A joint survival-longitudinal modelling approach for the dynamic prediction of rehospitalization in telemonitored chronic heart failure patients

Abstract

Telemonitoring in chronic heart failure involves remote monitoring, by clinicians, of daily patient measurements of biomarkers, such as blood pressure and heart rate. As a strategy in heart failure management, the aim is for clinicians to use these measurements to predict rehospitalization, so that intervention decisions can be made. This is important for clinical practice since heart failure patients have a very high rehospitalization rate. We present a dynamic prediction approach, based on calculating dynamically-updated patient-specific conditional survival probabilities, and their confidence intervals, from a joint model for the time-to-rehospitalization and the time-varying and possibly error-contaminated biomarker. We quantify the ability of the biomarker to discriminate between patients who are and those who are not going to get rehospitalized within a given time window of interest. This approach does not only provide a sound statistical modelling approach to the substantive problem, a problem which to the best of our knowledge has not previously been addressed using a statistical modelling approach, it also provides clinicians with a valuable additional tool on which to base their intervention decisions, and thus provides immense contribution to heart failure management.