Information Technology Reference
In-Depth Information
2.2
Data Mining in Healthcare
In the healthcare sector, DM is becoming more popular and essential. Healthcare
institutions face an increasing pressure to improve the quality of the healthcare while
reducing costs [7]. Moreover, huge amounts of data are collected every day by these
institutions and their complexity and volume makes them hard to analyze. DM is
capable to transform these data into useful information for decision-making [4].
Therefore, it is well suited to provide decision support in healthcare [7].
DM is capable to make predictions that can be useful in the medical and
biomedical context. The main goal of these predictions is decreasing the level of
subjectivity in the clinical decision-making process and improving the daily workflow
of the healthcare institution. The predictive models produce excellent knowledge to
support the healthcare professionals' work [8]. Today DM is used in a large number
of applications that have included both DM and Clinical Decision Support System
(CDSS) and it attracts considerable attention. It can be used in several contexts to
solve several problems: it can be used by healthcare insurers to help to detect fraud
and abuse; it can be used to help administrative decisions; it can be used to identify
best practices and effective treatments [4]. Healthcare organizations want DM to
improve physician practices, resource utilization as well as disease management [7].
3
Data Mining for Nosocomial Infection Prediction
DM technology can be applied to healthcare in order to build predictive models
providing predictions in real environments using, for that, real clinical data. In the
study of nosocomial infections it is important to know when these infections can
occur and that can be performed through the application of DM techniques to predict
the probability of the occurrence of an infection in the presence of certain variables.
Therefore, this DM study intends to develop DM models capable to classify a patient
as able to acquire or not a nosocomial infection according to the risk factors that
describe his/her clinical condition. It allows exploring the relationship between the
occurrence of infections, their characteristics and risk factors presented in the patients
hospitalized.
3.1
CRoss Industry Standard Process for Data Mining
There are several methodologies to implement the KDD process but the most used is
the CRoss Industry Standard Process for Data Mining (CRISP-DM). The DM study
presented in this work was performed using this methodology and the stages of this
methodology will be presented next.
It is important to emphasize that the sequence of the CRISP-DM stages is not rigid
because it is frequently necessary to return to previous steps since the result of one
stage determines the action to implement in the next stage. Besides that, the KDD
process is cyclical and does not finish with the last stage of the CRISP-DM because
the knowledge obtained during the CRISP-DM process can trigger new business
questions, frequently more focused, that force the cycle to restart again [9].
Search WWH ::
Custom Search