December 04, 2018
Machine learning techniques applied to telemonitoring datasets are better than existing algorithms at predicting chronic obstructive pulmonary disease (COPD) exacerbations requiring hospitalization and decisions to start corticosteroids, suggests a study published online in the Journal of Medical Internet Research.
“Telemonitoring of symptoms and physiological signs has been suggested as a means of early detection of COPD exacerbations, with a view to instituting timely treatment. However, algorithms to identify exacerbations result in frequent false-positive results and increased workload,” researchers wrote. “Machine learning, when applied to predictive modelling, can determine patterns of risk factors useful for improving prediction quality.”
For the study, researchers used an average 363 days of telemonitoring data for 135 patients. Data included symptoms, physiological measures, medication, COPD severity, and baseline demographic information. Weather data was also used to see whether it improved predictions.
From the data, researchers constructed predictive patterns and models fitted to training sets of patients. Then they compared performance of the predictive models with common symptom-counting algorithms.
When restricted to cases with complete data, the two most practical and traditional symptom-counting algorithms demonstrated area under the receiver operating characteristic curve (AUC) estimates of 0.60 and 0.58 for predicting admissions based on a single day's readings, according to the study. But in a real-world scenario with missing data and higher numbers of patient data and hospitalizations, algorithm performance fell.
Machine learning models performed significantly better, researchers reported. The best machine learning algorithm resulted in an aggregated AUC of 0.74. Adding weather data measurements did not improve the algorithm’s performance.
The machine learning algorithm was moderately superior to the best symptom-counting algorithm in predicting the need for corticosteroids, researchers reported.
“Early detection and management of COPD remains an important goal given its huge personal and economic costs,” researchers wrote. “Machine-learning approaches, which can be tailored to an individual's baseline profile and can learn from experience of the individual patient, … show promise in achieving this goal.”
Orchar P, Agakova A, Pinnock H, et al. Improving Prediction of Risk of Hospital Admission in Chronic Obstructive Pulmonary Disease: Application of Machine Learning to Telemonitoring Data. J Med Internet Res. 2018;20(9):e263. DOI: 10.2196/jmir.9227. Published September 21, 2018.
For articles by First Report Managed Care, click here
To view the First Report Managed Care print issue, click here