יום שבת, 28 במאי 2011

Mobile psychotherapy study + my wishes for biometrics sensors, behavioral trackers and AI analytics...

I just read this article by Margaret E Morris from Intel's Digital HEalth Group. It describes a study of 5 cases, of patients using mobile phones that assess their moods, offer interventions such as relaxation techniques and questions that help change negative interpretations. With this, once a week the patients/users had an interview in which a therapist reviewed their data with them and suggested routine solutions to recurrent problems. The studied cases showed improvement.

It seems the researchers got a few things very right:
1) Online assessment - If you want to change a person's behavior you need to map it. How to map it is a question of technology. Brick-and-mortar clinics require the patient to report what he remembers of his behaviors and cognitions in retrospect, and such reports, from outside an intense emotional experience, may suffer from distortion of perspective and rationalization.

2) Online Therapy - traditional psychotherapy requires a leap from the sofa to real life. A patient given a relaxation technique must be able to work it when it is relevant in his routine life. The traditional therapeutic idea of here and now, is what we currently refer to as online. Interventions are given when they are needed.

3) Data mining methodology -  As scholastic bible interpretations take in little data and make whole doctrines by it, so do therapists often take single incidents and observations and induce into principles. This situation was forced by time limitations, a patient's unwilingness or inability to tell it like it is, or just the barriers of language. However, in the age of information technology, data is superfluous and analysis usually means a data reduction funnel - making simple conclusions based on a lot of complex data. The described analysis that was described in the research could be viewed as data mining, and could be broadened to take in much more sets of data on the patient.

4) Mapping cause and affect - What the weekly analytic session provided the patients, was a link between events and their emotional effect. These are the two basic components of CBT. If a person could become aware of how events and situations prompt behavioral and cognitive responses, and more importantly that these are contingent with other responses, then a conscience change would be facilitated. In the article, the example was that a man who used to fight with his family when he came back from work noticed his energy dropped right after work, so he was advised by a therapist to take a pause for relaxation before getting out of the car.

I wish these points would be developed in the future:
1) More tracking and sensor based data - The device described in the study asked the patient/user to assess his own state and reply on diagnotic questions. this takes time and means that the device has to ask at certain time intervals or the user needs to turn to the device for assessment. I wish more tracking and sensor based data was available. If the patient's heart rate variability would indicate his arousal state, or if a phone tracker would would indicate that at a specific moment was talking on the phone with his wife, or even that in the conversation the speakers were interrupting each other's speech or that the trigger words "get milk" were mentioned... well, then there would be a lot more data to use and to relate to. But more importantly, a lot of the data would reflect aspects of the event that are not part of the patient's perception of the event, and may help bring a broader set of data into the patient's attention and help the patient "see it coming" and be prepared for an aversive situation.

2) Artificial Intelligence Analytics - when more data would be available, it would be harder for a human analyst to contain it. Artificial neuron networks can handle such capacities of data and make statisticly predictivve generalizations. What is more important, and will probably be the Nobel worthy discovery in all this post, is that they will rely less on pre concieved hypotheses and more on empirical findings. They may come to conclusions that humans, as sensitive as they maybe, are prone to overlook. Of course machine learning is also prone here or there. 

3) Sharing - Instead of having the therapist providing ideas and solutions, an experience sharing network could be created. People can learn from each other how to solve problems. For example: "67% percent of people of people who are asked to buy milk at lunch feel stressed during the rest of the day, 90% of them felt better by setting a reminder on their mobile, 10% of them got to feel better by having a long open talk with their life partner about personal space." Well, wouldn't you want to know that?

Illustration:



Ref:
Morris ME, Kathawala Q, Leen TK, Gorenstein EE, Guilak F, Labhard M, Deleeuw W
Mobile Therapy: Case Study Evaluations of a Cell Phone Application for Emotional Self-Awareness
J Med Internet Res 2010;12(2):e10
URL: http://www.jmir.org/2010/2/e10/
doi: 10.2196/jmir.1371
PMID: 20439251