Food for Thought: Quantifying Eating Disorders One (Mega)Byte at a Time with Data Driven Approaches

Eating disorders (ED) are mental health syndromes characterized by pathological behaviours with devastating impact on the physical health of patients. Therefore, both mental and physical care need to be combined during treatments for ED. Regarding physical health management, weight and cardiovascular risk factors are usually monitored. The psychopathology management follows similar heuristics as for other mental health disorders. Related to the complex mind-body interplay, recovery from ED is a slow and irregular process.
The treatment attribution could potentially be enhanced by building a better understanding of the continuous relations between bodily changes (i.e., weight, physiology) and the mental state in patients suffering from ED. Such relations could serve as markers to establish subgroups of patients whose mental and physical care can be managed similarly. For now, quantified clinical information on both dimensions is sparse in quantity (at sparse time points) and in quality (low variety of information), something that needs to be addressed before optimized patient status assessment and follow up is possible in clinical practice.
In this project, we investigate possible metrics that could inform the therapist about the patient status in relation to healthy controls, treatment outcomes, and scores from established psychometric tools. We focus on scenarios in which quantified data could aid data-driven decisions enabled by new technologies. We get acquainted with the challenges posed by the clinical reality, both by studying the usability of data resources available on site, and by exploring physiological sensing and emotional assessments for naturalistic follow-up of patients in treatment. In our explorations we employ descriptive statistics, inferential statistics for hypothesis testing, and machine learning methods for prediction.
Firstly, using information routinely collected during hospital contacts (e.g., weight, blood pressure, prior hospitalizations), we explore its relation to treatment outcomes. Secondly, we investigate whether autonomic nervous system’ biosignals (e.g., electrocardiogram, electrodermal dermal activity and skin temperature) and short self-reports on stress and feelings towards food are useful to characterize ED patients in treatment. Finally, a physical task designed to study ED physiological reactivity is assessed as a possible fitness evaluation method to be employed by subjects that cannot endure excessive effort in psychophysiological studies.