No abstract

Background: Unwanted pregnancy is the kind of pregnancy which is undesirable for at least one of the parents, and is accompanied by unfavorable consequences for the family and society. Objectives: In this study, three classification models have been used to predict the occurrence of unwanted pregnancies in the urban population in Khorramabad, Iran, and the performance of these models was compared. Methods: In this cross-sectional study, 467 multiparous mothers referred to the health centers of Khorramabad in 2012 were selected using a combination of cluster and stratified sampling, and the relevant variables were measured. The logistic regression, decision tree, and a neural network were implemented using SPSS version 21 and MATLAB version R2013a. To compare these models, the indices of sensitivity and specificity, the area under the ROC curve, and the correct percentage of the predictions were used. Results: Overall, the prevalence of unwanted pregnancies was 32.3%. The performance of the models based on the area under the ROC curve as the indicator was as follows: artificial neural networks (0.741), decision tree (0.731), and logistic regression (0.712). The highest sensitivity level belonged to the decision tree (73.5%), and the highest specificity level belonged to the artificial neural network (62.3%). Conclusions: Given the high prevalence of unwanted pregnancies in Khorramabad, Iran, it is necessary to revise and improve the family planning projects. In selecting the best classification method, if the researcher is interested in the better interpretability of the results, the use of the decision tree and logistic regression is recommended; however, if the researcher is interested in a higher prediction power of the model, the neural network is recommended.

Using simultaneous Bayesian modeling, an attempt is made to analyze data on the size of lymphedema occurring in the arms of breast cancer patients after breast cancer surgery (as the longitudinal data) and the time interval for disease progression (as the time-to-event occurrence). A model based on a multivariate skew t distribution is shown to provide the best fit. This outcome was confirmed by simulation studies too.