Invited presentation on "Conformance Verification when Dealing with Computerized and Human-Enhanced Processes" at the Workshop on Foundations of Biomedical Knowledge Representation (FBKR 2012), Lorentz Centre, the Netherlands.
DGT @ CTAC 2024 Valencia: Most crucial invest to digitalisation_Sven Zoelle_v...
FBKR 2012 - Montali - Conformance Verification when Dealing with Computerized and Human-Enhanced Processes
1. Stefano Bragaglia1, Federico Chesani1,
Paola Mello1, Marco Montali2
1DISI, University of Bologna
2KRDB, Free University of Bozen-Bolzano
Workshop on Foundations of Biomedical Knowledge Representation
01/11/2012 Lorentz Centre, Leiden
2. –
™CGs developed by applying evidence-based
medicine to large classes of abstract patients
™Assumptions
– Ideal patients
™ statistically relevant
™ with only the disease targeted by the CG
– Ideal physicians
– Ideal resources
™ ∞ resources
Ideal world
3. –
™Context and patients are not ideal
– Resources may be missing
– Each patient has her own story, condition, preferences
à Unforeseen situations are common
™CGs routinely adapted on a per-patient basis,
using the Basic Medical Knowledge (BMK)
™CGs enacted together with many additional
(local) rules and processes
™Physicians are not ideal
(maybe, they would need computerized support J )
Real World
4. –
™ Compliance The act of conforming as requested by the CG
™ Flexibility The ability of accommodating and promptly
adapting to change and unforeseen situations
Compliance vs Flexibility
Universe of Traces
Compliant
Traces
Compliance
Flexibility
5. –
™CGs propose a recommended behavior
™Many factors could lead healthcare professionals
in taking a different behavior
™We need to sort this discrepancy out!
™Goal of conformance checking: detect deviations
between the expected and the actual behavior
–I.e., provide to domain experts all useful information
to understand and explain these deviations
Conformance
6. –
™Not to be intended as a normative component
™“Global” usage (totality of cases): CGs
understanding and improvement
– Improvement of the organization
– Improvement of the CG model
™“Local” usage (single patient): decision support
– Track the state of affairs
(where is the patient located wrt the CG?)
– Report deviations
– Run-time and offline perspective
Usages of Conformance
8. –
™ Definitions and terminology, to describe terms and
applicability conditions of the CG
™ Workflows, characterizing the allowed courses of execution
™ Rules, to handle particular cases and exceptions, and
declarative fragments
™ Linguistic labels to explain features, conditions, criteria
– “Low”, “high”, “risky”, …
™ Temporal constraints (metric, repetitions, …)
– In addition to the ones imposed by workflows
What is a CG Model?
9. –
™ Interplay between CGs and BMK
– Complex interaction:theycan defeat each other depending
on the specific situation
– “Closed” vs “open” fragments of the CG
– Doctors always have the last word!
™ Interplay between workflows and rules
– Workflows: procedural knowledge
– Rules: declarative knowledge
™ Humans in the loop
– They are not web services!
– Missing a deadline for 50 ms is actually a deviation?
à “Grades” of conformance
Criticalities
13. –
™ Activities are
connected to an
expected lifecycle
– Internal states of
activities
– Transitions
triggered by
atomic events
Intra-Activity Perspective
active
completed
start
end
candidate
14. –
™ Correlation of
events to the
corresponding
lifecycle
™ “Next-transition”
expectation
™ Generation of
corresponding
“activity state”
fluent
Intra-Activity Conformance
active
completed
start
end
candidate
15. –
Inter-Activity Perspective11
Table 1. Basic workflow patterns in GLARE, and their corresponding enabling con-
ditions
Pattern Representation Enabling conditions
Sequence
A B
When A is completed, B becomes candidate
Exclusive choice
A
B
C
cond
else
When A is completed and cond holds, B becomes
candidate
When A is completed and cond does not hold, C
becomes candidate
Simple merge B
C
D
When B is completed, D becomes candidate
When C is completed, D becomes candidate
Parallel split
A B
C When A is completed, B and C become candidate
Synchronization
DB
C
When B and C are completed, D becomes candi-
date
16. –
™ Generation of “candidate” activity instances
– Todo list
™ Progression principle
– Every candidate activity is expected to be started
™ Enforcement of “closed” procedural knowledge
– Every non-candidate activity is expected not to be
started
– What about exceptions? (see next slide)
™ Closed time-window: every executed activity must
be completed before the end of the trace
Inter-Activity Conformance
18. –
Formalize the refinedmodel towards conformance checking
Refine CGs (GLARE) to accommodate BMK
Understand how CGs are interpretedby healthcare professionals
Collectingreal examples about BMK+CGs
Research agenda
[with Terenziani’s group]
19. –
™ Both BMK and CG may involve declarative and
procedural knowledge
™ Procedural knowledge fixes the sequencing of
actions to be done
™ Declarative knowledge captures constraints and
properties to be satisfied, without saying “how”
CG+BMK: Example
CG in GLARE [Terenziani et al.] BMK
Threats to patient’s life must be
addressed immediately.
