One of the most effective methods of treating depression involves increasing the quantity and frequency with which the person engages in pleasurable and rewarding activities. While this may sound simple for someone who is not depressed, it can be very difficult for a depressed person to find the motivation to participate in these positive activities.
Mobilyze provides didactic information, available either over the phone or via a web browser, that teaches the patient the kinds of things they would learn from a therapist. It also provides tools that support activities, which are typically prescribed in psychotherapy, such as monitoring activities to see relationships between activities and mood, scheduling positive activities, identifying ways in which the patient avoids engaging in activities that are likely to improve mood, and developing strategies to overcome that avoidance.
This type of intervention requires some consistent activity within the phone on the part of the patient, such as inputting information on activities, rating mood, and so on. It also requires that the patient engage in activities independent of the phone. As mentioned above, even when people are not depressed, their tolerance for such activities is often low, and among people with depression, where loss of motivation is a cardinal feature of the problem, non-adherence is common.
To address this, we are working to incorporate context sensing into Mobilyze, which will allow the phone to stay in continuous contact with the patient, interpreting his or her state in real time and reacting accordingly. The context sensing system uses data from the embedded sensors within the smartphone to develop algorithms aimed at predicting the user’s location, activity, social context and mood at any given point in time. We have elected to use only the embedded sensors, as we believe that asking the patient to attach and wear an external sensor would be one more point where non-adherence would be likely and treatment efficacy could be undermined.
To develop the algorithms, the user must go through a period of “training” the phone, which involves answering brief questions about these states until the phone develops an acceptable level of accuracy. Once Mobilyze is trained to identify the patient’s states, it can identify these states without any input from the patient.
Mobilyze can infer when the patient is engaging in the positive events prescribed as part of the treatment, in which case positive feedback is provided, and when the patient is not adherent, in which case encouragement and suggestions are provided.
In addition, context sensing can potentially identify risk states (e.g. at home on a weekend for more than 4 hours alone) or in states that are beneficial (e.g. out in a public place with friends) and respond accordingly with suggestions or positive feedback.
An initial evaluation of the first version of Mobilyze with 7 users showed promising results. The patients’ depressive symptoms improved significantly over 8 weeks of use. The context sensing system performed reasonably well, obtaining good predictive models for social context, and reasonable models for location and activity.
Models to predict mood did not perform well, in part due to problems in the measurement of mood.
We are currently testing a new version of Mobilyze that improves our capacity to collect, clean and analyze sensor data, as well as improved assessment models that are more sensitive to variations in mood.
Mark Weiser, often called the father of ubiquitous computing, once wrote that;
“The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it.”
mHealth interventions are more likely to succeed if they are integrated into the patterns of people’s lives, and make minimal external demands. Mobilyze aspires to use the mobile phone to slip into the patient’s life, gently nudging behavior towards activities that are more likely to improve mood.
