Background

Medication prescribing for the elderly is a complex task that requires special care and increased patient monitoring, while appropriate medications are vital for keeping elderly patients healthy, especially those with multiple diseases who use polypharmacy. Japan is known as the most aged country with 26.5% of the population older than 65 [1]. Therefore, a regulatory meeting was established in 2017 to discuss appropriate prescribing and medication use in the elderly and ensure adaptability to the changing medical status of patients [2]. Furthermore, this strategy is expected to reduce the side effects and polypharmacy and ensure reasonable medical costs [2]. Japan has been releasing new drugs to the world, and its regulatory authorities and pharmaceutical companies are expected to provide enough information and formulate precautions for the use of drugs in the elderly. However, since clinical trials mostly exclude the elderly, clinical information for this demographic is scarce for most new drugs, studies are eagerly anticipated, which is true in every country and region.

To aid physicians to prescribe appropriately, the Japan Geriatric Society (JGS) first published guidelines for safe pharmacotherapy in the elderly, and a list of potentially inappropriate medication uses in 2005 [3], which is recognized as the Japanese version of the Beers criteria [4]. It was updated in 2015, providing the Screening Tool for Older Persons’ Appropriate Prescriptions for the Japanese (STOPP-J) for drugs to be prescribed with special caution and drugs to be considered for treatment [5]. Ahead of this, in Europe, the Screening Tool of Older Persons’ potentially inappropriate Prescription (STOPP) [6, 7], proposed by the Ireland study group, is reported as a useful guide for identifying potentially inappropriate medications, particularly for hospital inpatients [8].

Recent evolutions in medical informatics and computerization have enabled researchers to use various databases and analytical tools in their studies. Studies using the databases of regulatory authorities, insurance claims, and medical records, as well as patients’ reports have become popular in public health disciplines and drug development phases recently as well as post-marketing phase. However, there are still challenges associated with the methods of collection, coding, and analysis of data for assessing the accuracy of medication use. The use of identical names or a systematic code for drugs enhances the efficiency of research with large-scale databases [9]. However, coding of medication data depends on the regulatory system, which varies between countries. For example, the National Drug Code in the US (https://www.fda.gov/Drugs/InformationOnDrugs/ucm142438.htm.) differs from the BNF codes (https://data.gov.uk/dataset/176ae264-2484-4afe-a297-d51798eb8228/resource/bac33489-b3dc-47ec-b688-da9cf40e25bd) in the UK. For research on medicine use in the elderly, the AGS Beers criteria [4], STOPP/START [7], and JGS guidelines [5] have been used. However, these guidelines mainly provide drug categories and medication considerations, which lack consistency (Table 1). In addition, they do not specifically define molecular entities, so that researchers have to select drug substances to be studied. Therefore, variation often occurs when using databases to investigate drug use. To facilitate computerized work with databases, Groot et al. [8] proposed a uniform coding for drugs approved in the Netherlands, compliant with the STOPP/START. On the other hand, Japan has several drug-coding systems, depending on the regulatory objective, such as those for labeling information, reimbursement of medical fees, and a third for logistics, but they are solely for domestic use. A variety of proprietary databases is currently available in Japan, and medical terms are mostly compared between them using the International Classification of Diseases (ICD) [10] or the Medical Dictionary for Regulatory Activities (MedDRA) [11], which are internationally recognized in pharmacoepidemiological studies. Unfortunately, due to nationally defined drug coding based on approval, the indications differ, and the classification of medication data may not be consistent across databases. Furthermore, additional efforts for drug identification and matching are necessary when coding systems are different. This process could introduce mismatching and misinterpretation flaws into studies using multiple databases. Therefore, it would be worthwhile to standardize drug codes for international use, e.g., global pharmacovigilance, as the aforementioned Dutch group did by identifying STOPP/START drugs at the substance level and using international coding systems [8]. However, while a local version of the Beers criteria and STOPP/START has been proposed in Japan [12, 13], no guidelines have been presented on how to encode medications. In this paper, we present a proposal to encode drugs in JGS medication guidelines using the Anatomical Therapeutic Chemical Classification System (ATC) [14], supporting the extraction and validity of the medication use data.

Table 1 Characteristics of pharmacotherapy criteria for older adults
Full size table

Methods

Process for listing drug substances

This study was set up voluntarily by three physicians (Akishita, Kojima, and Ishii) and two pharmacists (Akazawa and Nomura) who have experience with studies on drug use in the elderly, using both clinical observations and databases. The STOPP-J 2015 developed by the JGS through systematic review, repeated group discussion, and review by the related academic societies, followed by public consultation [15], was the basis for this study. This JGS tool for medications for older persons does not include details on drug dosage, frequency, or duration of administration, but rather includes drug categories or names. The first step involved drafting a list of approved proprietary names based on JGS guidelines with the support of the Japan Pharmaceutical Information Center (JAPIC), an organization that provides drug information and codes to the Ministry of Health, Labour and Welfare, and pharmaceutical companies. JAPIC provided the English and Japanese names of medicinal substances, supervised by Akazawa and Nomura. The National Health Insurance Drug Price List as of February 2017 [16] was referenced for substance names in Japanese. They included all relevant active substances approved in Japan, which were grouped into specific categories, e.g., the statin category included atorvastatin, simvastatin, pitavastatin, pravastatin, fluvastatin, and rosuvastatin.

