By Philippe Salamitou, SRETT CEO
Artificial intelligence and data science are complementary and powerful tools, sometimes the subject of fantasies, which must be evaluated in terms of their business uses.
Data science draws on various disciplines (mathematics, statistics, algorithms, advanced computing) to extrapolate knowledge from large amounts of raw data. It is used in the development of digital medical devices such as Vestalis for remote respiratory monitoring.
The Vestalis platform already capitalises on data science to:
1. Provide healthcare professionals with the most relevant indicators. Vestalis condenses essentially redundant information from connected devices and extracts indicators that help professionals decide whether or not to intervene with the patient. We have thus been able to show that a set of five pieces of information is statistically predictive of future compliance by patients undergoing CPAP treatment: average duration of use over seven days, regularity of duration of use over the last 14 days, average duration of use in the previous month, excessive leak alerts, and length of treatment.
2. Optimise alert algorithms. Within the framework defined by the HAS (1) and learned societies for remote monitoring, data science helps to determine which versions of algorithms are most suitable. A preliminary study conducted by Vestalis shows that while threshold alerts are not very sensitive to their settings in terms of their ability to detect specific situations, variation alerts are sensitive and probably require advanced statistical studies.
3. Develop more relevant alert algorithms. This involves exploiting information that is currently underused in home care, such as time series of data transmitted remotely by connected devices, the history of assessment results carried out by professionals, PROMs, PREMs and, more generally, all documented clinical data relating to the patient. Vestalis stores this data in a structured manner and facilitates its secondary use to enhance databases such as Renavo-tel, to conduct retrospective studies, and ultimately to develop decision support algorithms through machine learning.
4. Define care protocols based on data from telemonitoring: when should the first home visit take place after treatment has been initiated? Should it be triggered on the basis of an indicator? When should a patient who is not complying sufficiently be called? Following on from studies (supported by SRETT) already conducted on these issues, the statistical analysis of large volumes of telemonitoring data can now be used to develop new protocols.
A preliminary study conducted by Vestalis shows that while threshold alerts are not very sensitive to their settings in terms of their ability to detect specific situations, variation alerts are sensitive and probably require advanced statistical studies.
The entire Vestalis team remains committed, alongside its partners, to ensuring that data science contributes to the continuous improvement of respiratory patient care.
See you soon, The Vestalis team
References:
1. HAS opinion ‘Remote medical monitoring of patients with chronic respiratory failure’ dated 18 January 2022, available at: https://www.has-sante.fr/upload/docs/application/pdf/2022-01/avis_referentiel_insuffisance_respiratoire_chronique.pdf
2. Alves Pegoraro J, Guerder A, Similowski T, Salamitou P, Gonzalez-Bermejo J, Birmelé E. Detection of COPD exacerbations with continuous monitoring of breathing rate and inspiratory amplitude under oxygen therapy. BMC Med Inform Decis Mak. 25 February 2025;25(1):101.
3. PROMs (patient-reported outcome measures) assess the results of care as perceived by patients, while PREMs (patient-reported experience measures) assess the experience of care as experienced by patients.
4. Hoet F, Libert W, Sanida C, Van den Broecke S, Bruyneel AV, Bruyneel M. Telemonitoring in continuous positive airway pressure-treated patients improves delay to first intervention and early compliance: a randomised trial. Sleep Med. 2017 Nov;39:77-83