Application of Data Mining Techniques to Predict Urinary Fistula Surgical Repair Outcome: The Case of Addis Ababa Fistula Hospital, Addis Ababa, EthiopiaMinale Tefera1, Mitike Mola2, Getachew Jemaneh3 and Feleke Doyore4*
- *Corresponding Author:
- Feleke Doyore
Lecturer and researcher, Department of Public Health
Faculty of Medicine and Health Sciences
Wachemo University, Hossana, Ethiopia
Tel: +251916291489, 0932685424
E-mail: [email protected]
Received date: March 06, 2014; Accepted date: April 21, 2014; Published date: April 23, 2014
Citation: Tefera M, Mola M, Jemaneh G, Doyore F (2014) Application of Data Mining Techniques to Predict Urinary Fistula Surgical Repair Outcome: The Case of Addis Ababa Fistula Hospital, Addis Ababa, Ethiopia. J Health Med Informat 5:153. doi: 10.4172/2157-7420.1000153
Copyright: © 2014 Tefera M, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Background: Maternal outcomes are good in most countries of the developed world while the same is not true in many developing countries. The likelihood of the occurrence of incontinence after successful surgical repair makes predicting urinary fistula surgical repair outcome is important for decision making during treatment and follow up. Therefore, this research is aimed to apply data mining techniques to build a model that can assist in predicting surgical outcome of urinary fistula repair based on clinical assessments done just before surgical repair.
Methods: The six-step hybrid knowledge discovery process model is used as a framework for the overall activities in the study. 15961 instances that have undergone urinary fistula repair in Addis Ababa Fistula Hospital are used for both predictive association rule extraction and predictive model building. Apriori algorithm is used to extract association rules while classification algorithms J48, PART, Naïve Bayes and multinomial logistic regression are used to build predictive models. Support and confidence are used as interestingness measure for association rules while area under the WROC and ROC curve for each specific outcome is sequentially used to compare performances of models from the predictive algorithms.
Results: Predictive association rules from Apriori have shown frequent co-occurrence of less severity of injury with cured outcome. The predictive model from PART-M2-C0.05-Q1 scheme has shown an area under WROC curve of 0.742. Area under the ROC curve for residual outcome (ROCResidual=0.822) from this algorithm is better than Naïve Bayes and logistic, while the areas under the ROC curves for the other outcomes are greater than the model from J48.
Conclusion: Predictive model is developed with the use of PART-M2-C0.05-Q1. The predictive association rules and predictive model built with the use of data mining techniques can assist in predicting urinary fistula surgical repair outcome. Therefore, it is better in detecting residual outcome than the logistic regression model.