bedside

As the multivariate measurement of physiological variability provides additional information regarding a patient’s condition, machine learning is well suited to transform the variability data, in addition to any relevant clinical data, into predictive models that may be useful to help with clinical decision-making.

Patient monitoring routinely involves multiple waveforms, (e.g. ECG), sampled hundreds of times per second. Ignored are the patterns of variation of the inter-beat and inter-breath interval time series, where illness is characterized by their reduction and health is characterized by significant complex variation.^6,35,39 Over the last decade, my research team and I have pioneered the application of variability analysis at the bedside,^{5,23,24,27,33,35,36,40-43} transforming waveform monitoring into predictive models using artificial intelligence. We have used this approach to provide early detection of infection in neutropenic patients,²³ determine severity of illness in critically ill,²⁴ determine risk of future deterioration in emergency department patients with sepsis,⁴⁴ determine risk of extubation failure,^32,33 and predict likelihood of donation success after DCD. Further, we have integrated these variability-derived predictive models into clinical decision support software (CDSS) tools.

In collaboration with the Canadian Critical Care Trials Group (CCCTG), my team and I are implementing and evaluating the following variability-derived CDSS tools. (1) Extubation Advisor (EA) provides optimal prediction of extubation outcomes with standardized assessment of readiness for extubation. We built a prototype and observationally implemented EA at TOH last year, funded through TOHAMO Innovation grant (manuscript submitted). With co-PI Dr. Karen Burns (St Michael’s Hosp), we are planning multicenter interventional evaluation. (2) Donation Advisor was developed with co-PI Dr. Sonny Dhanani (CHEO), which identifies DCD candidates likely to be successful organ donors and objectively characterizes organ ischemia during DCD process. We have received funding from both Physicians’ Services Incorporated ($190k, 2019) and Health Canada (HC) grants ($610k, 2019) to perform multicenter implementation and evaluation of Donation Advisor, which we have initiated. Most recently, (3) Sepsis Advisor was developed with co-PI Dr. Doug Barnaby (NYC), which quantifies the risk of future deterioration in patients presenting with infection to guide management and optimize disposition decisions (i.e. transfer to ward, ICU, or home).  To our knowledge, no lab in the world has developed analogous tools that transform waveform monitoring into clinical decision support tools.

As CDSS tools require regulatory approval and integration with monitors and electronic health record, commercialization is required for clinical adoption; thus I founded Therapeutic Monitoring Systems (TMS – see www.therapeuticmonitoring.com) and long ago patented the underlying technology. With issued patents, proprietary data, and strong clinician scientist support, TMS attracted investment and is actively partnered with international companies to bring these CDSS tools to the bedside. The planned rigorous evaluation of these tools is required alongside their clinical implementation.

Relevant Papers:

1. Barnaby DP, Fernando SM, Herry CL, Scales NB, Gallagher EJ, Seely AJE Heart Rate Variability, Clinical and Laboratory Measures to Predict Future Deterioration in Patients Presenting With Sepsis. Shock 2019, Apr;51(4):416-422.
2. Godard S, Herry C, Westergaard P, Scales N, Brown SM, Burns K, Mehta S, Jacono FJ, Kubelik D, Maziak DE, Marshall J, Martin C, Seely AJE. Practice Variation in Spontaneous Breathing Trial Performance and Reporting,” Canadian Respiratory Journal, vol. 2016, Article ID 9848942, 10 pages, 2016.
3. Seely AJ, Bravi A, Herry C, Green G, Longtin A, Ramsay T, Fergusson D, McIntyre L, Kubelik D, Maziak DE, Ferguson N, Brown SM, Mehta S, Martin C, Rubenfeld G, Jacono FJ, Clifford G, Fazekas A, Marshall J. Do heart and respiratory rate variability improve prediction of extubation outcomes in critically ill patients? Crit Care. 2014, Apr 8;18(2):R65.
4. Herry CL, Townsend D, Green GC, Bravi A, Seely AJ. Segmentation and classification of capnograms: application in respiratory variability analysis. Physiol Meas. 2014, Dec; 35(12):2343-58.
5. Flouris AD, Poirier MP, Bravi A, Wright-Beatty HE, Herry C, Seely AJ, Kenny GP. Changes in heart rate variability during the induction and decay of heat acclimation. Eur J Appl Physiol. 2014, 18: 65.
6. Flouris AD, Bravi A, Wright-Beatty HE, Green G, Seely AJ, Kenny GP. Heart rate variability during exertional heat stress: effects of heat production and treatment. Eur J Appl Physiol. 2014, Jan 5.
7. Bravi A, Green G, Herry C, Wright HE, Longtin A, Kenny GP, Seely AJE.  Do physiological and pathological stresses produce different changes in heart rate variability?  Front Physiol. 2013, Jul 30;4:197.
8. Green, GC, Bradley B, Bravi A, Seely AJE.  Continuous multiorgan variability analysis to tract severity of organ failure in critically ill patients. J Crit Care. 2013, Oct;28(5):879. Epub 2013 May 29.
9. Armstrong RG, Ahmad A, Seely AJE, Kenny GP.  Heart rate variability and baroreceptor sensitivity following exercise-induced hyperthermia in endurance trained men.  Eur J Appl Physiol. 2012, Feb;112(2):501-11.
10. Bradley B, Green GC, Batkin I, Seely AJE.  Feasibility of continuous multiorgan variability analysis in the intensive care unit.  J Crit Care 2012, Apr;27(2):218.e9-20.
11. Bravi A, Green G, Longtin A, Seely AJE.  Monitoring and identification of sepsis development through a composite measure of heart rate variability. Plos One 2012, Sept 19;7(9): e45666. 
12. Bravi A, Longtin A, Seely AJE.  Review and classification of variability analysis techniques with clinical applications.  Biomed Eng Online. 2011, Oct 10;10:90. Review
13. Bunchan CA, Bravi A, Seely AJE.  Variability Analysis and the Diagnosis,Management, and Treatment of Sepsis.  Curr Infect Dis Rep 2012, Aug 5;14(5):512-521.
14. Armstrong RG, Kenny GP, Green G, Seely AJE. Diurnal variation in heart rate variability before and after maximal exercise testing. Chronobiol Int. 2011, May;28(4):344-51.
15. Armstrong RG, Seely AJE, Dilby D, WS, Kenny GP Journeay WS, Kenny GP. Cardiovascular and thermal responses to repeated head-up tilts following exercise-induced heat stress.  Aviat Space Environ Med. 2010, Jul;81(7):646-53.
16. Ahmad S, Ramsay T, Huebsch L, Flanagan S, McDiarmid S, Batkin I, McIntyre L, Sundaresan RS, Maziak DE, Shamji FM, Hebert P, Fergusson D, Tinmouth A, Seely AJE. Continuous multi-parameter heart rate variability analysis heralds onset of sepsis in adults. PLOS One 2009, Aug 2009: 14;4(8):e6642.
17. Seely AJE, Macklem PT. Complex Systems and the Technology of Variability Analysis, Critical Care 2004, 8(6):R367-84.