Luchtbrug is a national digital remote self-monitoring platform created with the goal of tracking the symptoms of asthma patients, to communicate with the health care professionals and find information on, e.g., personalized medication, rescue plan. One of the main goals of this kind of e-health is to reduce the number of regular visits to the outpatient clinic, and increase the monitoring frequency in the home situation. Thus, blended care is introduced with better monitoring and fewer visits. The data is collected from over 20 Dutch hospitals across the Netherlands and comes mostly in the form of various symptom questionnaires. The 68,000 questionnaires have been collected over a period from 2018 to 2023. This dataset is unique in its kind because of its size and scope. The main questionnaire used is the internationally accepted Asthma Control Test (ACT) [1] for children (7 questions, 7-11 years) and adults (5 questions, 12+ years) The questions relate to several features known to patients with asthma. In addition, the dataset includes results from the CARAT [2] questionnaire (related to Asthma and Allergy), consisting of 10 questions.
One important advantage of self-monitoring data lies in the high frequency of measurements, allowing to better separate the measurement error from the underlying signal. However, self-monitoring data comes with its own set of challenges as the lack of a controlled environmental setting increases the number of nuisance factors (time of day, mood, level of rest) [3]. In this project, we will analyze data collected via the Luchtbrug platform, which includes both questionnaire data, as well as clinical measurements on lung functions. We want to investigate whether we can identify unusual reporting behavior that suggests worsening or an impeding emergency. If one notices different reporting behavior, one would want to take some preventative measure like sending a push notification to alert the pulmonologist.
For the analysis, we will use statistical models from the areas of anomaly detection [4] and change point detection [5]. Part of the project would involve developing useful biomarkers that indicate serious events, e.g., higher-frequency measurements collected in a short period. For this purpose, clustering analysis [6] could prove useful in an initial step meant to identify different patient reporting patterns. Once these behaviors are assigned to possible clusters (infrequent reporting, regular reporting, only during serious events), we can more easily tell apart whether an anomalous event is happening, which would not fit into the typical measuring behavior showcased by the patient.
The main research question we want to answer with this project is whether we can detect the critical events that require hospitalization, emergency visits or acute consultations, using the data collected with the platform. Despite advances in asthma management, patients still frequently suffer from severe asthma exacerbations [7]. The early identification of these exacerbations by means of digital self-monitoring will lead to a more effective monitoring of asthma critical events and to more timely interventions of healthcare personnel. A secondary question would be to explore whether the varying reporting behaviors can be distinguished from possible seasonal effects, which imply that patients change their behavior depending on season.
Contact: Gabriel Bucur (RU); Marc Oppelaar (Radboudumc); Peter Merkus (Radboudumc).
[1] https://www.asthmacontroltest.com/en-gb/welcome/
[3] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3951782
[4] https://www.vmware.com/topics/anomaly-detection
[5] https://eecs.wsu.edu/~cook/pubs/kais16.2.pdf
[6] https://www.displayr.com/understanding-cluster-analysis-a-comprehensive-guide/