Chapter 3 Clinical Decision Support Flashcards
Decision analysis
Practical application of how people ought to make decisions
Uses rational procedure to identify likelihood and value (positive or negative) of all possible outcomes associated with each option
Expected value is likelihood times value
Clinical decision support systems
-Systematic and comprehensive software developed to help with decision analysis
-Any electronic tool that provides structured guidance
Heuristics
-mental shortcuts, information-processing rules that brain uses to produce decisions or judgments
-Can lead to biases
Cognitive biases
tendencies to think in certain ways that can lead to systematic deviations from a standard of rationality or good judgment
Availability bias
Tendency to overestimate the probability of usual events because of recent or memorable experiences
Representativeness bias
Tendency to overestimate unusual diseases or conditions due to matching pieces of the typical picture of that disease
Anchoring bias
Tendency to rely too heavily or anchor on one trait or piece of information when making decisions (usually the first piece of information acquired on that subject)
Blue based bias
Tendency to over or under estimate the probability of an outcome based on the perceived value associated with that outcome
Confirmation bias
Tendency to search for, interpret, focus on, and remember information in a way that confirms one’s preconceptions
Automation bias
Tendency to depend excessively on automated systems which can lead to erroneous information that can override or interfere with correct decisions
Clinical reasoning (4 models)
-cognitive process that clinicians use to discard or confirm a hypothesis
-Additive model: sum of diagnostic weights (positive weight if evidence supports diagnosis, negative weight does not support diagnosis)
-Bayesian model: method to calculate the probability of a condition based on the prevalence or pretest probability of the concern and other related events
-Algorithmic model: follow an internal flow sheet with series of branching logic
-Blois’ funnel: start with large differential diagnosis that gets whittled down based on medical history, physical exam, and diagnostic testing
Expected value
In decision analysis is the summation of the independent probabilities of events
Expected utility
In decision analysis includes expected value but also takes into account mitigating factors like risk aversion, personal preferences, or circumstances
-Can be used in decision analysis to adjust the value of the outcome based on the perceived utility of that outcome for that patient
Probability notation
-conditional probability
-addition rule: P(A)+P(~A)=1; (If A related to B) P(B)=P(A and B)+P(A’ and B)
-multiplication rule: P(A and B)=P(A)*P(B|A)
-outcome rule: P(A)+P(B)+P(C)…=1
Decision tree or graph notation
-square: decision node
-circle: change node
-each branch has probability
-probability of all branches of node must add up to 1
-triangle: outcome node
-outcomes are assigned value (cost, utility, QALY, etc.)
-if life and death are outcomes, life=1, death=0
Sensitivity analysis
what if analysis to determine how projected performance is affected by changes in the underlying assumptions
Standard gamble
-Way to assess utility or preference assessment
-Ask patient which of 2 choices they want: 1) continue life with the current medical condition, 2) choose the intervention with a defined risk of death
-adjust risk of death, until get to point of indifference, where he can’t really choose between the two options
-resulting value is the utility
Time Trade Off
-Way to assess utility or preference assessment
-sick patients estimate how many years of their life they would be willing to give up to live a certain number of years in full health
-TTO utility (the indifference point) is the length of remaining life in perfect health divided by length of remaining life with the evaluated health state
Visual analog
-Way to assess utility or preference assessment
-patients are asked to rate different health states on a marked or unmarked scale where 0=death, and 100=perfect health
Cost-effectiveness analysis
-Form of economic analysis that compares relative costs and outcomes (effects) of two or more courses of action
-Typically denominator is gain in health from a measure, an enumerator is cost associated with the health gain
-Often visualized on cost-effectiveness plan consisting of four quadrants (expensive/effectiveness)
Cost-benefit analysis
Assigns monetary value to the measure of effect
Incremental Cost/Effectiveness Ratio (ICER)
-Way to do cost-effectiveness analysis
-(Cost of A - Cost of B)/(Effect of A - Effect of B)
Cost-utility analysis
Similar to cost-effectiveness analysis
Computer interpretable guidelines (CIG)
Way to enact clinical practice guidelines into clinical workflow
Syntactic ambiguity
-Type of CIG barrier
