iDEA: Drexel E-repository and Archives >
Drexel Academic Community >
College of Arts and Sciences >
Research Day Posters (COAS) >
Development of a Bayesian network model schema that builds on existing FMECAs
Please use this identifier to cite or link to this item:
|Title: ||Development of a Bayesian network model schema that builds on existing FMECAs|
|Authors: ||Madsen, John|
|Issue Date: ||23-Apr-2009|
|Series/Report no.: ||Research Day 2009 Posters|
|Abstract: ||The Institute of Medicine’s 2000 report entitled 'To Err is Human' states that as many as 98,000 people die each year as a result of medical error in the United States. Subsequent studies indicate that this may be an underestimate. Awareness of the patient safety problem has led to widespread attempts to encourage quality improvement in America, from legislation requiring incident reporting to pay-for-performance programs. Evaluating and improving process design has been recognized as a critical element in improving patient safety. The Joint Commission on the Accreditation of Healthcare Organizations (JCAHO) recommends a technique called failure mode effect and criticality analysis (FMECA), which has been widely used in improving the safety of medical processes. This tool's uses are limited because of its inability to examine the possibility of multiple errors occurring in a process.
Methods Efforts were centered on the creation of a generalizable schema that could be used in creating models more descriptive of the possibility of multiple errors contributing to undesirable outcomes. For this purpose, we used a Bayesian Network (BN) to incorporate both the process flow diagram and the probabilities/frequencies of various failures and their consequences for a given procedure. Steps from process flow diagrams used in creating FMECAs are categorized into action steps and validation steps, which are organized with potential outcomes and probabilities into a resulting matrix that represents all possible combinations of errors as well as the probability that any given error (or combination of errors) will occur. The model is first used as an influence diagram to determine which possible branches in a chain of steps may be eliminated. Once branches with higher probabilities of error are eliminated, the streamlined BN will indicate the probability that any outcome is reached by any combination of steps with any combination of errors. An existing FMECA completed for blood transfusion is used to illustrate our method.
Results The resulting model is useful for several reasons. Decision analysis can be performed to ascertain what potential errors can simply be eliminated from the process. High probability errors are noticeable, but more importantly, dangerous combinations of error are highlighted. Depending on the level of specificity achieved in the initial FMECA, specific health outcomes can be attached to specific errors, creating a diagnostic tool for use in later root-cause analyses. The model can be adjusted readily, so proposed changes in the process can be examined in a hypothetical setting before being tested in an actual health care setting.
Conclusion Creation of a BN model increases the value of time intensive labor already performed during FMECAs. This method shares some of the benefits of more sophisticated modeling approaches but builds off of the widely used FMECA framework already recommended by JCAHO. It achieves the goal of determining which combinations of error lead to undesired outcomes.|
|Appears in Collections:||Research Day Posters (COAS)|
Items in iDEA are protected by copyright, with all rights reserved, unless otherwise indicated.