Emerging Paradigms of Cognition inMedical Decision-Making
Vimla L. Patel, PhD, DSc, FRSCLaboratory of Decision Making and Cognition
Department of Biomedical InformaticsColumbia University
New York, NY
Workshop onMedical Thinking: What We Know
University College, LondonJune 22-23, 2006
• CPOE system facilitated 22types of medication errors
• Fragmented displays preventcoherent medications view
• Pharmacy inventory mistakenfor dosage guidelines
• Separation of functions thatfacilitate double dosing andincompatible orders
• Inflexible ordering formatsgenerating wrong orders
• Three quarters of the housestaff reported observing eachof these error risks, indicatingthat they occur weekly or moreoften
Koppel R, Metlay JP,Cohen A, Abaluck B,Localio AR, Kimmel SE,et al.
Role of ComputerizedPhysician Order EntrySystems in FacilitatingMedication Errors
JAMA 2005;293:1197-203.
Another Case• Potassium chloride (KCl) ordered as IV
injection and as IV fluid additive usingEclipsys CPOE system
• 85-year-old patient, admitted to the medicalICU with septic shock and respiratory failure
• Patient received 316 mEq KCl over 42 hrs insetting of acute and chronic kidney failure
• High dose delivered due to errors andmisperceptions by several care providers
• Compounded errors propagated through thesystem over three days
Horsky J, Kuperman GJ, Patel VL. Comprehensive analysis of a medicationdosing error related to CPOE: A case report. J Am Med Inform Assoc2005;12:377-382.
Analysis of Failure• Misconceptions about the relation
between IV volume (humans) and timeduration (system)
• Misconception of latest and “dated”laboratory results
• Lack of Alerts when potassium valuereached a dangerous level
• Inadequate clinical user trainingregarding safe and efficient orderingpractices
Why should we be concerned aboutcognitive and social sciences?
• Human behavior is influenced by ourthoughts and our social values
• Cognitive / Social sciences providetheoretical & methodologicalfoundations for the study of humanbehavior
• Problems of changing behavior hasalways been an issue, but now we havebetter methods for investigation
• It isn’t sufficient to use the methodswithout understanding the rationale
What is Cognitive Science?
Multidisciplinary field incorporatingtheories and methods frompsychology, linguistics, philosophy,anthropology, and computer sciencein the investigation of cognitiveprocesses in humans and machines
Patel, V.L., Arocha, J.F. & Kaufman, D.R (2001) A Primer on Aspects of Cognition for Medical Informatics.Journal of the American Medical Informatics Association. 8; 324-343.
Cognitive Research
• Investigates psychological processesunderlying performance via an in-depthanalysis of cognitive steps that lead toobservable behavior
• Methodology: focus on understandingknowledge structures and mentalprocesses in cognitive activity
Typical Methods
• Naturalistic Field Studies: Ethnography• Participant Observation• Use of think aloud protocols• Study of Naturally Occurring Discourse:
Discourse analyses• Interviews: semi-structured
questionnaires• Case Studies• Video recordings
Methods of Analysis• Task and activity analysis• Representations of ideas and concepts
(Propositional)• Meaningful relations between ideas and
concepts (Semantic), higher levelunderstanding (conceptual), and contextsensitive (pragmatic) representations
• Dialogue analysis for team communication• Protocol analysis• Usability analyses
From Cognitive Scienceto Medical Cognition
Cognitive ScienceTheory
• Memory• Knowledge
Organization• Problem Solving• Heuristics/Strategies• Computational
Theory of Mind
Medical CognitionConceptual Frameworks
• Medical Problem Solving• Organization of Clinical
and Basic-ScienceKnowledge
• Diagnostic ReasoningStrategies
• Medical Decision Making
From Medical Cognition toBiomedical Informatics
Medical Cognition• Medical Problem Solving• Organization of
Knowledge• Diagnostic Reasoning
Strategies• Medical Decision Making• Text Comprehension and
Problem Representation• Development of Medical
Expertise• Medical Discourse
Biomedical Informatics• Knowledge and Data
Representation• Management of Medical
Information• Human-Computer
Interaction• Cognitive Models for
Enhancing DecisionSupport
• Cognitive Assessment ofUsability and Interfaces
• Targeted Training
Clinical Applications and Cognition
• Nature of Expertise andHeuristic Reasoning
• Effects of technology onReasoning
• Clinical decision making andmedical errors
Reference
• Patel, V.L., Arocha, J.F. & Kaufman, D.R. (1994)Diagnostic Reasoning and Expertise. The Psychologyof Learning and Motivation: Advances in Researchand Theory, 31, 137-252.
Forward-Directed Reasoning:An Example
Unemployed young male presents with fever, rigor,and chills. A puncture wound is observed on his arm.
