Fumito Tsuchiya

Voluntary Medication Error Reporting Program in a Japanese National University Hospital

BACKGROUND: In Japan, as in other countries, medical accidents arising from human error can seriously damage public confidence in medical services, as well as being intrinsically undesirable. OBJECTIVE: Errors voluntarily reported by the healthcare practitioners in our institution (Kanazawa University Hospital) were considered to assess the contributory factors by using the accumulated error database in the hospital information system. METHODS: Medical errors in our institution during the period from July 1, 2000, to June 30, 2002, were counted using the error reporting system database and were classified. RESULTS: The number of errors reported during the investigation period was 1378, of which 78% were reported by nursing staff. Medication errors involving administration of injectable or oral drugs to inpatients, dispensing, and prescription accounted for about 50% of that number. Among dispensing errors, 53% were detected by patients or their families and 36% by nurses. CONCLUSIONS: The best method of error prevention is to learn from previous errors. For this purpose, the error reporting program is effective. In patient safety management, it is important to take into account the potential risks of future errors, as well as to capture information about errors that have already happened. For safety management, adoption of appropriate information technology (e.g., implementation of a prescription order entry system) is effective in reducing medication errors. However, it is important to note that serious errors can also arise in computer-based systems.

Bactericidal effects of a high-power, red light-emitting diode on two periodontopathic bacteria in antimicrobial photodynamic therapy in vitro.

AIM   Light-emitting diodes have been investigated as new light activators for photodynamic therapy. We investigated the bactericidal effects of high-power, red light-emitting diodes on two periodontopathic bacteria in vitro. METHODS   A light-emitting diode (intensity: 1100 mW/cm(2) , peak wavelength: 650 nm) was used to irradiate a bacterial solution for either 10 or 20 s. Bacterial solutions (Porphyromonas gingivalis or Aggregatibacter actinomycetemcomitans) at a concentration of 2.5 × 10(6) c.f.u./mL were mixed with an equal volume of either methylene blue or toluidine blue O (0-20 μg/mL) and added to titer plate wells. The plate wells were irradiated with red light-emitting diode light from a distance of 22 or 40 mm. The contents were diluted, and 50 μL was smeared onto blood agar plates. After 1 week of culturing, bacterial c.f.u. were counted. RESULTS   The light-emitting diode energy density was estimated to be approximately 4 and 8 J/cm(2) after 10 and 20 s of irradiation, respectively. Red light-emitting diode irradiation for 10 s from a distance of 22 mm, combined with methylene blue at concentrations >10 μg/mL, completely killed Porphyromonas gingivalis and Aggregatibacter actinomycetemcomitans. CONCLUSION   High-power, red light-emitting diode irradiation with a low concentration of dye showed effective bactericidal effects against two periodontopathic bacteria.

Analysis on descriptions of precautionary statements in package inserts of medicines

Background To prevent medical accidents, users must be informed of the cautions written in medical package inserts. To realize countermeasures by utilizing information systems, we must also implement a drug information database. However, this is not easy to develop, since the descriptions in package inserts are too complex and their information poorly structured. It is necessary to analyze package insert information and propose a data structure. Methods We analyzed the descriptions of ‘precautions for application’ in package inserts via text mining methods. In order to summarize statements, we applied dependency analysis to statements and visualized their relations between predicate words and other words. Furthermore, we extracted words representing timing to execute the order. Results We found that there are four types of statements: direct orders such as “ ” (use), causative orders such as “ ” (make someone use), direct interdictions such as “ ” (do not use), and causative interdictions such as “ ” (do not make user use). As for words representing timing, we extracted six groups: ”at the time of delivery,” “at the time of preparation,” “in use,” “after use,” and “at the time of storage.” From these results, we obtained points of consideration concerning the subjects of orders in the statements and timing of their execution. Conclusion From the obtained knowledge, we can define the information structure used to describe the precautionary statement. It should contain information such as the actions described in the statement, the flag to express an order or interdiction, the subject to be ordered, and the timing.

A proposal for a drug information database and text templates for generating package inserts

To prevent prescription errors caused by information systems, a database to store complete and accurate drug information in a user-friendly format is needed. In previous studies, the primary method for obtaining data stored in a database is to extract drug information from package inserts by employing pattern matching or more sophisticated methods such as text mining. However, it is difficult to obtain a complete database because there is no strict rule concerning expressions used to describe drug information in package inserts. The authors’ strategy was to first build a database and then automatically generate package inserts by embedding data in the database using templates. To create this database, the support of pharmaceutical companies to input accurate data is required. It is expected that this system will work, because these companies can earn merit for newly developed drugs to decrease the effort to create package inserts from scratch. This study designed the table schemata for the database and text templates to generate the package inserts. To handle the variety of drug-specific information in the package inserts, this information in drug composition descriptions was replaced with labels and the replacement descriptions utilizing cluster analysis were analyzed. To improve the method by which frequently repeated ingredient information and/or supplementary information are stored, the method was modified by introducing repeat tags in the templates to indicate repetition and improving the insertion of data into the database. The validity of this method was confirmed by inputting the drug information described in existing package inserts and checking that the method could regenerate the descriptions in the original package insert. In future research, the table schemata and text templates will be extended to regenerate other information in the package inserts.

