masthead-services-03-1600x650.jpg

Blog, Comment & Insight

For Pharmaceutical Marketing Analytics

Using Artificial Intelligence to Improve Patient Adherence Results and Patient Outcomes

Posted by Dr. Andrée Bates | Mar 5, 2018 10:33:09 AM

Pharmaceutical companies spend a significant percentage of their marketing budgets to capture initial market share. Since the majority of spend goes towards attracting new customer, efforts at retaining existing customers are given short shrift, with a paltry percent of budgets focused on these. By focusing mostly on new customers, companies are missing a valuable opportunity to increase sales and market share.

Impact on Patients

Overall, non-adherence is responsible for an estimated 125,000 deaths a year in the United States, 33 to 69 percent of medication-related hospital admissions, and nearly 23 percent of all nursing home admissions. It is also a significant factor in drug-resistant infections such as HIV and antibiotic-resistant infections, posing both an individual and a public health risk.

Numerous studies find that patients who follow treatment recommendations have better health outcomes than those who don’t.  One, Horwitz et al., found patients who took 75 percent or less of the recommended dosage of beta blockers following a myocardial infarction were two to six times more likely than more adherent patients to die within a year of follow up.

Impact on Physicians

Non-adherence affects a physician’s ability to appropriately treat patients. For instance, if a patient is non-adherent or only partially adherent to prescribed medication and symptoms continue, physicians may assume a drug is not working and switch the patient to another medication. This can adversely affect the efficacy of a particular treatment regimen and may hurt the patient’s chances at improving.

Physicians feel the negative effects of non-adherence in other ways. With high compliance comes less disease-related medical costs for patients, fewer complications and less need for extensive visits or emergency hospital visits. However, with low compliance comes increased complications and hospitalization. In a study of diabetes patients, those who were 80-100 percent adherent had a low risk of hospitalization (13 percent). When adherence dropped to 60-79 percent, risk increased to 20 percent.  A drop in adherence down to 40-59 percent saw another jump in hospitalization risk, to 24 percent. Patients’ who are non-adherent have more problems, problems physicians may not be able to accurately diagnose, understand and treat.

Impact on Pharma

In todays’ cost constrained world, pharmaceutical companies can no longer ignore the hidden value available by increasing patient adherence. Today, an estimated 70 percent of patients who begin a Pharmaceutical therapy discontinue it within 1 year, even those with chronic conditions that require ongoing treatment or those taking chemotherapy to prevent cancer recurrence.  This costs the global Pharmaceutical Industry an estimated $30 Billion a year.

Put another way, increased adherence for a product with approximately $1 Billion in sales would translate to an additional $30 to $40 Million in annual revenue. In addition, since it costs six times more to attract new patients than to retain current patients, increasing the focus on, and yields from, adherence means additional money saved and earned.

The Solution: Using Artificial Intelligence to Predict and Enhance Outcomes

Companies implement numerous strategies to increase adherence and persistence with their products that have varying success as lack of adherence has many causes (over 250 have been documented in one study).

However, now with Artificial Intelligence techniques we have even more powerful approaches that allow stronger predictive insights that allow companies to predict which patients will cease therapies and what specific interventions will keep them on their treatment. 

Getting data on our customers allows us many types of insights. One project that we have done several times for various pharma clients is looking at patient data and identifying which individual patients will stop adhering to treatment in advance of them doing so, and then identifying what it would take to prevent them stopping treatment where this is medically indicated. Of course, if it is due to severe side effects or a medical reason for ceasing that treatment, then that is sensible and would not be indicated for intervention.

The projects in which we have done these types of patient adherence predictions and interventions tend to be injectables. I am not sure why but these are the client teams who have come to us with this challenge. The great thing about the clients we have had in this space to date is that they have a wealth of patient data because of the support they are providing their patients due to it being an injectable. We have patient notes from the visiting nurses who help them with their injections, we have delivery information about the drugs and when they are delivered to the patients, we have call center interactions from the patients and their point of contact. We even have sensor data from the needles themselves. And a lot more. Then we apply a combination of Natural Language Processing Algorithms and Machine Learning algorithms to the combined data in order to identify who will stop adherence, why, and what interventions would change the result and keep them adhering to their medications. This is fed automatically into a system for the nurses seeing the patients so each day they could see which patients were in danger and what interventions would be most effective, as well as to the patient liaison in the call centers. The brand team also get summaries of how many are currently predicted to cease treatment, and heat maps of where they are, and results of interventions in terms of staying on therapy.

The projects we have done in this space have been very successful when comparing the percentage of patients who ceased treatment before the algorithms to the percentage doing so after the systems have been introduced and therefore have improved patient outcomes and improved our client’s revenue.

Conclusion

Big data and AI allows the solving of numerous challenges within pharma but this particular one is a very important one when one considers the outcomes for patients who cease taking treatment when they should not. For companies wishing to consider this approach, please contact Eularis (http://www.eularis.com)  to discuss your situation confidentially.

Topics: Marketing Insights, Big Data, Artificial Intelligence, Advanced Analytics

Written by Dr. Andrée Bates