Actimetry Protocol in COPD Patients

COPD, Inactivity, Physical

COPD, Physical Inactivity, Medical Decision, Machine Learning, Respiratory Medicine

Recently, the principal investigator published an EI predictive Machine Learning algorithm based solely on clinical data, without any physical activity measures, collected from 1 409 patients. The GOLD standard of EI was defined on the basis of interrogation criteria. Patients considered as EI reported walking less than 10 minutes per day on average, and the pulmonologist judged that the patient had mainly domestic activities. Despite the subjective nature of the GOLD standard, the algorithm validated on a test sample had an error rate of only 13.7% (AUROC: 0.84, CI95% \[0.75-0.92\]). In the total study population (n=1409), 34% of patients were ultimately classified as EIs by the algorithm, in agreement with the results of studies using actimetry as the GOLD standard. The principal investigator now wish to verify and improve the validity of the MLA on a new smaller population of 104 patients, using a physiological GOLD standard such as three-dimensional actimetry.

It has been shown that COPD patients have a significantly decreased daily physical activity (DPA) compared to matched subjects. Moreover, the severity of inactivity is correlated with several prognostic indices such as the frequency of exacerbations, quality of life and mortality. These findings lead to the recommendation, with a level of evidence A, of DPA in the context of medically supervised respiratory rehabilitation programs and/or by encouraging patients to participate in programs promoting physical activity. However, despite the established benefits, it is estimated that this rehabilitative management actually involves only 10% of the patients who should benefit from it. Among the various causes of this situation, the underestimation of excessive inactivity (EI) by pulmonologists is one of the causes of this care deficit. Currently, only actimetry can accurately assess the patient's level of physical activity. To alert pulmonologists to this excessive situation justifying priority care, without resorting to actimetry, the aCCPP developed a Machine Learning Algorithm (MLA) based on clinical data from the Colibri-BPCO digital consultation that predicts excessive inactivity. In this study, the GOLD standard EI was defined using clinical criteria summarized below. EI patients reported walking for an average of less than 10 minutes per day, and the pulmonologist judged on questioning that the DPA was indeed essentially domestic. The objective of the MLA was to correctly classify EI subjects versus obviously active subjects hereafter referred to as Overtly Active (OA). The MLA was validated on a test sample with an error rate of 13.7% (AUROC: 0.84, IC95% \[0.75- 0.92\]). In the total population studied (n=1409), 34% of patients were finally classified as EIs, in line with the results of studies using actimetry as the GOLD standard. Following the publication of this work , the principal investigator would like to verify the validity of the algorithm on a new population using the recognized GOLD standard: three-dimensional actimetry measurements.

No related papers found yet.

None listed

France