Using wearable fitness tech to develop insurance wellness programs has been the subject of an industry-wide discussion for years. And as the technology improves, the conversation grows. At RGA, we wanted to gather meaningful data and gain experience over theory in our understanding of insurance wellness initiatives.
In 2016, RGA conducted an anonymous study among its employees and their friends and family to explore wearable fitness trackers’ potential application for insurance product development. The study included around 1,000 participants from 23 countries and was conducted over 12 weeks using five tracking devices.
Key insights for insurers included these five takeaways:
1. Accuracy Remains an Issue
Not all devices are created equal. For improved plan performance, insurers should consider mandating high-quality trackers, retrieving data from multiple devices, or limiting member benefits based on the quality of the device, particularly for those plans in which higher activity levels trigger additional benefits.
The study used multiple devices and confirmed our assumption there would be differences among them in consistency of measurements. For example, between the study’s two main devices – both wrist-based – steps recorded on one were, on average, around 8% higher than those recorded by the other. To be fair to their customers, insurers will need to reflect these differences in the activity targets they set.
Although we tested only wrist-based devices, it is notable that one participant recorded approximately 19,000 steps on his smartphone while simultaneously recording approximately 13,000 steps on a wrist-based device. This is a good illustration that the wrist may not be the best place to collect accurate, reliable biometric measurements. In order to obtain an accurate picture of activity, insurers will need to develop multi-device, multi-location solutions.
Our study provided an opportunity to begin investigating how to identify fraudulent activity, such as manufacturing steps by “wiggling your arm.” We found activity to be remarkably consistent in terms of level (steps per minute) and routine (day of week and time of day), and we observed fairly consistent bands of activity levels linked, very broadly, to environment (home, office, outdoors, etc.). Manufactured steps were identifiable as they occurred at activity levels not normally seen and at times of day when activity did not normally occur. This suggests fraudulent individuals will need to be quite sophisticated if they want to hide their “cheating.” However, a rather surprising result was that manufactured steps sometimes raised heart rate to a level above that which would have been expected, so was this cheating at all? Significant further work is required to generate an accurate and robust solution for identifying fraudulent activity, but the initial signs are promising.
Questions the industry should be asking:
- How can insurers easily evaluate the accuracy and reliability of wearable devices?
- How should insurers translate these differences to ensure fairness among customers?
- How do insurers build a multi-device, multi-location solution?
- How can insurers identify and limit fraudulent activity?