Display Settings:

Format

Send to:

Choose Destination
See comment in PubMed Commons below
Int J Qual Health Care. 2009 Aug;21(4):292-300. doi: 10.1093/intqhc/mzp023. Epub 2009 Jun 19.

Analysis of unintended events in hospitals: inter-rater reliability of constructing causal trees and classifying root causes.

Author information

  • 1Netherlands Institute for Health Services Research, PO Box 1568, NIVEL, 3500 BN Utrecht, The Netherlands. m.smits@nivel.nl

Abstract

BACKGROUND:

Root cause analysis is a method to examine causes of unintended events. PRISMA (Prevention and Recovery Information System for Monitoring and Analysis: is a root cause analysis tool. With PRISMA, events are described in causal trees and root causes are subsequently classified with the Eindhoven Classification Model (ECM). It is important that root cause analysis tools are reliable, because they form the basis for patient safety interventions.

OBJECTIVES:

Determining the inter-rater reliability of descriptions, number and classifications of root causes.

DESIGN:

Totally, 300 unintended event reports were sampled from a database of 2028 events in 30 hospital units. The reports were previously analysed using PRISMA by experienced analysts and were re-analysed to compare descriptions and number of root causes (n = 150) and to determine the inter-rater reliability of classifications (n = 150).

MAIN OUTCOME MEASURES:

Percentage agreement and Cohen's kappa (kappa).

RESULTS:

Agreement between descriptions of root causes was satisfactory: 54% agreement, 17% partial agreement and 29% no agreement. Inter-rater reliability of number of root causes was moderate (kappa = 0.46). Inter-rater reliability of classifying root causes with the ECM was substantial from highest category level (kappa = 0.71) to lowest subcategory level (kappa = 0.63). Most discrepancies occurred in classifying external causes.

CONCLUSIONS:

Results indicate that causal tree analysis with PRISMA is reliable. Analysts formulated similar root causes and agreed considerably on classifications, but showed variation in number of root causes. More training on disclosure of all relevant root causes is recommended as well as adjustment of the model by combining all external causes into one category.

PMID:
19542181
[PubMed - indexed for MEDLINE]
Free full text
PubMed Commons home

PubMed Commons

0 comments
How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for HighWire
    Loading ...
    Write to the Help Desk