Prediction of Anti-Cancer Medication Discontinuation via Patient Portal Messages and Structured Electronic Medical Records

点击翻译按钮获取中文摘要

Summary Cancer is a leading cause of morbidity in the United States, with more than half a million deaths estimated in 2019. Systemic cancer therapies are increasingly being designed as long-term oral anti-cancer medications, given the increased convenience of a self-administered regimen. For instance, patients with operable hormone-receptor-positive breast cancer are prescribed adjuvant oral hormonal therapy, with an expectation that they continue their regimen for a minimum of 5 years to maximize the benefits. Although many oral therapies have proven effective in mitigating cancer recurrence and mortality, discontinuation to these treatments are not uncommon. This is a concern because medication discontinuation before the completion of a prescribed treatment protocol leads to lower survival rates, increased risks of recurrence, and higher healthcare costs. To improve treatment adherence and promise better healthcare delivery, it is essential for healthcare providers to know when and why a cancer patient will discontinue their medications. While there have been various investigations into regimen discontinuation, the focus of these studies is either on knowledge discovery or intervention. While knowledge discovery focuses on characterizing the potential factors that lead to medication discontinuation, intervention aims to leverage discovered knowledge to design and test effective strategies to help patients adhere to treatments. Because there are thousands of cancer patients, it is impossible for healthcare providers to apply intervention to each of them. Limited medical resources need to be allocated efficiently, such that patients with a higher risk of discontinuing medications will receive greater, timely attention. Yet, the increasing integration of online communication and mobile computing technologies into the healthcare domain are generating massive quantities of patient-generated information. Thus, we propose to apply online patient-provider communications in a patient portal to supplement traditional EMR data to better understand a cancer patient’s medical experience. The central hypothesis of this project is that such communications together with structured EMRs can be applied to learn and forecast oral anti-cancer medication discontinuation. The specific aims of this project designed to test our central hypothesis are to 1) discover what has been communicated in a patient portal; 2) infer how patient portal messages and structured EMRs are associated with medication discontinuation; and 3) determine who are more likely to discontinue medications. To the best of our knowledge, this is the first study to apply the messages in a patient portal and structured EMRs to investigate medication discontinuation for cancer patients.