Causal Networks for Modeling Health Technology Utilization in Intensive Care Units
This study presents the application of Bayesian networks (Bn) to explain Neonatal Intensive Care Unit relationships. Information was compiled retrospectively from the medical records at two neonatal intensive care units of 523 neonates (63 deaths). A total of 31 variables were used for the model, eleven to characterize admission conditions and severity of illness as well as the 20 technologies. With mortality as the output variable, the K2 search algorithm and Geiger-Heckerman quality measures were used in the training that generated the Bn. Evidence propagation was used to assess the training, which yielded a sensitivity of 77.78% and a specificity of 91.30%, in the classification of mortality. Clinical criteria, correlations and logistical regression were used to analyse the relationships the model provided. The Bn found clinically coherent relationships as recognizable conditions that directly affect mortality such as congenital malformations are seen and it exposes the least effective technologies among those studied, bicarbonate treatment.
Unable to display preview. Download preview PDF.
- 4.Panerai, R.B., Chacón-Pacheco, M.L., Almeida, R.T.: Path Analysis in Health Technology Assessment. In: International Society of Technology Assessment in Health Care, Seventh Ann. Meet., Helsinki, Finland (1991)Google Scholar
- 8.Sierra, B., Serrano, N., Larranaga, P., Plasencia, E.J., Inza, I., Jimenez, J.J., Revuelta, P., Mora, M.L.: Using Bayesian networks in the construction of a bi-level multi-classifier. A case study using intensive care unit patients data. Artif. Intell. Med. 22(3), 233–248 (2001)CrossRefGoogle Scholar
- 9.Heckerman, D.: Bayesian networks for knowledge discovery. In: Fayyad, U.M., Piatesky-Shapiro, G., Padhraic, S., Uthurusamy, R. (eds.) Advances in Knowledge Discovery and Data Mining, pp. 273–305. Mit Press, Melo Park (1996)Google Scholar
- 10.Castillo, E., Gutiérrez, J.M., Hadi, A.S.: Expert systems and probabilistic network models. Springer, New York (1997)Google Scholar
- 11.Gaiger, D., Heckerman, D.: A characterization of the Dirichlet distribution with application to learning Bayesian networks. In: Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence, pp. 196–207. Morgan Kaufmann Pub., San Francisco (1995)Google Scholar