Conference article

International results have recently questioned the value of providing telehealthcare to all COPD patients. Results from the Danish TeleCare North trial nuanced the debate by concluding that telehealthcare would most likely only be cost-effective for patients in a subgroup of severe COPD. Machine-learning methods have been suggested as a strategy to target telehealthcare even better than clinical subgroups. Data from the TeleCare North trial was used to fit classification models in order to explore this feasibility. Three models were applied: a simple decision tree, logistic regression and a linear support vector machine. Results indicate that classification methods can be used to predict patient-level cost-effectiveness with a relatively high precision. With these methods, it is feasible to target telehealthcare even better in order to maximize survival and health-related quality of life while not overusing scarce health resources as argued by health economists and clinical advocates of rational medicine.

Pattern Recognition, Automated/classification; Cost-Benefit Analysis, Telemedicine; Pulmonary Disease, Chronic Obstructive, Denmark.