Institutional members access full text with Ovid®

Using Present-On-Admission Coding to Improve Exclusion Rules for Quality Metrics: The Case of Failure-to-Rescue

Needleman, Jack PhD, FAAN*,†; Buerhaus, Peter I. PhD, RN, FAAN; Vanderboom, Catherine PhD, RN§; Harris, Marcelline PhD, RN

doi: 10.1097/MLR.0b013e31829808de
Original Articles

Background: The Agency for Healthcare Research and Quality (AHRQ) patient safety indicator “death among surgical inpatients with serious treatable complications” (failure-to-rescue) uses rules to exclude complications presumed to be present-on-admission (POA). Like other administrative data-based quality measures, exclusion rules were developed with limited information on whether complications were POA. We examine whether the accuracy of failure-to-rescue exclusion rules can be improved with data with good POA indicators.

Methods: POA-coded data from 243,825 discharges from a large academic medical center were used to develop 3 failure-to-rescue exclusion rules. Data from 82,871 discharges from California hospitals screened for good POA coding practices was used as a validation sample. The AHRQ failure-to-rescue measure and 3 new measures based on alternative exclusion rules were compared on sensitivity, specificity, and C-statistics for prediction of POA status. Using data from the AHRQ HCUP National Inpatient Sample, the alternative specifications were tested for sensitivity to nurse staffing.

Results: The AHRQ exclusion rules had sensitivity of 18.5%, specificity 92.1%, and a C-statistic of 0.553. All POA-informed specifications of exclusion rules improved the C-statistic of the failure-to-rescue measure and its sensitivity, with modest losses of specificity. For all tested specifications, higher licensed hours and proportions of registered nurse were statistically significant and associated with lower risk of death.

Conclusions: Failure-to-rescue is a robust quality measure, sensitive to nursing across alternative exclusion rule specifications. Despite expanded POA coding, exclusion-based rules are needed to analyze datasets not coded for POA, legacy datasets, and datasets with poor POA coding. POA-informed construction of exclusions significantly improves rules identifying POA complications.

*Department of Health Policy and Management, UCLA Fielding School of Public Health

UCLA Patient Safety Institute, Los Angeles, CA

Department of Nursing, Center for Interdisciplinary Health Workforce Studies, Institute for Medicine and Public Health, Vanderbilt University Medical Center, Nashville, TN

§Mayo Clinic, Rochester, MN

School of Nursing, University of Michigan, Ann Arbor, MI

Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Website, www.lww-medicalcare.com.

Supported in part by a grant from the Robert Wood Johnson Foundation Interdisciplinary Nursing Quality Initiative (INQRI).

The authors declare no conflict of interest.

Reprints: Jack Needleman, PhD, FAAN, Department of Health Policy and Management, UCLA Fielding School of Public Health, 650 Charles E. Young Dr. S., Room 31-236B CHS, Los Angeles, CA 90095-1772. E-mail: needlema@ucla.edu.

© 2013 by Lippincott Williams & Wilkins.