Comfirmation as a Key for Patient Safety: a Network Analysis of Incident Reports

According to the results of the statistical analysis on incident reports released by the Japanese Ministry of Health, Labour and Welfare, drug-related medical accidents account for more than half of all medical accidents. With respect to the causes of medical accidents, those caused by human carelessness, such as “incorrect drugs” or “incorrect drug dosages,” account for more than 40% of all medical accidents. Therefore, establishing a method to prevent the occurrence of accidents due to human carelessness is a highly contributing factor in achieving safe, worry-free medical care. Accordingly, the purpose of this research is to construct a knowledge base that is useful for examining accident prevention measures by specifying a drug group closely related to the accident as well as the causes of the accident through the structuring of bibliographical information. Natural language processing and network analysis are used as research methods. The network analysis method facilitates analysis on various relationships that are not only direct relationships but also indirect relationships, which are at least two steps away from the former. The incident reports in Japan are used as the subjects of this research. First, the frequently appearing characteristic words were extracted according to the three categories of the accident’s background, details, and solutions described in the report, and then networks using these extracted words were created. Next, using the pharmaceutical dictionary and our own rules for dissolving the “discrepancy in descriptions,” drug information was extracted from the accident report, and a drug network was created. From the above analyses it has become clear that inadequate information on drug labels, physician’s instructions, and patient misunderstanding are the more important factors in medical accidents. Moreover, drugs and the combinations of drugs that frequently cause accidents have been revealed. The POAS (Point of Act System) is known as a cutting-edge medical information system that prevents confirmation errors made by humans. Finally, using the above results, the degree of effectiveness of POAS in the prevention of drug-related accidents was calculated. Of all the drugs that were returned through POAS, the ones that had actually caused medical accidents accounted for 31.6%. In addition, with regard to the ten most frequently returned drugs through POAS, the Pearson’s Correlation between the number of drugs returned and the number of accident cases was high at 0.508, thus confirming the effectiveness of POAS implementation. From this point forward, in order to achieve an even more effective analysis, the following factors are important in incident reports: (i) unification of descriptions pertaining to drug(s) and prevention of discrepancy in descriptions; and (ii) encouraging more descriptions pertaining to the confirmation details.