| Learning Objective #1: Explain the Japanese nationwide human error/clinical incident reporting system | |||
| Learning Objective #2: Discuss the data mining method application for massive incident database |
Objective: To test relations among variables hypothesized to affect incident of drain and tube.
Sample: 7,301 incident cases reported in 2003.
Method: Constitution of a causal model: Factor analysis was executed by using 35 items to be related to incidents. Principal factor method was used to extraction of factors. Four factors were decided by scree plot and promax rotation was used. As a result of having analyzed a factor again, four factors and 15 items were extracted. Four factors were named "Bad state of medical workers" Lack of knowledge and experience of medical worker" "Deficiency of an administration system" "Lack in situation" form item contents. A causal model comprising of two hypotheses was constituted. (1) Deficiency of an administration system is caused bad state of medical workers and lack of knowledge and experience of medical workers. (2) Bad state of medical workers and lack of knowledge and experience of medical workers are caused an error of judgment and lack of confirmation.
Analysis: Validity of the causal model assumed was viewed by covariance structure analysis.
Results: To evaluate the overall fit of the estimated model, GFI and AGFI were squared. GFI= .99, AGFI= .98. To evaluate the partial e fit of the causal model, Coefficient of determination of endogenous variables and influence index to observed variables from construct were squared. Coefficient of determination were 0.37 and 0.40. As regard influence index, those of "Lack in situation" were 0.13 and 0.20. The others were more than 0.29. Coefficients between constructs were more than 0.30.
Conclusions: It was confirmed that the estimated model explained variance covariance multiplication well.