C4.5: programs for machine learning
C4.5: programs for machine learning
Workflow mining: a survey of issues and approaches
Data & Knowledge Engineering
GLIF3: a representation format for sharable computer-interpretable clinical practice guidelines
Journal of Biomedical Informatics
The SDA Model: A Set Theory Approach
CBMS '07 Proceedings of the Twentieth IEEE International Symposium on Computer-Based Medical Systems
Temporal Constraints Approximation from Data about Medical Procedures
CBMS '07 Proceedings of the Twentieth IEEE International Symposium on Computer-Based Medical Systems
A Causal Modeling Framework for Generating Clinical Practice Guidelines from Data
AIME '07 Proceedings of the 11th conference on Artificial Intelligence in Medicine
Mining hospital data to learn SDA* clinical algorithms
AIME'07 Proceedings of the 2007 conference on Knowledge management for health care procedures
Generating macro-temporality in timed transition diagrams
AIME'07 Proceedings of the 2007 conference on Knowledge management for health care procedures
Detecting dominant alternative interventions to reduce treatment costs
KR4HC'11 Proceedings of the 3rd international conference on Knowledge Representation for Health-Care
Editorial: Bayesian networks in biomedicine and health-care
Artificial Intelligence in Medicine
BPM' 2012 Proceedings of the 2012 international conference on Process Support and Knowledge Representation in Health Care
Hi-index | 12.05 |
Objective: To propose a methodology to automatically induce state-decision-action diagrams from health-care databases and electronic health records in order to show health-care professionals an explicit representation of the past health-care procedures carried out in a health-care organization and to use these representations to study the deviations with respect to official and predefined protocols and clinical algorithms. Materials and methods: The methodology is based on two data and knowledge structures: episode of care database and set of rules. These two structures contain, respectively, patient data from health-care centres and the translation rules which are used to adapt the data of the episode of care database to the terminology we want the resulting state-decision-action diagram to have. The data expressed in the new terminology is used to generate the final state-decision-action diagram by means of a machine learning method. We have performed several tests on the treatment of hypertension with data from the SAGESSA Health-care Group in Spain. The state-decision-action diagrams obtained have been analyzed at the level of their ability to predict correct treatments and at the level of their adherence to the clinical algorithms published by four official health-care organizations. Results: The state-decision-action diagrams obtained represent an average 94.6% of the treatments in the database, only excluding some atypical cases. Moreover, these diagrams show a high level of adherence to the treatment proposed by the National Heart Foundation of Australia and the Spanish Society for Hypertension with about 90.4% of coincident treatment. Conclusions: A new methodology has been developed and validated which automatically induces state-decision-action diagrams which can be used as a graphical representation of the health-care procedures carried out in health-care organizations. The methodology is also a tool to study the adherence of these health-care procedures to the official standards.