Pilot study tests wearable sensors and smartphone-based data collection for diabetes

mHealth Research Digest

Background

In recent years, wearable biophysical sensors have rapidly increased in functionality and number while decreasing in size. These sensors monitor an array of biometrics including blood glucose, heart rate, respiratory rate blood pressure, motion, and acceleration.

Contemporaneous development of diminutive location sensing devices, including Global Positioning System (GPS) receivers allow for study of geographic influences on human health to a unforeseen level of sophistication. Reams of data collected by such monitoring devices can be transmitted wirelessly to and processed by smartphones, which continue to advance in computing power and portability.

GPS allows for tracking of ones space-time path; this data can elucidate periods of movement or exercise and stationary activities. A number of studies have used GPS to provide these detailed, objective accounts of human activity.

However, few have combined GPS tracking with real-time monitoring of physiologic metrics that examine the connection between human geography, health behaviors, and disease. Doherty et al. recently published an account of a pilot study that utilized a continuous multi-sensor monitoring system of metrics of human health (heart rate, glucose) and daily activities in type 2 diabetics.

Design

The monitoring system included a Blackberry smartphone with custom software, GPS receiver and ECG/3-axis accelerometer connected wirelessly to the smartphone through Bluetooth and implanted continuous glucose monitor that stored data over the study period rather than transmitting in real-time. The Blackberry platform was chosen because it allows for data tracking custom software to run continuously in the background.

Subjects additionally kept a food and medicine diary and completed web-based prompted activity recall diary that used the collected GPS/accelerometer data. A graphical user interface that presented the multiple data streams in real-time was developed for decision support, pattern elucidation, and caregiver engagement. 40 type 2 diabetic patients were monitored with the aforementioned system for a period of 72 hours. During this period, a research assistant was monitoring all incoming data and keeping track of any hardware or software issues that arose.

Results

Of the 40 enrolled patients, three did not complete the 72-hour study period, with all three citing irritation from the subcutaneously implanted continuous glucose monitor. The foremost challenge with the smartphone platform was battery life, even with the addition of an extended battery and external GPS receiver. The volume of incoming data provided further stress on the battery; the ECG alone transmitted up to 25 data points per second.

A challenge to GPS tracking noted by the authors was blockage of signal transmission resulting in unknown location. Most of these lost data points, 86.3%, were due to indoor activities and few drops occurred during travel periods. While mapping GPS data revealed a variety of salient information regarding human activity, creating algorithms for automated analysis proved challenging.

However, the algorithm produced by the authors predicted activity with 95% accuracy and took 60 seconds to process 24 hours worth of data. The output of the automated activity detection software was used to prompt users for additional details and allow participants to correct any inaccurately analyzed data.

For example, if the software determined that a patient was at home for a period of time, the user could then modify that with the type of activity being performed at that location. Accelerometers and ECG tracking allowed for detection of stationary physical activity.

The continuous glucose sensor was implanted under the skin and monitored by a device worn at the hip. Patients had to measure blood glucose using their glucometer and manually enter readings into the device for calibration four times per day. Reading were extracted every 10 seconds and recorded by the monitor at 5 minute averages. Patients additionally used a food and medicine diary over the study period.  Writings in the diary were recorded with a Bluetooth-enabled pen that wirelessly transmitted what was written for storage in the smartphone. The data was analyzed by Nutritionist Pro data analysis software. Outside of the three subjects that ended the study early, nearly 100% reporting of meals was found. Although medication use was always recorded, dosages were absent in 25% of recordings.

After completing the 72-hour study period, 29 patients were contacted and asked about their experience. 79% of patients said that they would do the study again, while 18% said “maybe” and one patient said “no.” Patients reported some difficulty with the prompted recall diary but 78% felt it accurately represented their activities even before they corrected errors and entered descriptions.

Discussion

This study offers an enticing glimpse into a possible healthcare future, in which patients are able access to their blood pressure, glucose, and other metrics in real-time. Such a reality would advance disease self-management immeasurably forward; chronic disease patients could quantify the effects of various lifestyle decisions and actions on their health immediately.

In addition, large numbers of people utilizing these devices would create exciting opportunities for analysis of lifestyle behaviors on health at a population level. As smartphones continue to advance in processing capability and battery life, and body sensors miniaturize further, streams of data from these devices will allow both physicians and patients insight into disease processes never before capable.