-Sentence construction could have more than 1 meaning
Semantic ambiguity
-Type of CIG barrier
-One word having different meanings
Pragmatic ambiguity
-Type of CIG barrier
-Inconsistent or conflicting recommendations
Under-specification
-Type of CIG barrier
Strength qualifiers
-Type of CIG barrier
-“should be effective”
Passive voice
-Type of CIG barrier
-“should be performed”
Task-Network-Model
-Most CIGs model guidelines via TNM, which contains a flowchart of specific tasks
Arden Syntax
-language for encoding medical knowledge
-does not use TNM
-are collection of Medical Logic Modules (in which each MLM represents a single decision)
-most appropriate for simple alerts in reminder systems
-less than ideal for complex multi-step guidelines
Five rights of clinical decision support
-Information, Person, Format, Channel, Time
Types of alerts
-Interruptive (or modal)
-Non-interruptive (or modeless)
Leapfrog group
-Created in Nov 2000
-In response to the 1999 Institute of Medicine report “To Err is Human”
-Report underscored cost and frequency of adverse drug events
-Chose to address Computer Physician Order Entry (CPOE) as the first task, due to potential to lower patient harm from medications
Commission for Certification of Healthcare Information Technology
Is developing standards for CDS
Data-Information-Knowledge-Wisdom pyramid (funnel)
Data -> Information -> Knowledge -> Understanding -> Wisdom
-Made prominent by Russell Ackoff in 1989 but pyramid is recorded as fat back as 1934 was referenced in a song by Frank Zappa “Packard Goose”
-Emphasis in this pyramid is on knowledge being actionable
Knowledge modeling
-process of creating computer interpretable model of knowledge or standard specifications about a kind of process
-approaches include: clinical algorithms, bayesian statistics, production rules, scoring and heuristics
Clinical algorithm
-Approach for knowledge modeling
-systematic process through an ordered sequence of steps, with each step dependent on the outcome of the previous step
-Follows path through a flow chart
-Is a precursor to clinical guideline
Bayes theorem
-Approach for knowledge modeling
-probability of disease can be calculated from prior probability of disease and probability of findings occurring in disease
-But limited because findings in diseases are not conditionally independent and diseases are not mutually exclusive
Production rules
-Approach for knowledge modeling
-Encodes knowledge as “if-then” rules
-Backward chaining: starts with goal
-Forward chaining: follows defined path, algorithmic approach
MYCIN
-Developed by Edward Shortliffe’s PhD dissertation in 1970s
-Uses backward chaining deduction system to help physicians diagnose 2 types of bacterial infections
-Rule bases were large and difficult to main
-Took a long time to enter in information
Scoring and heuristics
-Approach for knowledge modeling
-Knowledge is represented as profiles of findings found in diseases
-more scalable than production rules
INTERNIST-1
-Example of system that used scoring and heuristics
-Responds to user with follow up questions to narrow field of possible diagnoses
-Used taboos (men can’t have ovarian cancer)
-But long learning curve and time consuming data entry
-Knowledge base was incomplete, and couldn’t construct differential diagnosis
DxPlain
-Another example of system that used scoring and heuristics, still in existence
As low as reasonably practicable
Software developers have responsibility to reduce risk of using CDS to a level that is as low as reasonably practicable
End-User Licensing Agreement
Many systems have EULAs which claim that software is provided as is without any claim or warranty of quality, fitness for purpose, completeness, accuracy or freedom from errors
Legal, ethical, and regulatory issues of CDS
information overload, copy and paste, order sets, alert fatigue privacy breaches
Fox’s 4 primary approaches to safety and quality in CDS
-Use rigorous software engineering to ensure reliability
-Systematic quality control for the medical component of the CDS
-Hazard management during system operation
-Comprehensive auditing to allow quality reviews
Hazards and Operability Analysis (HAZOP)
To analyze risk of CDS and stratify into 4 categories
-Risk Level 1: significant and avoidable hazards that could be caused by CDS
-Risk Level 2: no direct hazard associated, but beneficial intervention is overlooked (fails to warn physician to check lead levels)
-Risk Level 3: no direct hazard, fails to anticipate future conditions
-Risk Level 4: no identified risks
Public health informatics (7 elements of)
-planning and system design
-data collection
-data management
-analysis
-interpretation
-dissemination
-application to public health program
Syndromic surveillance
The continuous, systematic collection, analysis and interpretation of health-related data needed for the planning, implementation, and evaluation of public health practice
