Personalized Depression Treatment
For many suffering from depression, traditional therapy and medications are not effective. The individual approach to treatment could be the solution.
Cue is an intervention platform that transforms sensors that are passively gathered from smartphones into personalised micro-interventions that improve mental health. We analyzed the most effective-fit personal ML models for each subject using Shapley values to understand their feature predictors and reveal distinct features that deterministically change mood with time.
Predictors of Mood
Depression is a major cause of mental illness around the world.1 Yet, only half of those with the condition receive treatment. To improve outcomes, healthcare professionals must be able identify and treat patients who are the most likely to respond to specific treatments.
The ability to tailor depression treatments is one method of doing this. By using mobile phone sensors as well as an artificial intelligence voice assistant, and other digital tools researchers at the University of Illinois Chicago (UIC) are developing new methods to predict which patients will benefit from the treatments they receive. Two grants were awarded that total over $10 million, they will make use of these techniques to determine biological and behavioral predictors of responses to antidepressant medications as well as psychotherapy.
So far, the majority of research on factors that predict depression treatment effectiveness has focused on sociodemographic and clinical characteristics. These include demographic variables like age, sex and education, clinical characteristics such as the severity of symptoms and comorbidities and biological indicators such as neuroimaging and genetic variation.
While many of these variables can be predicted by the information in medical records, only a few studies have employed longitudinal data to determine the causes of mood among individuals. Few also take into account the fact that mood can vary significantly between individuals. It is therefore important to develop methods which permit the identification and quantification of individual differences in mood predictors and treatment effects, for instance.
The team's new approach uses daily, in-person evaluations of mood and lifestyle variables using a smartphone app called AWARE, a cognitive evaluation with the BiAffect app and electroencephalography -- an imaging technique that monitors brain activity. The team will then create algorithms to recognize patterns of behavior and emotions that are unique to each individual.
The team also developed a machine-learning algorithm that can model dynamic predictors for each person's mood for depression. The algorithm combines these individual characteristics into a distinctive "digital phenotype" for each participant.
This digital phenotype has been correlated with CAT DI scores which is a psychometrically validated symptom severity scale. However the correlation was tinny (Pearson's r = 0.08, BH-adjusted P-value of 3.55 x 10-03) and varied widely among individuals.
Predictors of symptoms
Depression is the most common cause of disability around the world1, however, it is often untreated and misdiagnosed. Depression disorders are usually not treated due to the stigma associated with them and the absence of effective treatments.
To aid in the development of a personalized treatment plan to improve treatment, identifying the factors that predict the severity of symptoms is crucial. However, the methods used to predict symptoms depend on the clinical interview which is unreliable and only detects a tiny variety of characteristics associated with depression.2
Machine learning can improve the accuracy of the diagnosis and treatment of depression by combining continuous, digital behavioral patterns gathered from sensors on smartphones with a validated mental health tracker online (the Computerized Adaptive Testing Depression Inventory CAT-DI). Digital phenotypes can be used to are able to capture a variety of distinct actions and behaviors that are difficult to capture through interviews and permit continuous and high-resolution measurements.
The study included University of California Los Angeles (UCLA) students experiencing mild to severe depressive symptoms participating in the Screening and Treatment for Anxiety and Depression (STAND) program29, which was developed under the UCLA Depression Grand Challenge. Participants were sent online for support or to clinical treatment depending on the degree of their depression. Participants who scored a high on the CAT-DI of 35 or 65 were assigned to online support with the help of a peer coach. those who scored 75 patients were referred to in-person clinics for psychotherapy.
Participants were asked a series of questions at the beginning of the study about their demographics and psychosocial traits. These included age, sex, education, work, and financial status; if they were partnered, divorced or single; the frequency of suicidal thoughts, intentions, or attempts; and the frequency at which they drank alcohol. Participants also scored their level of depression symptom severity on a scale of 0-100 using the CAT-DI. The CAT DI assessment was carried out every two weeks for those who received online support and weekly for those who received in-person support.
Predictors of Treatment Response
Research is focusing on personalization of depression treatment. Many studies are aimed at finding predictors that can help clinicians identify the most effective medications for each person. Particularly, pharmacogenetics is able to identify genetic variants that determine how the body's metabolism reacts to antidepressants. This allows doctors select medications that will likely work best for each patient, while minimizing the time and effort needed for trial-and-error treatments and eliminating any adverse negative effects.
Another promising approach is to build prediction models that combine clinical data and neural imaging data. These models can be used to determine which variables are most predictive of a specific outcome, like whether a drug will improve mood or symptoms. These models can be used to determine the response of a patient to treatment, allowing doctors maximize the effectiveness.
A new generation uses machine learning techniques like the supervised and classification algorithms, regularized logistic regression and tree-based methods to combine the effects of several variables and improve predictive accuracy. These models have proven to be effective in forecasting treatment outcomes, such as the response to antidepressants. These approaches are becoming more popular in psychiatry and will likely become the norm in the future clinical practice.
In addition to prediction models based on ML The study of the mechanisms behind depression is continuing. Recent findings suggest that depression is related to dysfunctions in specific neural networks. This theory suggests that individualized depression treatment will be built around targeted treatments that target these neural circuits to restore normal functioning.
One method of doing this is to use internet-based interventions which can offer an personalized and customized experience for patients. For instance, one study discovered that a web-based treatment was more effective than standard treatment in improving symptoms and providing an improved quality of life for those suffering from MDD. Additionally, a randomized controlled study of a customized approach to depression treatment showed sustained improvement and reduced adverse effects in a large proportion of participants.
Predictors of side effects
In the treatment of depression one of the most difficult aspects is predicting and determining which antidepressant medications will have very little or no negative side negative effects. Many patients are prescribed various medications before finding a medication that is safe and effective. Pharmacogenetics provides an exciting new method for an efficient and specific approach to choosing antidepressant medications.
A variety of predictors are available to determine which antidepressant is best to prescribe, including gene variants, phenotypes of patients (e.g. sexual orientation, gender or ethnicity) and comorbidities. To determine the most reliable and accurate predictors for a specific treatment, controlled trials that are randomized with larger samples will be required. depression treatment options is due to the fact that the identification of interaction effects or moderators may be much more difficult in trials that only consider a single episode of treatment per participant instead of multiple sessions of treatment over time.
In addition, predicting a patient's response will likely require information about the comorbidities, symptoms profiles and the patient's subjective perception of effectiveness and tolerability. At present, only a handful of easily assessable sociodemographic variables and clinical variables appear to be consistently associated with response to MDD. These include age, gender and race/ethnicity, SES, BMI and the presence of alexithymia.
The application of pharmacogenetics in depression treatment is still in its early stages and there are many hurdles to overcome. First, a clear understanding of the genetic mechanisms is needed, as is a clear definition of what is a reliable predictor of treatment response. Ethics, such as privacy, and the responsible use genetic information must also be considered. In the long term pharmacogenetics can offer a chance to lessen the stigma associated with mental health care and improve treatment outcomes for those struggling with depression. But, like all approaches to psychiatry, careful consideration and implementation is necessary. At present, the most effective method is to provide patients with a variety of effective depression medication options and encourage them to talk with their physicians about their experiences and concerns.