Endometriosis and Its Correlation with Lifestyle Factors and Health Indicators : A Data Mining Approach Using R and Python

Around 10% of women in fertile age have endometriosis, despite this there is little known about its origin. It can take years from the first experienced symptoms to an established diagnosis, which is done using invasive methods. A database from the Lucy application containing 11720 questionnaires with up to 41 questions each was examined for signs of associations between lifestyle factors and health indicators with endometriosis. The database was restructured and unusable questionnaires were pruned, leaving 5719 questionnaires for analysis. Questionnaires were assigned a label depending on whether the respondent had an endometriosis diagnosis, some other diagnosis or no diagnosis. Histograms were created for 36 questions to visualize answer distributions among the labels while wordclouds were created for 3 questions to determine commonly used words. Pearson’s chi-square test was performed on 24 questions to determine whether there were statistically significant differences in how respondents answered with regards to their diagnosis. Then, 26 questions were divided between seven question groups based on their similarities. The relative prevalence of endometriosis for every answer alternative was determined and compared withthe average for that group. This was then followed up by a multi-correspondence analysis for every question group. It was found that there are associations between endometriosis and 12 health indicators. These being: affected by infertility, always being tired, regarding own health as bad, having more than 100 birthmarks, severe chronic pelvic pain, dysmenorrhea, regularly whole-body pain, muscle stiffness, neck and shoulder tension, forgetfulness, concentration difficulties, regularly experiencing that stress worsens physical symptoms. No statistically significant association between lifestyle factors and endometriosis could be found in the survey database. There were findings of socioeconomic differences between respondents with endometriosis and those without a diagnosis which may have affected the result.