Patient engagement platforms are becoming more and more popular with healthcare providers and even pharmaceutical companies. The concept of patient engagement addresses unmet needs around a patient centric approach: empowering patients and family members to better manage their health and care. Sounds great, doesn’t it. So how come so many patient engagement platforms are still far from delivering on the promised value and ROI?

The answer lies in understanding the expected values and identifying the execution hurdles. Expected values are as diversified as any promising new technology gets: doctor-patient relationship management, integrating relevant info across sources, family sharing, adherence monitoring, incentive management, just to name a few. Harnessing of Big Data is probably the most common theme across expected values.

Dig data which is often referred to as the 3Vs – Volume, Velocity and Variety – is the foundation of all expected values. It includes the life time doctor-patient communications in real time and across channels; the crowd sourced data from online communities; the endless sensor data from wearable technology; the multi dimensional ways to incentivize families; and more. While collecting and storing big data has been a significant IT challenge, the real hurdle lies in the ability to extract insights from Big Data. 

Machine learning algorithms help us do that. Unlike an expert system that requires advance judgment rules, machine learning algorithms let the data ‘teach’ us what the insights are. They can help us learn what data is relevant for each user (patient, family or clinician), how to avoid alert fatigue, what drives adherence, etc. Machine Learning algorithms help us make actionable sense out of Big Data.

But for such algorithms to work, we need to ‘feed’ them with data first. Some reference set of data often called training set. The inherent nature of these algorithms renders them useless without a critical mass of data to learn from. In today’s pace users have limited patience to wait for a technology to start adding value.  Cold start is the problem of not having enough data for machine learning algorithms to add value.

One way to overcome cold start is to focus. Imagine how much data you need in order to teach an algorithm to identify all animals in the safari, vs. the amount of data needed to only identify the giraffe… similarly patient engagement platforms that try to provide personalized data to all patients across variety of complex conditions will require a lot to overcome the cold start problem. Whereas a focused patient engagement platform stands a better chance to create value and become a trusted go-to source.

It comes as no surprise that specialty pharmaceutical companies (often focused on Orphan drugs) are ahead of the patient engagement curve, as described by Mark Lush et al. in a Deloitte article. Health care professionals or disciplines are no different. The diversity and multi dimensional aspect of the adoption challenge can be overwhelming. That’s why, focusing on specific clinician disciplines can help identify adoption hurdles and provide the agreeable level of telehealth services.

Service providers and payers are collaborating on risk sharing and result based models in population health management. These models allow both providers and payers to get better care and financial outcomes, but they also expose both to risk. A key success factor is to have strong risk prediction models. Innovative work in actuarial fields today includes use of machine learning algorithms in risk modeling. Leading to better risk prediction of specific population health management.

To summarize, the best preventive medicine for patient engagement platforms against catching a cold start is to focus. Focus, not only leads to expertise but it also enables machine learning algorithms to extract insights from big data in order to generate value in relevant time.