Tracking Mental Well Being by Using Sensing Technology

Serious Mental Illnesses also known as SMI’s, are amongst the most pressing healthcare concerns as people suffering from these illnesses die, on average, 25 years earlier than others. On top of this, these illnesses are, in the US, in the top 5 conditions for direct medical expenditure with annual costs that exceed $30 billion and require lifelong management and monitoring which is hard to maintain. Early warning signs of change in mental well being or relapse are often looked over and missed, and there have been studies showing how behaviour-monitoring devices can assist in noticing these declines in mental health early as well as self tracking. As the capacity to automatically track self-reported behaviours is expanding, there is a possibility to dramatically improve illness management in a cost effective manner. Moreover this method is unobtrusive meaning there can be discrepancies between sensing capacity and patient acceptance.

Sensing can be used as a personalised treatment which ill explain further down in this essay. Despite all this, clinicians still haven’t accepted technological support as much as they could benefit from it. Low acceptance can also lead to low adherence and a rise in stigma which is what peopler trying to avoid. If developed closely and properly with patients and their clinicians, sensing technologies would be accepted more widely and provide a more cost efficient and effective method to manage SMI’s. Mental health has been known to be linked to behaviours, decrease in physical activity or social interactions and emotional responsiveness are known symptoms. There are many more behavioural changes that can be tracked, that I’ll go over in more detail later. Sensing physical activities like decrease or increase in physical movements, depending on which SMI we’re talking about, can be tracked using smartphone applications like the data of the accelerometer, as well as the compass. They show us in much more detail when someone’s walking and if they’re climbing stairs and its direction. This can be extra helpful when talking about eating disorders as patients often do more exercise and climb up and down stairs repeatedly for a long amount of time to burn the calories off. Sensing social engagement might be the hardest one to track as capturing and analysing social encounters can be hard to do without trespassing the users privacy. Nonetheless there is a growing body of evidence showing that changes in vocal and speech patterns are signs in decline of emotional health.

Privacy-sensitive audio can be used to show us the duration and change in pitch and much more of users conversations. All of this can be achieved using the microphones on smartphones, but we can also identify daily stress using this method. Usually, physiological symptoms of stress are measured through sensors, like chemical analysis, but these method use direct interaction and are thus less likely to be used by patients. New techniques have been developed to passively detect stress episodes using smartphone microphones and are made to adapt to each individual and can be used even when the microphone’s position relative to the speaker and the room is nonstatic. To track sleeping patterns, patients usually have to wear specialised sleep sensors which most likely wouldn’t work with them, as the burden placed on users while sleeping must be minimised so that there aren’t any other factors influencing their health. Smartphone features like light sensor, wether the phone is in use etc, can be used to estimate sleep duration with a regression model. Even though this has a large error margin compared to other methods like instrumenting the bed, this is less obtrusive and so has a more sustained use. Moreover, depending on the users will, the system can offer a more detailed record of their sleep duration. The authors have unpublished experimental data from college students that show interaction patterns alone can be used to estimate sleep duration with 85% accuracy, as smartphone use is increasing among youth. On top of this, it’s also important to go above and beyond detecting the illnesses, and give patients feedback. When going to clinicians to get help for SMI’s, the patients only get their feedback at the next appointment which could be a weeks away while using behavioural sensing gives immediate feedback, although there is no guarantee, the feedback must consist of easy to understand visualisations so users can understand their illnesses. The writers of this article have come up with an app called BeWell, it measures the physical activities, social engagement, and sleep patterns in an ambient display so users don’t feel different from others. Most apps give general advice that doesn’t fit each individual and might lead them back into bad habits as it fails to relate to the user.

The best way to fix this would be to use context fit algorithms. This wouldn’t only make individuals more engaged in getting help but also help them understand and learn their illnesses. The issue with this is that the creation of these messages would need the help of medical professionals which would cost even more to the overtaxed healthcare system. Personalised interventions could be more meaningful and potentially more convincing lessons for the patients, though there already exists psycho-educationalinterventions they usually are uniform approaches and so don’t stick to the individuals. Pharmacological treatments and psychosocial treatments, for an SMI, are grounded in its pathology and are centered on an evidence-based understanding of the illness; so it would also be beneficial to consider each SMI’s neural and behavioural characteristics. To be effective, sensing solutions must consider factors that affect patient adherence. Several factors have been associated with nonadherence, mainly the stigma that SMI’s carry. This is what a lot of patients refuse to use devices that will make them feel like they stand out by having higher levels of privacy using smartphones, it could be possible to reduce the potential stigma as we would ensure that the user-facing aspects of the system visible to others would be less visible. An important aspect of rich sensing method is gaining the users confidence which can be influenced by the technological reliability as well as its accuracy. To gain this trust it’s crucial to tell the users about the data’s limitations and the systems uncertainty.

