INTEGRATING CASE-BASED REASONING WITH
EVIDENCE-BASED PRACTICE FOR DECISION SUPPORT
Expedito Carlos Lopes and Ulrich Schiel
Computing and Systems Department, Federal University of Campina Grande, Brazil
Keywords: Case-based reasoning, Evidence-based practice, Decision support.
Abstract: Evidence-Based Practice (EBP), an emergent paradigm, uses the premise that decision making can be based
on scientific proofs available in reliable data bases, usually found on sites over the Internet. However, the
procedures of the EBP do not provide mechanisms for retention of information and knowledge strategic of
the individual solutions, which could facilitate the learning of different end-users, in the future. On the other
hand, Case-Based Reasoning (CBR) uses the history of similar cases to support decision making. But, the
retrieval of cases may not be sufficient to give support to the solution of problems. Since both research
evidences as well as similar cases are important for decision-making, this paper proposes the integration of
the two paradigms for problem-solution support, regarding complex problems. An example of the justice
area is presented.
1 INTRODUCTION
Problem is the main element that triggers the process
of decision making. When it presents difficulties
associated with the absence of available solutions, or
even when it demands a great effort to understand
why it happens or which are its causes, it is
considered a complex problem (Loriggio, 2002).
In 1992, emerged the Evidence-Based Medicine
(EBM) that has its primary focus on provide
effective counseling to help patients with terminal or
chronic illness to make decision in order to cure of
illness, extend or increase the quality of their life
(Sacket et al., 2001).
Supported in the EBM, was standardized the
Evidence-Based Practice (EBP) that uses the
premise that decision making can be based on
scientific proofs available in reliable data bases,
usually found on sites over the Internet, derived
from primary studies (e.g. randomized clinical trials)
and/or secondary studies (systematic review and
meta analysis) carried out by independent research
groups (e.g. Cochrane Collaboration on medical
area, and Campbell Collaboration concerning crime
& justice and education areas). EBP on the medical
area involves complex and conscientious decision-
making, based not only on the available evidences
but also on patient characteristics, situations, and
preferences (Sackett et al., 2001). EBP has expanded
itself to the healthcare area in general.
Although with some alterations, this paradigm
presents itself in several different areas, such as
criminal justice, including politics for crime
prevention, considering the potential offensive risk
of the defendants (Warren, 2007).
On the other hand, Case-Based Reasoning
(CBR) uses the history of similar cases to support
decision making (Pal and Shiu, 2004), instead of
using research evidences. CBR fails when no similar
case can be found in the Case-Base. Moreover, the
retrieval of cases may be not sufficient to give
support to the solution of complex problems.
But, EBP procedures do not provide mechanisms
for retention of information and knowledge strategic
of the individual solutions, which could facilitate the
learning of different end-users, in the future, and
preserving evidences used, since they can be
modified or removed of the Internet later.
Since both research evidences and similar cases
are important for decision-making, the integration of
the two paradigms constitutes an interesting research
topic to the solution of complex problems. Thus,
complex cases and research evidences can be
expressed together in a knowledge representation
and the integration of EBP and CBR techniques is
considered as support for the development of
applications that encompass decision making.
368
Carlos Lopes E. and Schiel U. (2009).
INTEGRATING CASE-BASED REASONING WITH EVIDENCE-BASED PRACTICE FOR DECISION SUPPORT.
In Proceedings of the 11th International Conference on Enterprise Information Systems - Artificial Intelligence and Decision Support Systems, pages
368-371
DOI: 10.5220/0002007403680371
Copyright
c
SciTePress
Figure 1: Basic Cycle of Evidences and Cases Processing.
2 INTEGRATING EBP
PROCEDURES WITH CBR
FOR DECISION SUPPORT
In the EBP, according to Sackett (2001), what is
objectively searched is "the integration of best
evidence from research, clinical skill and preference
of the patient." Clinical knowledge and experience
are used to identify the health state and diagnosis of
patients, their individual risks and benefits of
proposed interventions. Moreover, the patients worth
about preferences regarding private, personal
concerns and expectations brought to the
appointment are considered in clinical decisions.