An hearth failure is a life threat.
Diuretic therapy is a possible immediate
response for acute heart failure.
Electrocardiographic
study
Echocardiographic
study
Coronary
angiorgraphy
20. –
™ The interplay between the two kinds of knowledge
occurs at execution time
™ Brainstorming with physicians led to a specialized
activity life cycle
– Capturing the semantics of “executing activities” from
the viewpoint of domain experts
– Pointing out where BMK-driven decision making
comes into play
– Showing that data are as much as important as the
process
Binding CG with BMK
21. –
™ BMK
– Eligibility checks
(preconditions)
– Abnormality
checks to identify
exceptional cases
™ Before the
activityexecution
™ During the
activityexecution
Revised Lifecycle
ready
candidate
active
completed
discarded aborted
preconditions
∧ ¬abnormal
else
start
end
failure
∨ abnormal
22. –
Conformance with CG+BMK
™ Ready and candidate states collapsed
™ Expected life cycle à triggered by logical conditions
™ Real life cycle à triggered by event occurrences
™ Conformance: detect and show deviations
expected real
candidate
active
completed
discarded aborted
start end
failure
∨ abort
abort
ready
candidate
active
completed
discarded aborted
preconditions
∧ ¬abnormal
else
start
end
failure
∨ abnormal
23. –
™ Proposed in 1986 by Kowalsky and Sergot
™ Events
™ Fluents, i.e. properties whose truth value can change
along time
™ Domain axioms: link the happening of events with
the change of truth value of fluents
Representing the current
state: Event Calculus
24. –
The Simple EC Ontology
1 2 3 4 5 6 7 8 9 10 11 12 13 14
initiates(a,f,3) terminates(b,f,12)
happens(a,3)
holds_at(f,7)
declip
clip
0
f
f holds in (3,12]
a b
25. –
An example…
17
Fig. 4. EC-based conformance evaluation of a CG execution.
• Reification of deviations as special fluents
• Expectations not explicitly represented
26. –™ Events
– Somethinghappened (what)
– At a time point (when)
™ Fluents
– Properties/status of the system
– Affectedby events
™ Expectations
– About events
– About fluents
™ Achievement properties (existentially quantified)
™ Maintenance properties (universally quantified)
– Only positive vs. positive/negative expectations
Declarative Conformance:
few concepts
27. –
™ Matching function: return a score if an observed event
matches any (positive/negative) expectation
™ Should support different semantics
– Ontologies
– Fuzzy concepts
– Temporal reasoning
™ Fulfillment
– an event matching a positive expectation has happened
– No event matchingnegative expectation has happened
– Achievement/maintenance propertiesare treated almost
similarly…
Events, fluents, expectations
and…
28. –
™ Violation
– an event matching a positive expectation did not
happen
– An event matching negative expectation has happened
– Fluents: strong negation vs. weak negation, in case of
maintenance properties
Events, fluents, expectations
and…
29. –™ Work in progress!!!
™ Based on Drools/Java and Drools Chance
CG representation and
expectations: ECE rules
18
rule "Risk factor evaluation "
when
$pat : Patient( ... ) // patient identifier
// evaluation of risk factor and confidence degree
$risk : EvaluatedRisk ( $phys , $pat , $disease , $factor , $conf )
$factor == "high"
$conf >= "medium"
then
expect InitiateTreatment ( $pat , $disease , this after [0,1 hour] $risk )
on fulfillment { // if the treatment is initiated
/* some increase in patient health */
}
on violation { // if the treatment is not initiated
alert( ... );
}
end
Fig. 5. An example of ECE-Rule [4].
30. –
ECE rules…
native matching mechanism supported by Drools
derived by fuzzy ontologies.
rule "Fuzzy evaluation of conformance "
when
Order ($e: expectedInDays )
DeliveryLog (
$d: delay ~InTime $e
, @imperfect(kind =="userOp")
$p: packaging nec ~isA " GoodPackaging ")
then
println("Degree of Delivery Conformance : " +
Drools.degree);
end
Fig. 6. A rule that checks the conformance of a delivery