Using Statistics and Data Mining Approaches to Analyze Male Sexual Behaviors and Use of Erectile Dysfunction Drugs Based on Large Questionnaire Data.

Zhi Qiao,Xiang Li,Haifeng Liu,Lei Zhang,Junyang Cao,Guotong Xie,Nan Qin,Hui Jiang,Haocheng Lin
DOI: https://doi.org/10.3233/978-1-61499-753-5-206
2017-01-01
Abstract:The prevalence of erectile dysfunction (ED) has been extensively studied worldwide. Erectile dysfunction drugs has shown great efficacy in preventing male erectile dysfunction. In order to help doctors know drug taken preference of patients and better prescribe, it is crucial to analyze who actually take erectile dysfunction drugs and the relation between sexual behaviors and drug use. Existing clinical studies usually used descriptive statistics and regression analysis based on small volume of data. In this paper, based on big volume of data (48,630 questionnaires), we use data mining approaches besides statistics and regression analysis to comprehensively analyze the relation between male sexual behaviors and use of erectile dysfunction drugs for unravelling the characteristic of patients who take erectile dysfunction drugs. We firstly analyze the impact of multiple sexual behavior factors on whether to use the erectile dysfunction drugs. Then, we explore to mine the Decision Rules for Stratification to discover patients who are more likely to take drugs. Based on the decision rules, the patients can be partitioned into four potential groups for use of erectile dysfunction: high potential group, intermediate potential-1 group, intermediate potential-2 group and low potential group. Experimental results show 1) the sexual behavior factors, erectile hardness and time length to prepare (how long to prepares for sexual behaviors ahead of time), have bigger impacts both in correlation analysis and potential drug taking patients discovering; 2) odds ratio between patients identified as low potential and high potential was 6.098 (95% confidence interval, 5.159-7.209) with statistically significant differences in taking drug potential detected between all potential groups.
What problem does this paper attempt to address?