Systematically, the EBP includes the steps:
1. Transformation of the problem into a question to
be answered;
2. Identifying the best evidence to answer the raised
question;
3. Critical analysis of the evidence in order to
validate its applicability and its impact in relation
to the question;
4. Integrating clinical skills, values and cultural
aspects of the patient related to the previously
extracted procedures;
5. Evaluation of the efficiency and effectiveness in
the results of each step.
In crime prevention, the concept of EBP is defined
as: the correlations practice that has been proven
through scientific research aimed to reduce the
recidivism of offenders. In Warren (2007), EPB
principles used by judicial units in order to reduce
criminal recidivism are presented. EBP considers the
risk and needs principle of the offender, use of
risk/needs assessment instruments, integration of
treatment and community-based sanctions.
In CBR, the approach to get a solution for a
certain problem is based on a comparative analysis
of previous realities with a new similar reality. There
is an analogical reasoning process based on the
degree of similarity in form of deductive inference
(Pal and Shiu, 2004; Russel and Norvig, 2003).
CBR and EBP have common aspects such as (i)
use of procedures to support decision-making; (ii) be
based on a problem presented by a motivator agent;
(iii) use of search and similarity measures involving
the current situation and previously stored situations;
(iv) they presents the results in ranking schemes.
The two approaches can be considered
complementary for a good decision-making. While
CBR takes advantage retrieving the history of
previous cases, whereas EBP aims finding
documents containing updated evidence based
consolidated research results (see Figure 1).
INTEGRATING CASE-BASED REASONING WITH EVIDENCE-BASED PRACTICE FOR DECISION SUPPORT
369
While CBR search uses a structured case-base,
usually presented in the organizational environment
of the problem, the EBP approach searches evidence
in unstructured document usually obtained over the
Internet.
We describe the integration preserving
compatibility with the classic model on CBR.
1. Retrieve cases - Obtain cases with similar
problems to new case and antecedent cases of the
motivator agent. In the retrieve the search is textual
and the key terms in the new case are confronted
with the contents of the attached evidence, in
addition to the content of the founded cases. The
similarity calculation considers the attached
evidence with its associated weight.
2. Retrieve and process research evidence – It
contains four sub-activities:
2.1 - Transform problem in question – it aims to
build a template, from a given problem, to conduct
the research for documents containing evidence;
2.2 - Evidence Retrieval – obtain documents that
contain evidences considering the template
previously made. It is based on Information
Retrieval techniques involving similarity metrics and
ranking schemes;
2.3 - Critical Evaluation – this sub-activity includes
steps for the selection, the rejection and the
acceptance of documents, besides of evidences
extraction;
2.4 - Selects the best research evidence – involves
the consolidation of evidences showing result
presentation and discussion, considering aspect
several, e.g., practical decisions, methodological
limitations, quality of original material, etc. (Muñoz
et al., 2002).
3. Reuses – The first reuse activity treats the
construction of new solutions, with mechanisms for
adjustments and adaptations, considering (i) similar
cases solutions (or part of them), (ii) historical of
motivator agent cases, and (iii) the best research
evidence confronted with attached evidences to
similar cases. It corresponds to classic solution on
CBR. The second reuse activity must collect the
values and preferences of the motivator agent and
request his/her participation in decision-making.
Reuse examples: (i) medical diagnosis and
treatment; (ii) judgment and designation of
activities/service institutions to penal execution.
4. Revise and Retain – The same of the classic
model on CBR.
3 EXAMPLE – APPLICATION TO
CRIMINAL JUSTICE
This example is applied on a real case of abortion
from the State of Bahia Court, Brazil, whose
practice is considered a crime. The subject of the
new case is “voluntary abortion”.
The Table 1 contains the general evidence and
case structure for criminal justice. A prototype was
developed in Java language that interacts with the
framework JColibri using the CBR model (Bello-
Tomás, 2004).
