Retrospective development and evaluation of prognostic models for exacerbation event prediction in patients with Chronic Obstructive Pulmonary Disease using data self-reported to a digital health application [article]

Francis P Chmiel, Dan K Burns, John Brian Pickering, Alison Blythin, Thomas M. A. Wilkinson, Michael Boniface
2020 medRxiv   pre-print
Self-reporting digital applications provide a way of remotely monitoring and managing patients with chronic conditions in the community. Leveraging the data collected by these applications in prognostic models could provide increased personalisation of care and reduce the burden of care for people who live with chronic conditions. This study evaluated the predictive ability of prognostic models for prediction of acute exacerbation events in people with Chronic Obstructive Pulmonary Disease
more » ... data self-reported to a digital health application. Methods Retrospective study evaluating the use of symptom and Chronic Obstructive Pulmonary Disease assessment test data self-reported to a digital health application (myCOPD) in predicting acute exacerbation events. We include data from 2,374 patients who made a total of 68,139 self-reports. We evaluated the degree to which the different variables self-reported to the application are predictive of exacerbation events and developed both heuristic and machine-learnt models to predict whether the patient will report an exacerbation event within three days of self-reporting to the application. The model's predictive ability was evaluated on self-reports from an independent set of patients. Findings Users self-reported symptoms and standard Chronic Obstructive Pulmonary Disease assessment tests display correlation with future exacerbation events. Both a baseline model (AUROC 0.655 (95 % CI: 0.689-0.676)) and a machine learnt model (AUROC 0.727 (95 % CI: 0.720-0.735)) showed moderate ability in predicting exacerbation events occurring within three days of a given self-report. While the baseline model obtained a fixed sensitivity and specificity of 0.551 (95 % CI: 0.508-0.596) and 0.759 (95 % CI: 0.752-0.767) respectively, the sensitivity and specificity of the machine-learnt model can be tuned by dichotomizing the continuous predictions it provides with different thresholds. Interpretation Data self-reported to healthcare applications designed to remotely monitor patients with Chronic Obstructive Pulmonary Disease can be used to predict acute exacerbation events with moderate performance. This could increase personalisation of care by allowing pre-emptive action to be taken to mitigate the risk of future exacerbation events. It is plausible future studies could improve the accuracy of these models by either the inclusion of symptom information recorded with greater granularity or including variables not considered in our study, for example vital signs, information on activity, local environmental data, and lifestyle information.
doi:10.1101/2020.11.30.20237727 fatcat:32yzhnchrzdd7daaqprv5wjpzm