Functional Data Analysis of Blood Pressure Across and During Dialysis Sessions

• Activity Number: 76
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Benjamin Goldstein*+ and Trevor Hastie and Wolfgang Winkelmayer
• Companies: Stanford University and Stanford University and Stanford University
• Keywords: Functional Data ; Electronic Health Records

Abstract

Electronic health records (EHRs) are increasingly common in medical settings where historic demographic and real-time biometric data (e.g. blood pressure [BP]) are collected and stored. One such setting is hemodialysis (HD) where a patient will regularly (3x/week) receive 3-4hrs of service over a period of months or years. Integrated in the EHR database is detailed information about demographics, laboratory results, and real-time measurements. Of particular medical interest is how a patients' hemodynamics vary over the course of and across HD sessions. A challenge in these settings is how to appropriately analyze such data. Using dense real-time data we illustrate the various medically relevant questions that can be addressed using a functional data approach. For example, BP is observed to vary based on ones HD schedule, supporting observations that the HD schedule is related to adverse events.