Measuring Changes in the Distribution of Incident-Outcome Severities and the Method's Implications for Health Care Policy

• Activity Number: 82
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): William Goodman*+ and Kim Sears
• Companies: University of Ontario Institute of Technology and Queen's University
• Keywords: Incident-outcome severities ; severity-distribution analysis ; healthcare policy ; health and safety statistics ; medical errors ; medication administration errors

Abstract

Healthcare policy makers are highly constrained by costs when deciding among alternative measures. Not all dangers which could be addressed pose the same relative risks that, if an incident occurs, the outcomes will be severe. This paper describes methods to assess, statistically, the effect sizes and significance, of observed systematic upward (or downward) shifts in the distributions of incident-outcome severities, among alternative scenarios. Adapting resampling-based methods developed originally for industry and government safety data, to identify factors most associated with more extreme outcomes, we demonstrate their potential to inform policy that addresses widely recognized risks to hospital patients from medication administration errors. Data were collected from registered nurses at three Canadian pediatric care centers. With the proposed metrics, we find that if a medication error occurs, only four of the 16 factors considered significantly correlated with increased tendencies towards more severe-level outcomes. For avoiding the most serious outcomes of medication errors, targeting the identified factors would be more cost effective than more scattered approaches.