Using Artificial Technology to Provide Targeted Adherence Messaging to Patients Using Oral Chemotherapy Agents

Author: Daniel Tomaszewski

The advent and expanded availability of oral oncolytics has revolutionized the treatment of cancer. The diverse mechanisms of actions and treatment targets of novel oral oncolytics have expanded availability of treatment to otherwise terminal cancers. This is particularly true in the treatment of systemic and metastatic cancers. Oral oncolytics have drastically improved treatment for conditions like Chronic Myeloid Leukemia and advanced stage Melanoma and have come with the convenience of oral dosing. However, the use of oral oncolytics has created a new challenge, the increased potential of medication non-adherence. With the many considerations that exist in the treatment of systemic cancers and the increasing use of oral oncolytics in their treatment, it is important to both study the effects of non-adherence and develop interventions that aim to improve adherence. To date, limited research has been conducted evaluating the impact adherence may have on disease progression and treatment failure. Through the use of pharmacy claims data, levels of adherence to oral oncolytics can be calculated. By mining electronic health records for treatment outcomes, it would be possible to evaluate the clinical impact of non-adherence. Assuming adherence can be directly linked to poor clinical outcomes, the inherent question becomes how to best approach improving non-adherence. Non-adherence has been shown to be the result of a wide variety of factors that vary substantially from one patient to the next. The complex nature of medication adherence requires an intervention that addressed the diversity of these challenges on individual medication use. One approach that has begun to be explored is the use of artificial intelligence to develop personalized patient messaging. Through techniques known as Reinforcement Learning, artificial intelligence can be implored to develop tailored text messages that are personalized to individual causes of non-adherence. Such programming requires the development of multiple algorithms that allow patients to provide short responses to early messages to first establish the likely reason for non-adherence. By using AI to understand the cause of non-adherence and to tailor messaging to patients, the intervention will be more likely to impact the root cause for each individual patient. Additionally, AI will allow for longitudinal adaptation of messaging as patient needs/concerns may change over time.