The accuracy can be improved by making users fill out reports for a designated training period and report anomalous cases, patients with SMI’s are prone to have memory and cognitive deficits which can mean that clinicians won’t know whether the sensing data is misrepresentative or if the patient’s experiencing a period of reduced self awareness. Another major factor in this system is respecting the users privacy. The sensing algorithms should only store as much raw sensed data as the user is comfortable sharing and as is clinically meaningful, this is why the authors have come up with a system that does not record raw audio but instead destructively processes audio data on the fly to extract and store features that are useful to infer the presence and style of speech, but not sufficient enough to reconstruct the spoken words. It is important that the user has control over how their data is being viewed and used, so they play an active role in their treatment and understand illness, which will help them adhere more to treatments. This would include deleting data and muting sensing at certain times. This will also help combat the stigma as they show that they are still in control of themselves. Moodrythm is an app whose goal is to find that balance between sensing and acceptance mostly used for bipolar disorder. This illness is associated with poor functional and clinical results, high suicide rates and large costs societal costs.

The app has a specific goal as it is only for one SMI, which is to give users a routine as that is one of the main challenges for people suffering from bipolar disorder and a lack of routine can lead them to have new depressive and manic episodes. The app focuses on helping them maintain this routine, and uses the social rhythm metric, a validated five-item self-assessment of the regularity of daily routines and moods. With this daily tracking it is easy to see how a regular routine can positively affect patients moods, which can help clinicians stabilise the routines even more, however, self-assessment and self-reporting by the patients can’t always be reliable as they often misjudge themselves during manic episodes. Smartphone sensing capabilities monitor how often the user has social interactions and when they’re awake or asleep, it uses SRM(social rhythm metric) to make a model for a set of self-reported tasks with therapeutic goals. They use a reward sensitive neural characteristic associated with bipolar disorder and virtual badges to reward users that complete the given tasks, this has been proven to help motivate users and help them adhere to treatments. This app computes during the day the frequency and duration of the interactions of the users based on the analysis of the audio data, on which it bases the completion times for the tasks and the goals for social engagements for the user.

For the sleep pattern, they have a sleep module that uses the applications from their smartphones like the accelerometer to estimate how long a user sleeps with approximately 85% to 90% accuracy. SRM uses an algorithm where the higher your score is, the more stable your routine is. Their homepage includes floating bubbles which represents daily events, these bubbles can move, and the more stable the routine, the less the bubbles move, the bubbles change colours, from green, if it’s stable, to red if it’s not. Therapists help patients get an insight on their disorder and how they can improve their daily routine so that their mental health improves. While developing this app, they worked with three clinicians and three patients, where they closely studied the link between patients acceptance and their understanding of the data to see if it could help them or their clinicians see from another perspective. Even though the link between technology and its utility in clinics has been proven, it’s still not currently used as a part of it. Though this does have challenges, like the fact that clinicians are used to only having limited data on their patients, and not their daily data, so it would be hard to go over all of this data for each patient.

This could be solved by destructively processing the location data and only telling the clinicians view changes in location and how many times they have been there which makes it a bit easier to process. In conclusion, the authors are making it very clear that there are more benefits than drawbacks in using rich sensing technology and smartphones during and after the recovery process of the patients, as it is more cost-efficient and time-effective for both the patients and the clinicians. It helps the patient as they get a better understanding of their illness and help them get a daily routine. I find that the authors are right for the most part as it seems to be very beneficial also for the people surrounding the patients. It also seems to be very cost efficient compared to the other methods so I do think it would be useful to incorporate technology more into clinical practises. Though if this gets into the hands of the wrong people they could use this to manipulate others. One of the issues that would be solved is the amount of money spent on the recovery of these illnesses. An ethical issue on this subject is the privacy, even though they have found ways of improving their privacy settings, this could still be an issue especially because it would keep track of everything they do everyday. Fields involved in this matter are cybersecurity, to keep the data of the patients safe as cyberattacks are increasing, software engineering to produce the applications, and psychology and medicine to cure the patients.