On the role of continuous physiological monitoring and time-series analysis in medical prognosis

Technological advances allow the continuous monitoring of a wide variety of physiological variables. Our recent results indicate that the statistics of the resulting time series reflects the dynamics of the underlying homeostatic regulation mechanisms. In the present contribution, we explore the role time series and continuous monitoring may play in medical prognosis and what their relation is to clinical prognostic scales that are used in medical practice. We hypothesize that prognostic scales based on time series may be more sensible, i.e., may predict health risks in an earlier state before the appearance of symptoms, but that these scales will necessarily be less specific, i.e., they reflect a wide variety of conditions of internal (such as stress, infection, disease or aging) or external origins (physical activity, outside weather conditions, etc.). Clinical prognostic scales are by construction more specific, but they tend to be based on symptoms and less sensitive to detect subclinical alterations. We propose that time-series based scales and clinical scales may play complementary roles in medical prognosis, where the former can serve as screening tests to identify an as large as possible number of at-risk individuals (true positives), and the latter as confirmation tests to avoid giving unneeded treatment to healthy subjects (true negatives).