The Figure 2 presents data of the problem (fact
and his/her author, witness and material proofs).
Figure 2: Data of the problem.
Table 1: Evidence and Case Structure.
Structure Attributes
Problem (fact) Description, Objective, Circumstances, Author (Name, Adress, District, Conduct, Behaviour,
Needs, Risk condition, Preferences), Witness proof (Name, Content, Date), Material proof
(Organization, Content, Date, Responsible person)
Research
Evidence
Question, Defined terms, Keywords for search, Sources, Search criteria, Result (Type of study,
Considerations), Discussion
Solution (trial) Sentence, Criteria, Recommendation (Permit, Institution)
Outcome Final situation, Consequences, Final comments
Learning Procedures to case solution, Comments search for evidences
ICEIS 2009 - International Conference on Enterprise Information Systems
370
Figure 3: Retrieved case for judgment support.
One similar case about therapeutic abortion was
found in the Evidences and Cases Base, but it isn’t
sufficient to give support to the solution, forcing the
judge to search by research evidences (see Figure 3).
The activity Retrieve and process research
evidence began with the transformation of the
problem in a question. Initially, defined terms,
keywords, sources and search criteria are informed.
After analysis of the documents retrieved, the
history of retrieve, the results and discussion are
informed agree with the Figure 4.
Figure 4: Template of the search for evidence.
The first activity reuse was based in evidences
attached to the case, discussion, and the court
sentence was in favour of the "voluntary interruption
of pregnancy", based on criteria of the Brazilian
Penal Code. Aiming to learning, procedures to case
solution describe the steps that it took to reach this
verdict coming to be used in the activity retain. In
the second activity reuse, the preferences of the
pregnant woman were agree with the list of
indicated hospitals (institutions) and medical
surgeons of the court. In the decision with
recommendation, the court attended the
recommendation of the concerned pregnant. It was
consigned a charter authorizing the surgery (permit)
as soon as possible and awaited the final outcome.
4 CONCLUSIONS
This work proposes the integration of CBR with
EBP for decision support. The activities of the basic
cycle of CBR and the systematic steps of the EBP in
the medical and justice areas served of bases for
built this propose.
The main contributions are: (i) representation of
the EBP procedures; (ii) knowledge representation
to retain cases and evidences together; (iii) extension
of CBR regarding evidences attached to the cases;
and (iv) the representation of the EBP procedures
integrated to CBR.
Future researches encompass the formalization
of the computing model towards Evidence and Case-
Based Reasoning and the creation of a semi-
automatic Evidence-Oriented Information Extractor.
REFERENCES
Bello-Tomás, J. J., Gonzáles-Calero, P. A., Diaz-Agudo,
B. (2004). JColibri: An Object-Oriented Framework
for Building CBR Systems. ECCBR (pp. 32–46).
Madrid, Spain.
Loriggio, A. (2002). De onde vêm os problemas. São
Paulo, Brazil: Negócio Editora.
Muñoz, S.I.S., Takayanagui, A.M.M., Santos, C.B.,
Sanchez-Sweatman, O. (2002). Systematic literature
review and meta-analysis. Proc. 8th Brazilian Nursing
Communication Symposium. São Paulo, Brazil.
Pal, S., Shiu, S.C.K. (2004). Foundations of Soft Case-
Based Reasoning. New Jersey: Wiley-Intersciense
publication.
Russel, S., Norvig, P. (2003). Artificial Intelligence: A
Modern Approach. New Jersey: Prentice Hall.
Sackett, D.L., Straus, S.E., Richardson, W.S., Rosenberg,
W., Haynes, R. B. (2001). Evidence-based medicine:
how to practice and teach EBM. New York: Elsevier
Health Sciences.
Warren, R.K. (2007). Evidence-Based Practice to Reduce
Recidivism: Implications for State Judiciaries.
Available: http://works.bepress.com/roger_warren/1.
INTEGRATING CASE-BASED REASONING WITH EVIDENCE-BASED PRACTICE FOR DECISION SUPPORT
371