The analysis of near-miss cases using data-mining approach

We applied the data mining technique to medical near-miss cases collected by a foundation related to the Japanese Health, Labor and Welfare Ministry, and extracted information such as pairs of confusing medicines, the cause of near-miss cases in some situations, which cannot be obtained by simple aggregation calculations and descriptive statistics. We also introduce the results of text mining applied to the free-description data regarding the backgrounds and causes of near-miss cases and their counter measures.

Similarity Index for Sound-alikeness of Drug Names with Pitch Accents

Abstract Drug name similarity is one of major reasons of medical accidents. In order to prevent from the accidents, one of the best ways is to avoid approving drugs that has the names similar to that of existing drugs. It is well-known that there are two kinds of drug name similarity, look-alikeness and sound-alikeness. Nabeta et. al. proposed a look-alikeness similarity index,which excludes the sound-alikeness. Though, in Japan, oral prescription is basically prohibited, emergent situation can force a doctor to prescribe orally. In such a situation, medical accidents can occur. In this study, we proposed a sound-alikeness similarity index based on quantitative similarity of consonants. The consonant similarity was proposed based on The International Phonetic Alphabet (IPA). Overall drug name similarity is calculated based on Letter Sequence Kernel (LSK). The similarity calculation method takes account of the effect of plural pitch accents. We divided a drug name into some pieces at the position where a pitch accent changes, applied LSK to each of them, and combined them to obtain the value of the similarity index. The similarity index proposed in this study achieved relatively high correlation to the results of our experiment, r ≃ 0.8.

Construction and analysis of database on outer cases of medicines

This study reduced the burden on medical staffs by determining 37 kinds of attributes based on the outer cases of medicines collected from seven pharmacies. We constructed a database on the outer cases of medicines and analysis of it provided useful knowledge.

The similarity index of character shape of medicine names based on character shape similarity (II)

The similarity of drug names in Japanese such as ??? (Amaryl) and ??? (Almarl) causes confusion over drug names and can lead to medical errors. In order to prevent such errors, methods of computing their similarity have been proposed. However, there are no studies that take account of character shape similarity quantitatively. In a previous study, we calculated the character shape similarity by template matching technique and proposed a method of measuring medicine name similarity based on it. Although we obtained a high correlation coefficient between the medicine name similarity values and subjective evaluation, we observed some character pairs that are not similar. In this study, we improved the method of computing the character shape similarity based on the characteristic points of character and compared it with advanced methods.

The Evaluation of Pharmaceutical Package Designs for the Elderly People

In recent years, many medical accidents have been caused by the confusing designs of pharmaceutical packages and displays. For osteoporosis treatment, a common disorder of elderly females, improving display visibility and operability is especially important. However, since the osteoporosis drugs to be taken once a week are sold in various package designs, differences of viewability, understandability, and operability must be clarified from the differences of package design. We performed experiments as follows.

Application of Data Mining and Text Mining to the Analysis of Medical near Miss Cases

Not only the side effects of medicines themselves, but also their abuse, namely the lack of safety in drug usage, can cause serious medical accidents. The latter applies to the case of the mix-up of medicines, double dose or insufficient dose. Medical equipments can also cause accidents because of wrong treatment, such as wrong input to equipments and wrong power-off. In order to avoid such accidents, it is necessary to investigate past cases to identify their causes and work out counter measures. Medical near-miss cases caused by wrong treatment with the medicines or the medical equipments are strongly related to medical accidents that occur due to the lack in safety of usage. Medical near-miss cases are incidents, which could be medical accidentsavoided owing to certain factors, and happen more frequently than medical accidents. Incorporating Heinrich’s law, which shows the tendency of frequency and seriousness of industrial accidents, we estimate that near-miss cases happen three hundred times per one serious medical accident or thirty minor accidents. This can be interpreted as there being many causes of medical accidents, most of which are eliminated by certain suppression factors, which lead to near-miss cases. The rest of the causes lead to medical accidents. From this perspective, we can expect that both medical accidents and near-miss cases originate from the same type of causes, which suggests that the analysis of data on near-miss cases is valid to investigate the cause of medical accidents, since their occurrence frequency is much larger than that of medical accidents. For the reasons stated above, we analyze the data of medical near-miss cases related to drugs and medical equipments, which have been collected in previous years to determine the root cause of medical accidents caused by the neglect of safety of usage. Though simple aggregation calculations and descriptive statistics have already been applied to them, the analyses are too simple to extract sufficient information such as pairs of medicines that tend to be confused, and the relationships between the contents of incidents and the causes. To realize such analyses, we utilize data mining techniques such as decision-tree and market-basket analysis, and text-mining techniques such as the word linking method. The related works analyzing medical databy utilizing natural language processing ormachine learningwere introduced by Hripcsak et al (Hripcsak et al., 2003), who suggested the framework todetect events such as medical errors or adverse outcome.Recently,