Premium
Designing highly reliable adverse‐event detection systems to predict subsequent claims
Author(s) -
Helmchen Lorens A.,
Burke Maureen E.,
Wojtusiak Janusz
Publication year - 2015
Publication title -
journal of healthcare risk management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.221
H-Index - 16
eISSN - 2040-0861
pISSN - 1074-4797
DOI - 10.1002/jhrm.21167
Subject(s) - event (particle physics) , matching (statistics) , adverse effect , warrant , health care , patient safety , sample (material) , computer science , risk analysis (engineering) , medicine , actuarial science , medical emergency , business , chemistry , physics , finance , pathology , quantum mechanics , chromatography , economics , economic growth
Background To respond proactively to patient safety events, many healthcare organizations have been enhancing and customizing their event reporting systems. Yet an indiscriminate expansion of the range and number of event reports may reduce, rather than raise, risk managers’ ability to detect events that warrant a response. To avoid becoming overwhelmed by too many event reports that have little immediate operational value, risk managers therefore require a concurrent and complementary refinement of their data‐processing capabilities. Objective To examine the extent to which adverse event reports can predict subsequent claims. Data and Methods The study sample included all adverse event reports and all records of closed claims that related to patient care episodes between July 1, 2006, and May 31, 2009, at a large hospital system in northern Virginia. After matching closed claims to event reports, we fitted multivariate predictive models to identify event report entries that predict future claims. Results During the period under study, 20 151 event reports and 94 claims were filed across the health system. We were able to match 60 claims (63.8%) to at least 1 preceding event report, implying that only 0.3% of event reports preceded a subsequent matching claim. The superior prediction model identified 90% of eventual matched claims by retaining only 20% of all event reports. Conclusion Simple prediction algorithms can supplement expert judgment by screening for reports that are likely to result in a claim, thereby enabling risk managers to evaluate adverse event reports more expeditiously and to identify, and ultimately prevent, serious safety lapses more reliably.