Patient-oriented Data Mining Approach for Chronic Disease

Background: In recent decades, chronic diseases appear increasingly in the list of top ten leading causes of death. Therefore, many studies were conducted to discover risk factors for populations by engaging data mining techniques. Even though such studies include patients consent for use in data analysis, the results of studies are most likely for research purposes only and there are limited studies reporting how patients understand those results. Objective: Our objective is to develop a patient-oriented data mining approach which could help patients and their health professionals better understand and retrieve the mining results. This approach could bridge the gap between researchers and patients to better increase population-based awareness of the development of the chronic condition. Methods: As a pilot study, we collected measurement data and symptom data of breast cancer survivors, who are at-lifetime-risk of developing lymphedema. We then applied a data mining algorithm and fed the results into a visualization module. Results: Instead of searching a huge pool of data mining results, patients and their health professionals could select options from different categories, such as percentage of limb volume change, and retrieve the results more specifically relevant to them. Discussion/Conclusion: This approach could integrate patient demographic and treatment data for future work. Therefore, it could provide better personalized data mining results based on data from different groups of patients.