Novel use of REDCap to develop an advanced platform to display predictive analytics and track compliance with Enhanced Recovery After Surgery for pancreaticoduodenectomy

Abstract

Background: Prediction models are increasingly being used with clinical practice guidelines to inform decision making. Enhanced Recovery After Surgery (ERAS^®) protocols are standardized care pathways that incorporate evidence-based practices to improve patient outcomes. Predictive analytics incorporated within a data management system, such as Research Electronic Data Capture (REDCap), may help clinicians estimate risk probabilities and track compliance with standardized care practices.

Methods: Predictive models were developed from retrospective data on 400 patients who underwent pancreaticoduodenectomy from 2008 through 2014. The REDCap was programmed to display predictive analytics and create a data tracking system that met ERAS guidelines. Based on predictive scores for serious complication, 30-day readmission, and 30-day mortality, we developed targeted interventions to decrease readmissions and postoperative laboratory tests.

Results: Predictive models demonstrated a receiver-operating characteristic area (ROC) ranges of 641-856. After implementing the REDCap platform, the readmission rate for high-risk patients decreased 15.8% during the initial three months following ERAS implementation. Based on predictive outputs, patients with a low-risk score received a limited set of postoperative laboratory tests. Targeted interventions to decrease hospital readmission for high-risk patients included home care orders and post-discharge instructions.

Conclusions: The REDCap platform offers hospitals a practical option to display predictive analytics and create a data tracking program that meets ERAS guidelines. Prediction models programmed into REDCap offer clinicians a support tool to assess the probability of patient outcomes. Risk calculations based on predictive scores enabled clinicians to titrate postoperative laboratory tests and develop post-discharge home care orders.

Keywords: Clinical prediction rule; Data management systems; Mortality; Outcome; Pancreaticoduodenectomy; Patient readmission.