AIMed NORTH AMERICA, CALIFORNIA 11–14 DECEMBER 2019 The R Th Role o of C Cogniti tive ve Sc Science i in Dev Devel elop oping Safe afe an and Ac Accep eptab able e Cl Clini nica cal Tech chno nologies Vimla L. Patel, PhD, DSc, FRSC Senior Research Scientist and Director Center for Cognitive Studies in Medicine and Public Health New York Academy of Medicine, New York City Professor (Adjunct of Biomedical Informatics Columbia University, New York, NY Arizona State University, Phoenix, AZ www.aimed.events/northamerica-2019 www.lohia-patel.net @VimlaLPatel
A Case of Potassium Chloride Overdose Potassium chloride (KCl) ordered as IV injection and as IV fluid additive using a CPOE system • 85-year-old patient, admitted to the medical ICU with septic shock and respiratory failure • Patient received 316 mEq KCl over 42 hrs in setting of acute and chronic kidney failure • High dose delivered due to errors and misperceptions by several care providers • Compounded errors propagated through the system over three days Horsky, J., Kuperman, G.J., & Patel, V.L. (2005). Comprehensive analysis of a medication dosing error related to CPOE: A case report. Journal of the American Medical Informatics Association; 12 : 377-382 .
JAMIA 2005 12:377-82
Understanding What is Underneath
Saturday, Jan 31, 2004 r e e m b e m a m N u a t y N c N n c i e e r n n r s d d e e s e u n e i n d o g d i u t s t a u r a e D e n e r q O G t o B E B R S e R r F A 1 Active potassium Potassium 10 mEq q1h IV Sat 1/31/04 Sat 1/31/04 1:30 pm Chloride Chloride Inj 1:30 PM 5:00 PM A 2 DC D/C D5W NULL IV Fluids, Sat 1/31/04 Sat 1/31/04 Crystalloi 1:30 PM 1:30 PM d A 3 Active D5W w/ for one liter IV Fluids, Sat 1/31/04 Sat 1/31/04 40meq KCl/L Crystalloi 1:31 PM 1:35 PM d A 4 DC D/C D5W w/ NULL IV Fluids, Sat 1/31/04 Sat 1/31/04 40meq KCl/L Crystalloi 1:35 PM 1:35 PM d A 5 Active D5W Enter IV Fluids, Sat 1/31/04 Sat 1/31/04 additives Crystalloi 1:35 PM 2:26 PM d A 6 Active D5W w/ 100 Enter IV Fluids, Sat 1/31/04 Sat 1/31/04 meq KCl/L additives Crystalloi 1:36 PM 2:26 PM d A 7 DC potassium KCl 20 mEq 20 mEq NUL IVP Sat 1/31/04 Sat 1/31/04 Chloride/wate B 1:37 PM 1:37 PM Inj - Central r 2:26 pm A 8 DC D/C D5W NULL IV Fluids, Sat 1/31/04 Sat 1/31/04 Crystalloi 2:26 PM 2:26 PM d A 9 DC D/C D5W w/ NULL IV Fluids, Sat 1/31/04 Sat 1/31/04 100 meq Crystalloi 2:26 PM 2:26 PM KCl/L d A 10 Active D5W w/ Enter IV Fluids, Sat 1/31/04 Sat 1/31/04 80meq KCl/L additives Crystalloi 2:42 PM 3:34 PM d 3:34 pm A 11 DC D/C D5W w/ NULL IV Fluids, Sat 1/31/04 Sat 1/31/04 80meq KCl/L Crystalloi 3:34 PM 3:34 PM d A 12 Active D5W w/ Enter IV Fluids, Sat 1/31/04 Mon 2/2/04 80meq KCl/L Crystalloi 3:34 PM 9:39 AM d
Swiss Cheese Model of Patient Safety Errors Patient needs medication . There are many opportunities for error detection & correction: AI tool for automatic R1 real time detection ? R2 R3 Pharmacy/ Nurse Patient receives medication overdose
Intelligent Support for Augmenting BMJ Quality and Safety . 24, 468–474. Human Error Detection and Correction
Cognitive Science and AI Cognitive science or the science of cognition, includes numerous subfields of psychology, philosophy, linguistics, cognitive anthropology, neuroscience and computer science. Various field within cognitive science, influence AI research, where one branch of AI system developers try to mimic cognitive processes (not machine learning): Augmenting human intelligence Patel, V.L., & Kaufman, D.R. (2013). Cognitive Science and Biomedical Informatics. In E. H. Shortliffe & J. J. Cimino (Eds.), Biomedical Informatics: Computer Applications in Health Care and Biomedicine (4 ed., pp. 133-185). New York, NY: Springer-Verlag.
Message 1: Role of Human Cognition § The overdosing case study exemplifies a larger problem: Decision support systems, as intelligent as they may be, seldom align with the mental processes underlying clinical decisions. § We cannot underestimate the important role of human mind § We cannot view AI as passive tools, they influence our thinking, as much as we shape their behavior Patel, V.L., Kushniruk, A.W., Yang, S., & Yale, J.F. (2000) Impact of a computerized patient record system on medical data collection, organization, and reasoning. Journal of the American Medical Informatics Association; 7(6): 569-585
Message 2: Understand the Real Clinical World • There are hidden variables that can never be taken into account even if you have billions of data points. All risks cannot be computed • To function in a complex and uncertain world, clinicians need: • To have at their disposal a simple yet effective toolset for decision-making • To develop the ability to ignore information that is not critical or important • To know when to trust their own judgment rather than relying completely on complex models Patel, V.L., Kaufman, D.R., & Kannampallil, T.G. (2019). Diagnostic Reasoning and Expertise in Healthcare . In: The Oxford Handbook of Expertise: Research & Application.
Message 3: Future Role of Cognition and AI Future looks towards Cognitive AI • Deep learning AI systems crunch data • Symbolic AI, deals with human-like reasoning • The unique hybrid combination of the two plus the explainability (transparency) of symbolic AI is the next step towards the future of AI and decision making
Thank you vpatel@nyam.org http://www.lodhia-patel.net/
Recommend
More recommend