Medication incompliance and vital signs in heart failure patients

Abstract

The success of medication treatment is a crucial part of chronic cardiac patients' therapy. Health professionals are requested to dynamically optimize this procedure (medication initiation, titration, change of medication plans) for each patient. Assessment of the person's response to treatment, and assessment of adherence to treatment are essential in supporting these medical decisions. In this work, methods that apply biosignal analysis on vital signs data are presented, aiming to detect differences due to medication incompliance. These data, although gathered in controlled conditions, encompass physiological fluctuations introduced by different factors of daily life, such as medication timing and activity. Heart Failure data are normalized to personal levels, and classification models trained with a pair of features (SBP and HR in semirecumbent position) succeed in achieving accuracy over 97% in a crossvalidation setup, in detecting 48 hrs incompliance. Additionally, the differences in incompliance patterns between heart failure and hypertensive subjects are discussed. These results constitute a promising step towards application of vital signs measurement and analysis in homecare for incompliance detection.