Trajectory Analysis for Postoperative Pain Using Electronic Health Records- a Non-Parametric Method with Robust Linear Regression and K-Medians Cluster Analysis

• Activity Number: 170 - Nonparametric Methods for Longitudinal and Survival Data
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Yingjie Weng* and Lu Tian and Karishma Desai and Dario Tedesco and Wen-wai Yim and Steven Asch and Ian Carroll and Catherine Curtin and Kathryn McDonald and Tina Hernandez-Boussard
• Companies: Stanford University School of Medicine and Stanford University and Department of Surgery, Stanford University School of Medicine and Department of Surgery, Stanford University School of Medicine and Department of Surgery, Stanford University School of Medicine and Center for Innovation to Implementation, VA Palo Alto Health Care System and Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University and Department of Surgery, Stanford University School of Medicine and Center for Health Policy, Primary Care and Outcomes Research, Stanford University and Departments of Medicine, Biomedical Data Science, and Surgery, Stanford University
• Keywords: Electronic Health Record ; Pain Trajectory ; Non-parametric ; Robust Linear Regression ; K-medians Cluster Analysis

Abstract

Pain scores are widely monitored using electronic health records (EHR). Current methods fail to leverage all data available in EHR. We developed a non-parametric method with robust linear regression and K-medians cluster analysis to analyze longitudinal pain patterns in EHR. A robust polynomial regression on log-transformed pain score with time and its power transformations as the independent variable of interest was fitted for each inpatient stay. The time coefficients constructed clusters. A generalized mixed effects model adjusting for important patient and clinical characteristics estimated the cluster-specific pain scale trajectory. Logistical regressions examined the association between pain trajectories and patient outcomes. We identified 3627 patients receiving knee replacement surgery at a large medical center, 2009-2016; patients had an average of 25 pain scores. Four clusters were identified and one distinct pain trajectory was associated with higher risk of postoperative infections at 30, 60, and 90-day intervals. Our novel method successfully categorized patients' pain trajectories and identified a cluster of patients at higher risk for postoperative infections.