Facebook Twitter LinkedIn Instagram Email Printer Google Plus
Engage Consumers May 09, 2016

Digital Health: An Employer Framework for Evaluating What Works

Chief Technology Officer, SimpleTherapy
Key Takeaway
Evaluate #digitalhealth on prediction, proof, personalization, participation, practicality - @simpletherapy CTO

Arpit Khemka is the Chief Technology Officer of Fremont, CA-based SimpleTherapy, a fully online alternative to traditional physical therapy. SimpleTherapy’s personalized pain recovery system is designed to address pain from 80 percent of all types of musculoskeletal pain and physical therapy prescriptions, head-to-toe, using its adaptive, online sequence of 15 minute video-guided sessions that are personalized for each user. Two of the largest five nationwide health insurance companies partner with SimpleTherapy to offer members online exercise therapy and the company has expanded to work directly with large employers from the piloting to the full-scale commercialization process, according to CEO Helena Plater-Zyberk. Based on these experiences, Khemka presents here the “Five Ps Framework” for evaluating a digital health solution:

An unprecedented number of digital health applications, tools, and platforms are transforming the way employees can engage in managing their own health. According to late 2015 analysis, "there are roughly 165,000 apps in play from some 40,000 vendors, half of which have entered the market in just the past two years."

One of the biggest mistakes employers can make is selecting an innovative app or platform without considering what end users are willing to sustainably integrate into their daily lives. Even the coolest technology will be ineffective if employees aren’t willing to use it.

When selecting which of these solutions to introduce for employee populations, health benefits leaders can evaluate the myriad options along five key criteria: Prediction, Proof Points, Personalization, Participation, and Practicality, or "The Five Ps Framework":

1) Prediction

Predictive modeling is applied in many consumer settings. It’s how Netflix selects movies to feature, how Facebook targets advertising, and how Google determines the correction needed to fix the search bar typo. Mining insights from historical data sets enables more effective and less costly choices. Take Jane, a 52-year-old overweight female who has had knee surgery within the last five years. Analyzing a dataset on individuals with similar profiles shows that Jane has a 75 percent chance of suffering another costly musculoskeletal episode within the next six months. To avoid this needless pain and expense, Jane can be targeted to engage in exercise therapy programs, specifically programs tailored to prevent the musculoskeletal injuries that are most common to her job-specific role. Prediction used for prevention can improve the health of individuals while reducing cost. Accurate predictions rely on having data in sufficient volumes to make the models work. According to the American Medical Informatics Association, the volume of healthcare data is expected to increase to 25 million terabytes by 2020, an almost inconceivable amount through which to comb and learn. For perspective, one terabytes is what it takes to store 2 million photos.  

Ask: Has the company already amassed the necessary data to conduct meaningful predictive analytics?

 

2) Proof Points

Before piloting to an active employee population, the startup should be able to demonstrate they can generate meaningful outcomes by sharing the results of an independently reviewed clinical trial. In an Institutional Review Board (IRB)-approved clinical trial, participants receive specific interventions according to the research plan or protocol created by the investigators. These interventions may be drugs, devices, procedures, or behavioral changes. For a digital health company, the tested intervention will be a platform, a tool, or software such as an algorithm built to help consumers engage in managing their own health. A clinical trial may compare this new medical approach to a standard one that is already available, to a placebo that contains no active ingredients, or to no intervention. The investigators determine the safety and efficacy of the intervention by measuring certain outcomes in the participants.  

Ask: Does the company have data to show from a properly conducted clinical trial to serve as a building block for predictive analytics capabilities and to build confidence in the ability of the product to scale?

 

3) Personalization

Predictive analysis applied to individuals is one of the hottest concepts to come along in the last decade. Personalization is made possible through a complex system of customer loyalty data, metadata, and cloud computing that enables the continuous collection, storage, combination, and analysis of “big data” about each of us from a number of disparate sources. For example, when there are thousands of data points on 45-year-old men, all with body mass indexes (BMI) of 37 and similar self-reported pain scores, treatments for knee pain can be tailored to these individuals. Certain exercises may not be well-tolerated by these individuals, and therefore removed from an exercise therapy regimen. Currently, new drugs that put into market are closely followed for side effects.

Ask: Does the company have the capability to pinpoint certain side effects and adverse reactions based on past drug and other data?

 

4) Participation

A good digital health product requires participation, particularly if it wishes to successfully utilize user data to drive its predictive analytics and personalization. Netflix and OkCupid, for example, improve their algorithms when users rate movies and potential partners. Google and Facebook monitor browsing habits. These systems are designed to be so easy to participate in that users are almost completely unaware of how much data they are providing at any given moment. The importance of making a product more engaging cannot be understated; however, tracking an individual’s progress and challenging them relative to their behavior and engagement patterns across your system is equally important. Analyzing these patterns, such as when and how an individual responds to incentives coming from a predictive model, allows for personalized interventions. These results then become sustainable as the interventions generated by the predictive model continuously adapt to individuals, rolling up to significant population change. Prediction paired with personalization generates more participation.

Ask: How is the product designed to engage and drive willing and informed user participation?

 

5) Practicality

The best way to go about collecting user data and increasing participation is to make the product or service so practical and easy to use that it becomes effortless. In the case of digital health, this can mean eliminating the need to purchase expensive hardware or cumbersome non-integrative software, such as an Xbox, that the user doesn’t already own and use. Creating a product that is convenient, practical, and can run on existing platforms or devices minimizes the risk of losing relevancy.

Ask: Are special devices or tools needed to use the product?

Insights in your inbox

Subscribe