Quantile Regression and Change Point Analysis of Remote Patient Monitoring Data from CardioMEMS HF System

Quantile regression provides a sophisticated analytical approach for clinical data, offering deeper insights into patient variability and risk assessment than conventional regression techniques. This study delves into the mathematical underpinnings of quantile regression, highlighting its advantages over ordinary least squares (OLS) regression, particularly in the context of predicting diastolic pulmonary artery pressure (PAP). We explore how this method can be applied to guide clinical interventions more effectively. By leveraging quantile regression’s ability to model different parts of the outcome distribution, we demonstrate its potential to enhance patient care through improved identification of high-risk individuals and the development of more personalized treatment strategies. We then present a comprehensive framework for analyzing multiple changepoints in both mean and variance of time series data, with specific application to diastolic PAP monitoring in cardiac patients. The study analyzes patient cases with monitoring periods of approximately 590 days, examining their diastolic PAP measurements and clinical interventions. Through rigorous statistical analysis, we identify distinct pressure control regimes characterized by changes in both mean and variance. The methodology combines changepoint detection in mean-variance parameters with residual analysis to provide insights into pressure stability and intervention effectiveness. Key findings demonstrate that optimal pressure control can be achieved through different patterns of intervention and maintenance, with some patients showing improved stability at higher mean pressures, while others maintain lower pressures with greater intervention frequency. The research reveals that pressure stability, rather than absolute pressure values alone, may be a crucial indicator of successful management.