Haizhe Jin1, Qingxing Qu2, Masahiko Munechika3, Masataka Sano4, Chisato Kajihara5, Vincent G Duffy6, Han Chen7. 1. From the Department of Industrial Engineering. 2. Industrial Engineering, Northeastern University, Shenyang, China. 3. Department of Industrial and Management Systems Engineering, Waseda University, Shinjuku, Tokyo. 4. Department of Management Information Science, Chiba Institute of Technology, Narashino, Chiba, Japan. 5. Department of Industrial and Management Systems Engineering, Waseda University, Shinjuku, Tokyo, Japan. 6. School of Industrial Engineering, Purdue University, West Lafayette, Indiana. 7. Department of Industrial Engineering, Northeastern University, Shenyang, China.
Abstract
OBJECTIVES: Medical errors are the manifestation of the defects occurring in medical processes. Extracting and identifying defects as medical error factors from these processes are an effective approach to prevent medical errors. However, it is a difficult and time-consuming task and requires an analyst with a professional medical background. The issues of identifying a method to extract medical error factors and reduce the extraction difficulty need to be resolved. METHODS: In this research, a systematic methodology to extract and identify error factors in the medical administration process was proposed. The design of the error report, extraction of the error factors, and identification of the error factors were analyzed. RESULTS: Based on 624 medical error cases across four medical institutes in both Japan and China, 19 error-related items and their levels were extracted. After which, they were closely related to 12 error factors. The relational model between the error-related items and error factors was established based on a genetic algorithm (GA)-back-propagation neural network (BPNN) model. Additionally, compared to GA-BPNN, BPNN, partial least squares regression and support vector regression, GA-BPNN exhibited a higher overall prediction accuracy, being able to promptly identify the error factors from the error-related items. CONCLUSIONS: The combination of "error-related items, their different levels, and the GA-BPNN model" was proposed as an error-factor identification technology, which could automatically identify medical error factors.
OBJECTIVES: Medical errors are the manifestation of the defects occurring in medical processes. Extracting and identifying defects as medical error factors from these processes are an effective approach to prevent medical errors. However, it is a difficult and time-consuming task and requires an analyst with a professional medical background. The issues of identifying a method to extract medical error factors and reduce the extraction difficulty need to be resolved. METHODS: In this research, a systematic methodology to extract and identify error factors in the medical administration process was proposed. The design of the error report, extraction of the error factors, and identification of the error factors were analyzed. RESULTS: Based on 624 medical error cases across four medical institutes in both Japan and China, 19 error-related items and their levels were extracted. After which, they were closely related to 12 error factors. The relational model between the error-related items and error factors was established based on a genetic algorithm (GA)-back-propagation neural network (BPNN) model. Additionally, compared to GA-BPNN, BPNN, partial least squares regression and support vector regression, GA-BPNN exhibited a higher overall prediction accuracy, being able to promptly identify the error factors from the error-related items. CONCLUSIONS: The combination of "error-related items, their different levels, and the GA-BPNN model" was proposed as an error-factor identification technology, which could automatically identify medical error factors.