[Skip to Navigation]
Views 91
Citations 0
Comment & Response
August 2016

Question About the Analysis of Missing Data and Identification of an Error—Reply

Author Affiliations
  • 1National Surgery Office, Veterans Health Administration, Glendale, Colorado
  • 2The George Washington University, Washington, DC
JAMA Surg. 2016;151(8):782. doi:10.1001/jamasurg.2016.0130

In Reply We appreciate the thoughtful letter by Wada et al in response to our study.1 In Veterans Affairs Surgical Quality Improvement Program (VASQIP)–assessed cases, data for risk assessment and postoperative events are obtained via 1 of 2 methods. The majority of data are collected manually by a surgical quality nurse, including preoperative, intraoperative, and 30-day postoperative outcome data. The manually collected data are augmented by automatic data transmission from other VA information systems, including demographic characteristics, admission dates, and laboratory values. Of the 305 data fields collected in a VASQIP-assessed case, 82 variables are used for risk prediction and outcome measurement. These fields, when missing, are filled using an IVEWare multiple imputation program.2 Overall, 47 preoperative variables (57.3%) are collected manually, 12 intraoperative variables (14.6%) are collected automatically, and 23 postoperative variables (28.0%) are collected manually.

×