Accountable care organizations are incentivized to keep their patients healthy and shrink healthcare expenditures. Ensuring that patients with chronic diseases take their medications as prescribed and stay out of the hospital is a priority for ACOs. As a result, it is imperative that they implement the best possible medication adherence programs that apply predictive analytics to both anticipate which patients will likely not be adherent to their medications and forecast the most effective intervention strategy for each individual patient.
Medication adherence programs today
Most interventions are currently assigned to patients using rules-based systems. Pre-defined rules activate interventions based on specific medical events and profiles. For example, under a rules-based system, every patient older than 50 that is 1 day late to refill an RAS antagonist prescription might receive an automated call reminder.
Rules-based systems do not account for the wide variation in patient behavior that stems from the interaction of a whole host of patient characteristics and extrinsic factors. Some examples of patient characteristics include history of adherence, responses to past medication adherence interventions, medical history, relationships to healthcare professionals, age, race, income level and education level, among many other characteristics. Extrinsic factors include, but are not limited to, the cost of medication, drug side effects, the number of medications prescribed, severity of disease, co-morbidities and healthcare provider characteristics.
In the example above, not all 50-year-old patients who are one day late to refill their renin-angiotensin system antagonist medication will become adherent to their medication after an automated call reminder. Some patients will fill their medication within a day without the reminder, and other patients are going to need a much more intensive intervention. As a result, under a rules-based system, unneeded interventions are delivered to some patients or interventions that are unnecessarily intensive. Other patients receive ineffective interventions. In addition, rules-based systems cannot learn from patient behavior. If a particular automated call reminder did not work the first time on a specific patient, it is probably not going to work the second time the patient is delayed in refilling their prescription. This information is useful in predicting what other types of interventions will work on that patient and which patients the automated call reminder is likely to be effective.
Personalized medication adherence: The next frontier
Personalized medicine has traditionally referred to the process of finding the best therapy for a patient through analysis of a patient's specific biology. Since no two people share exactly the same DNA, personalized medicine alters and targets therapies to each individual patient. However, a drug must enter a patient's body to improve a patient's health, no matter how advanced and effective a pill is. Just as medicine has become personalized, we must also take an individualized approach to improve medication adherence.
Medication adherence interventions come through many communication channels and use differing approaches — all intended to convince patients to become adherent to their medication. An intervention can be a reminder, a conversation or counseling. It can be educational or motivational. Live calls, automated calls, mail or face-to-face conversations are all considered traditional intervention channels.
Technological advances in the last few years have produced smartphone apps and pill bottles that track adherence. Some employers are implementing incentive programs to entice people to take their medications as directed. Behavioral outcomes, intervention costs and return on investment from the various types and combinations of interventions differ significantly from patient to patient.
Personalized medication adherence programs initiate interventions in a patient-specific manner to successfully affect behavior change with the lowest cost interventions, maximizing return on investment. First, predictive behavioral modeling forecasts who is likely to not be adherent to their medication. Personalized medication adherence interventions are then applied to the high-risk population. The behavioral outcomes of the interventions are used to continually improve the predictive model and hone in on the best intervention strategy for each patient. Through predictive analytics, personalized medication adherence programs give the optimal intervention at the optimal time to the right patient.
ACOs can become leaders in medication adherence
It is imperative that an ACO's patient population is adherent to their medications to improve health across a population and reduce healthcare costs. ACOs can become leaders in the next generation of medication adherence programs by implementing the following plan.
1. Predict medication adherence at the individual patient level. Employ predictive analytics to estimate the probability that each individual member within the population will be adherent to their medication, identify at-risk patients and group patients into prioritized categories. This information can be used to spread resources across the entire population to maximally increase medication adherence.
2. Deliver patient-specific adherence interventions. Choose several intervention communication channels and methods that can be tailored to each individual patient to optimize return on investment. The programs may be built in-house or outsourced, but the adherence interventions must be able to be personalized.
3. Create a medication adherence program that can learn from patients. Machine learning and advanced statistics models can learn from new and existing patients to continually zero in on the best intervention method for each patient.
4. Track all data. The most effective programs will be those that track data from all sources — claims, lab results, electronic health records, health apps, health risk assessments and other new sources of patient data. Data collection is essential for medication adherence programs to continually learn what interventions will work best for different combinations of patient characteristics and extrinsic factors.
Clifford Jones is CEO of AllazoHealth, a firm that applies predictive analytics to patient data in order to better solve the problem of medication non-adherence for health insurers, PBMs, ACOs and integrated delivery networks. Before founding AllazoHealth, Clifford developed CVS Caremark’s award-winning “Pharmacy Advisor” medication adherence program, which earned the “2011 Rx Benefit Innovation Award” from the Pharmacy Benefit Management Institute and a “Best Practices in Health Care Consumer Protection and Empowerment Award” from URAC. Earlier in his career, he led the development of analytics software for Boston Consulting Group’s healthcare practice. Clifford graduated from the University of Pennsylvania’s dual-degree management and technology program where he studied management, engineering and mathematics.
Lee Cooper was a summer intern at AllazoHealth and is pursuing a joint JD/MBA degree at Columbia University. Prior to graduate school, Lee was a strategy consultant at Artisan Healthcare Consulting, focusing on commercial strategy and launch planning in biopharmaceuticals. Lee received his bachelor of arts degree from Dartmouth College.
Prior to joining AllazoHealth, Rebecca Elwork worked in roles in finance and quality improvement at NYU Langone Medical Center and for a healthcare consulting firm that advised self-insured organizations on the best strategies to decrease employees’ health care costs and improve health outcomes. She also tried her paw at launching a start-up intended to help pet owners find highly rated pet service providers and schedule appointments online. Rebecca holds a bachelor of arts degree in biological basis of behavior from the University of Pennsylvania and master's degree in health services administration from the University of Michigan.
More Articles on ACOs:
Advocate, Blue Cross ACO Held Down Costs in 2012
4 Accountable Care Models for Rural Providers
It's Not Easy to Engage Patients — Even for ACOs