Evidence-based decision making is a $3.7 billion market with the potential to grow to more than $10 billion by 2017. For healthcare providers, payers and pharmacy groups data analytics offers a way to improve adherence among their customers. Meanwhile, one company has developed an algorithm it says not only can anticipate which patients will have medication adherence issues, but also identify the best intervention strategy for each of those customers.
Clifford Jones is the CEO of Allazo Health, a startup that has developed a system that he says can reduce healthcare costs and improve health plan ratings. Jones traces his enthusiasm for using predictive analytics to reduce healthcare costs back to CVS Caremark where he worked on a team with Douglas Ghertner (now the CEO of Change Healthcare) and developed Pharmacy Advisor, an analytics tool to improve adherence for people with chronic conditions, particularly diabetes.
It was a member of accelerator Blueprint Health’s class last year and works out of its shared office space in New York.
“We did a lot of research of which interventions work and which ones don’t. One of the triggers would be when a patient was late to fill their prescription. What program would drive diabetes behavior change without breaking the budget?”
Jones said instead of relying on trigger-based interventions, his company’s tool predicts how to influence patients. It began by building algorithms that worked and proving that out with a patient population. The two year old company began with cardiovascular conditions and diabetes and initially focused on health plans. This year it began targeting ACOs.
Medication is just one component of adherence but that’s what makes picking the right intervention program at the right time even more critical. There are a lot of different types of interventions: face-to-face, phone, e-mail, motivational counseling, incentive programs, engaging physicians and family members, social support, apps and devices that help people manage their medication, even gamification.
Interventions needs to deliver more value. There needs to be better level of coordination to better drive behavior change, said Jones. “The most advanced solutions are not where they could be. The good thing about what we’re doing is it’s actionable now.”
One defining moment for the company was getting the data that showed its algorithm was working really well. Jones points out that traditional predictive analytics techniques aren’t well suited to predicting patients’ medication adherence and how interventions will impact their adherence. “We knew that we would have to develop novel analytic techniques to solve the problem, and this proved to be challenging, but we were successful.”
One area where Jones said its analytics tool could benefit health plans is with Medicare Advantage star ratings as well as their HEDIS Score, or Health Care Effectiveness Data and Information Set. It’s used to measure many different health issues and it also gives employers a way to compare the effectiveness of different health plans. But with passage of the Affordable Care Act, the score is one of the quality measure that determines if these plans get a bonus.
“A lot of entrepreneurship is timing and we’ve been blessed with really good timing,” said Jones. “We’ve been able to develop technology in advance of where everyone wants it and everyone wants it now.”
Congratulations Clifford on this great initiative and to your team at Allazo Health. Data is crucial for analytical modelling and accurate data can predict many behavioral patterns which will lead to various intervention strategies. Because of the multiple stakeholders in the care delivery and prescription cycle involves 1) physician prescribing medication, 2) consumer receiving from pharmacy, 3) Consumer taking the medication on time 4) Consumer refilling or performing the renewals and starting the step 1 or 2 all over. In order to calculate the effective predictive analysis, the detailed data will be required. Missing any data in that order can introduce significant amount of error in predictive analysis. Will be curious to know more about this.
Predictive modeling is really tough when one has absolute, clean data; some say impossible. I'd really like to see effect-cause-effect with this, and I seriously doubt it passes any reasonable standard of scientific rigors.
RT “@medcitynews: Health IT startup can predict adherence and forecast best intervention strategy http://t.co/hQGLkqk0”