UnemployedYoung Male
PunctureWound
IntravenousDrug User
ContaminatedNeedle Infection
Data Hypothesis
Backward-Directed Reasoning:An Example
Elderly woman presents with signs of myxedema. She alsopresents with respiratory failure, which is unexplained for bymyxedema.
Data Hypothesis
RespiratoryFailure Hypoventilation Hypometabolic
State
Myxedema
Case Interpretation by a Novice45 yr.-old
male
MyocardialInfarction
4-hour hx of chest pain
central, crushing chest pain
faintness
sweating
mild cough
OtherDiagnosis
Case Interpretation by an Intermediate45 yr.-old
male
MyocardialInfarction
4-hour hx of chest pain
central, crushing chest painfaintness
sweating
mild cough
OtherDiagnoses
Case Interpretation by an Advanced Subject
45 yr.-old male
MyocardialInfarction
4-hour hx of chest pain
central, crushing chest painfaintness
sweating
mild cough
OtherDiagnoses
AorticDissection asymmetric BP
Constraints on Decision Making• Cognitive
– Memory– Knowledge– Inferences– Strategies
• Socio-Cultural– Group norms– Policies and Practices
• Organizational/Situational– Environmental– Team and Resources
Clinical Applications and Cognition
• Nature of Expertise andHeuristic Reasoning
• Effects of technology onReasoning
• Clinical decision making andmedical errors
Effect of an EMR System onHuman Cognition
• Transition from paper records to EMRand back to paper record
• Impact on knowledge organization,reasoning
• Information and other technologiesare not merely tools to expedite,facilitate and enable the execution oftask
Patel V, Kushniruk A,Yang S,Yale J-F, Impact of a computer-based patient record system on data collection,knowledgeorganization and reasoning. JAMIA,7(6)569-85,2001
Category of Information Hand-WrittenPatient Record
Computer-BasedPatient Record
1. Chief Complaint 10 282. Past Medical History 13 133. Life Style 33 194. Psychological Profile 10 115. Family History 7 146. History of Present Illness 55 277. Review of Systems 52 88. Physical Examination 60 559. Diagnosis 14 910. Investigation 29 1711. Treatment 21 24
TOTAL ENTRIES 304 225
Information in EMR and Hand-WrittenRecords
First section from paper-based record (Pre-EMR)
74 year old woman, whose diagnosis was made inFebruary, as she complained of polyuria/nocturiaand fatigue for a few years. She was told hersugar was very high and she was sent to Dr. K.,who started her on Diabeta 5 mg/d and sent her toDr. S. in ophthalmology who reported normalretina. She lost weight, her polyuria improved, herbladder urgency got better, and her glucosevalues improved dramatically. She does nomonitoring at home. She had to be hospitalizedfor an ankle fracture after falling on ice, for 3months. At follow-up, Dr. K. seemed pleased withthe results.
First Section from Electronic Medical Record(EMR)
CHIEF COMPLAINT: Type II diabetes mellitus
PERSONAL HISTORYSURGICAL: cholecystectomy: Age 60 years old
MEDICAL: hypothyroidism: asymptomatic since 25 years
LIFE STYLEMEDICATION
DIABETA (Tab 2.5 MG)Sig: 1 tab(s) Oral before breakfast
SYNTHROID (Tab 0.125 MG)Sig: 1 tab(s) Oral before breakfast
HABITS: smoking: 0 alcohol: 0
First Section from Paper-Based Record(Post-EMR)
Diabetes type I X age 4Currently on N54 - N28
R6 - R2 Measure with OT IIGlucose levels: <130 130-180 >180
AM IIIIIII IIIIIIIIIIIIIII Lunch Supper IIIIIIIIII Bedtime IIIIIIIII IIIIIIIIIIII
Last HbA1C since April 96: 7.4/7.2/6.7/6.6/8.9 - highervalues in log book
Retinopathy: NIL March 97Nephropathy: NIL Oct. 96
Electronic Medical Record Return to Paper Record
Same as EMR!
Multiple Hypotheses
HypothesesPatient Data
Diagnostic Reasoning
Patient Data
Paper Record
SYSTEMSLEVEL
DIAGNOSTICLEVEL
INTERMEDIATECONSTRUCT
FINDINGLEVEL
OBSERVATIONLEVEL
C1 C2
D1 D2 D3
FA1 FA2 FA3 FA4 FA5
F1 F2 F3 F4 F5 F6 F7 F8 F9
O7
O8
O9
O10
O11
O12
O1
O2
O3
O4
O5
O6
+ + +
Structure ofMedical
Knowledge inProblemSolving
Influence of Technology onHuman Cognition
• Information and other technologies are notmerely tools to expedite, facilitate and enablethe execution of tasks They have profound and enduring
consequences• Optimal design requires sensitivity to internal
organization of concepts by human beings Utility and acceptance depend on
designer’s recognition that the system is amediator of cognition (re-organizescognition)
Clinical Applications and Cognition
• Nature of Expertise andHeuristic Reasoning
• Effects of technology onReasoning
• Clinical decision making andmedical errors
Some References
• Patel, V.L., Kaufman, D.A., Arocha, J.F. (2002)Emerging Paradigms of Cognition and MedicalDecision Making, Journal of Biomedical Informatics,35, 52-75.
