Empirical developments in repeated assessment of individuals in their natural environment
Ecological Momentary assessment (EMA), also referred to Experience Sampling Method (ESM), is becoming increasingly popular. And with good reason: Obtaining information on single subjects is valuable. A precursor of such EMA is a longitudinal design, where only three to approximately six measurements are obtained from each subject. Now, with smart phones almost universally available, it has become relatively easy to obtain large amounts of information from single subjects, much more than six. But: How should you design such studies? How many times a day should you measure? What kind response (continuous or discrete) should you work with? Many questions exist about such studies.
Our symposium is about how such studies can be designed and about the type of analyses that reveal new insights from these large amounts of data. Below, you can find the program.
Attendance is FREE, but register below.
13.30 – 13.45 Opening
13.45 – 14.30 Eiko Fried (Leiden University) – WARN-D: Building a personalized early warning system for depression
14.30 – 15.15 Laura Bringmann (University of Groningen) – Making ESM ready for clinical practice: From complex networks to basic measurements
15.15 – 15.45 Break
15.45 – 16.30 Loes Keijsers (Erasmus University Rotterdam) – TBA
16.30 – 18.00 Drinks
You can find the abstract of the talks below.
May 11, 2023
Pieter de la Court Building, room 1A.01
Wassenaarseweg 52, Leiden University
Attendance is FREE.
If you want to become a member of VVSOR, see vvsor.nl/join
Laura F. Bringmann
Making ESM ready for clinical practice: From complex networks to basic measurements
Data obtained from ESM studies is commonly used to infer psychological networks, which are subsequently employed in clinical practice. However, predictive accuracy studies show that these networks are prone to overfitting, indicating that they predominantly capture noise, and therefore, are unsuitable for use in clinical practice. Consequently, I suggest that we should revert to the fundamental challenges that ESM data still faces, and concentrate on resolving them instead of focusing on complex networks. Specifically, we must address the issues regarding the use of VAS or Likert scales, how participants interpret these scales and items, and how interpretations evolve over time. Furthermore, the categories utilized at present seem to be too general for clinical use. Hence, to make ESM ready for clinical use, we should improve the measurements. In my lab, we are presently developing several methods to advance the situation. These include combining social networks with ESM measures, and employing self-learning items to get personalized responses on the context.
title: “WARN-D: Building a personalized early warning system for depression”
abstract: “Depression is common, debilitating, often chronic. Given that only 50% of patients improve under initial treatment, experts agree that prevention is the most effective way to change depression’s global disease burden. The biggest barrier to successful prevention is to identify those at risk for depression in the near future. To close this gap, we are conducting the WARN-D study, an effort to build a personalized early warning system for depression. In the study, we follow 2000 students over 2 years, using smartphones, smartwatches, systems science, and machine learning. Collected data will be utilized to build a personalized prediction model for depression onset. Overall, WARN-D will function similar to a weather forecast, with the core difference that one can only seek shelter from a thunderstorm and clean up afterwards, while depression may be successfully prevented before it occurs.”