How do people use data to make decisions? In a new study with Dr. Laurel Messer and colleagues, we explored the use of continuous glucose monitors (CGM) by adolescents with type 1 diabetes. The study is based on the idea of situational awareness as a predictor of successful self-management.
Some of the obvious ways in which people might use data are as an immediate check on their current blood sugar, or as a tool to watch patterns of change in their personal blood sugar data over the course of a day. Perhaps contrary to what one might expect, neither the length of time since a person's last glucose check or the total number of times that they checked their blood sugar during the day predicted whether they were able to successfully manage their diabetes. (Successful control was based on a widely-accepted "time in range" or TIR measure of blood sugar over the course of the day). In fact, the more often someone looked at their CGM device, the worse their diabetes control was on that day! Just paying more attention to data, then, is not a clear recipe for success.
Similarly, there was an inverse correlation between the number of alert messages someone received and their success controlling their diabetes (that one might have been expected, though, just because someone with worse control would tend to get more alerts). As has been seen in studies of nurses' behavior, receiving more messages might actually lead to "alert fatigue" that interferes with behavior change instead of supporting it. On the other hand, when adolescents actually responded to an alert message from their pump, their TIR did modestly improve. An even stronger predictive relationship was seen when adolescents used their pump proactively, entering a carbohydrate value for meals that they ate. Using the CGM device in that way was the strongest single predictor of successful diabetes self-management in this study. Adolescents' thoughtful interaction with the pump, rather than just viewing a blood sugar number or passively receiving a message, seems to have been key to their success. This is of course difficult for technology designers to automate, because the interaction needs to be started by the end-user rather than by the device.
One other finding speaks to the cognitive aspects of situational awareness, and supports my earlier argument that situational awareness is less about Narrative-level logic and attention, and more about Intuitive-level awareness and expertise. We asked adolescents to predict their current or future blood sugar value, without looking at the value on the CGM device. The accuracy of those predictions was a significant predictor of their time in range. In particular, adolescents' blood glucose control was worse when they underestimated their current values (meaning that they thought they were doing better than they were). Adolescents had better time in range values when they didn't make that type of mistake (over-estimating their current blood sugar was less harmful to self-management). Because the adolescents were asked to make these predictions before checking their actual blood sugar result, accuracy means something more than just looking at the available data. It suggests that some adolescents had an intuitive-level knowledge of their current blood sugar that was separate from using the data available from their CGM device.
What does this study tell us about using data to make decisions? My take-away is that new data sources like CGM sensor devices can be a useful training tool, but that our eventual goal in interacting with these devices should be to develop our own expertise. Simply having data available or providing alerts didn't help in this study, and indeed it was potentially harmful due to alert fatigue. But proactively interacting with the data source to gain knowledge was helpful. And ultimately, people who had an accurate sense of their own body state separate from the data source were among the most successful in managing their diabetes, a finding consistent with the view that situational awareness is largely an Intuitive-level process that happens outside of conscious awareness. Sensor devices can provide us with helpful data, but we may be most successful over the long term if we use those data in a deliberate way, with the idea that eventually we won't need the device anymore.
Comments
Post a Comment