• Patel, V.L., Arocha, J.F., & Zhang, J. (2005). Thinkingand reasoning in medicine. In K.J. Holyoak & R.G.Morrison (Eds.), Handbook of thinking and reasoning(pp. 2298-2538). Cambridge: Cambridge UniversityPress.
Studying Decision MakingClassical Approach:• Adopts normative or rational
decision models• Models tend to be
prescriptive• Any systematic deviations
from normative standardsare seen as decision biases
New Framework: Expanding theScope of Decision Making
• More descriptively adequateaccount of decision making
• Explain adaptive as well assuboptimal characteristics ofdecision makers
• Decisions embedded in abroader social andtechnological context
Progress and Occurrence of Medical Error
Near Miss
Boundary
Normal Routine
Adverse EventReport
Death
Boundary
• Need a better understanding ofcognitive and situational demands oncompetent performance of clinicians inteam work
• Must recognize that acquiringexpertise implies developing an abilityto adapt flexibly to ever changingcomplex situations
• Need baseline information on howclinicians make decisions in theireveryday tasks
Clinical Cognition:Implications for Informatics
Cognition in the Wild
• Distributed cognition• External and Internal resourses
Cohen, T., Blatter, B., Almeida, C., Shortliffe, E., &Patel, V. (in press). Distributed cognition in thePsychiatric Emergency Department: A cognitiveblueprint of a collaboration in context. ArtificialIntelligence in Medicine.
Shadowing of medicalteam personnel during‘Crucial Periods’ pertinentto the individual.
Mapping the activities to theICU/ER layout and time-stamping each interaction orevent.
Conducting brief interviews to gaininsight on infrastructure, roles,shifts, timings.
Obtaining log files of the clinicalinformation systems andattempting correlation withobservational data.
Specific Methods
Distributed Cognition
• Cockpits, navigation• Shift in focus
A solitary individual
Groups of individuals in context
Collaborative Cognition1. Team members2. RepresentationDATA
DATADATA
3. Data sourcesMultiple
Intellectual Partnership
• Distributed cognition– Human-computer interaction analysis paradigm
PDA• Knowledge residespartly in theenvironment
Intellectual Partnership
Coordinating Coordinating internalinternal (user (user’’s mind) ands mind) andexternalexternal (interface, environment) resources (interface, environment) resources
PDA
Resident 1
Attending 1
Social worker
Cognition Distributed Across Time
Nurse
Attending 2
Resident 2
Attending 1
Resident 4
Nurse
Social worker
DURATION OF STAY
Night Resident
Night Nurse
Mobile Crisis
“Area B”core ‘mini-team’
Cognition Distributed Across Teams
Family
Managed Care Liaison
Security
Inpatient Admission
OPD
PatientNurse
Resident
Attending
SW/PSAC
Homeless Outreach
NYPD
Night NurseNurseNurse
DURATION OF STAY
Resident 1
Attending 1
Social worker
Bridging by External Representation
Attending 2
Night Resident
Resident 3
Attending 1
chart chart chart
Limitations ofNaturalistic Decision Making
• Massive amount of data• Labor intensive analysis• No specific standard to evaluate
against• Limited generalizability• Bench to Bedside needs also to
include Bedside to Bench
Conclusions
• Normative models, despite grandcontributions, cannot accuratelydescribe the complexity of real-lifedecision making.
• A more descriptive naturalisticapproaches contribute significantlyto overcome this limitation
• A broader framework to include boththe approaches is proposed
Some Lessons• It is human to make errors, but we also learn
from errors• Challenge is to reduce errors while improving
our ability to recognize and correct errorsbefore they do irreparable harm
• Technology can help but also introduce newopportunities for errors
• We need a better understanding of cognitiveand situational demands on competentperformance
• We must recognize that acquiring expertiseimplies developing an ability to adapt flexiblyto ever changing situations
BiomedicalInformaticsTextbook
(3rd edition)Springer Verlag - 2006
BioChapter 4:
Cognitive Science inBiomedical Informatics
—VL Patel and DR Kaufman—
Thank You
[email protected]://www.dbmi.columbia.edu/patel/