In the second step, the physicians, Kojima and Ishii, reviewed the drafted drug names in parallel from the perspectives of clinical treatment of the elderly, supervised by Akishita to reach a consensus. Since the JGS guidelines suggest controlling long-term medication for older persons to avoid untoward systemic adverse events occurring in any case where a drug is used or unused, external drugs and injections were excluded, except for self-injection. Simultaneously, the pharmacists matched the Japanese drug names with the national price list and the English drug names with ATC codes according to pharmacological criteria. Yonekawa helped encoding work. All possible oral indications were considered. Table 2 shows the criteria used to choose substances, categories, and codes.

Table 2 Procedures and concepts for listing drugs and codes
Full size table

Drug classification systems

There are two major global drug classification systems; the Anatomical Therapeutic Classification by the European Pharmaceutical Market Research Association (EPhMRA) [17] and the ATC created by the World Health Organization (WHO) Collaborating Centre for Drug Statistics [14]. Our study used the WHO ATC classification, which has codes at the substance level known as the 5th level.

In Japan, several drug-coding systems exist. The Japan standard commodity classification includes classes for drugs, which appear similar to the 3rd level of the ATC. However, it should be used with caution since it has not been updated since 1990 and, therefore, numerous new drugs are coded as “others.” Based on this system, the National Health Insurance Drug Price List [16, 18] provides 7-digit drug codes at substance level, similar to the ATC 5th level, and 12-digit codes at the product level. In contrast with the ATC process, we extracted the code using the first 7-digit  numeric code from the 12-digit alphanumeric code to represent the substance level.

The selected drugs and the corresponding codes, proposed first by Nomura and Yonekawa, were compared with those formulated separately by the Japanese system vendor, Data Horizon Corporation (https://www.dhorizon.co.jp). Then, the differences were checked and returned to both the authors and the corporation to reach a consensus.

Results

The drug list is presented in Tables 3, 4, and 5. Of the 236 encoded drug substances, 197 matched the 5th level of the ATC, along with 10 of 26 combinations. No ATC was available at the 5th level for 39 substances and, therefore, they were identified as 4th level substances. If multiple ATCs at the 5th level were available for one substance, the best pharmacological match or the indication-matched ATC was selected and presented with the rest of the possible ATCs. These lists are available as a PDF and spreadsheet at http://docrd.jp/ftp_up/STOPP-J%20List.pdf and http://docrd.jp/ftp_up/STOPP-J%20List.xlsx, and also on the JGS web page for the STOPP-J in Japanese, http://www.jpn-geriat-soc.or.jp/tool/pdf/list_02.pdf and http://www.jpn-geriat-soc.or.jp/tool/xls/list_03.xlsx. Since the JGS’s list was prepared as a support tool for daily medical practices, medicines rarely used or withdrawn were excluded from our list. Medicines used for short-term treatments were also excluded. The STOPP-J shows all Insulin products as drugs to be prescribed with special caution, however, if describing more accurately it recommends prescribers consider to stop sliding scale administration. This indicates that insulins can be prescribed and, therefore, they are excluded from coding.

Table 3 Proposal for drug coding of “List of drugs to be prescribed with special caution” 1)
Full size table
Table 4 Proposal for drug coding of “List of drugs to consider starting” 1)
Full size table
Table 5 Proposal for coding the combination drugs of “List of drugs to be prescribed with special caution” and “List of drugs to consider starting”
Full size table

Combination products were separately listed (Table 5) and were divided into three groups, consisting of one where the combination of exact substances was found in the ATC, such as amlodipine besylate and irbesartan (C09DB05), the ATC representing combination such as levodopa and decarboxylase inhibitor (N04BA02), and another where each substance had an individual ATC. Among the constituent substances in the latter case, the ATCs were presented only for the JGS listed substances. For example, only aspirin was selected from BUFFERIN Combination Tablet® with aspirin, aluminum glycinate, and magnesium carbonate. If the combination product consisted of the same substances in different proportions, different codes were assigned by the price list. For example, combinations of atorvastatin calcium hydrate, and amlodipine besylate were coded 2190101, 2190102, 2190103, or 2109104, depending on their compounding ratio. In addition, drugs can be categorized differently in the Japanese pricing list and have several codes; for example, BUFFERIN Combination Tablet® is 1143010 as “antipyretics, analgesics, and anti-inflammatory agents” and 3399100 as “other agents relating to blood and body fluids” All codes are listed in the tables.

Discussion

Identification and application of STOPP-J drug substances

The efforts to reduce inappropriate drug use in elderly patients are likely to have a substantial impact on reducing drug-related morbidity. One major required step is a change in the prescription behavior of physicians, which is influenced by their knowledge and alert systems involving pharmacists, computerized reminders [19], and promotional information from pharmaceutical companies [20]. The current JGS guidelines provide concept and review steps for prescribing to the elderly but do not fully detail specific substances. Thus, our computerized database of standard drug substances, reflecting the STOPP-J with a corresponding coding system, will provide an efficient way to improve physician knowledge about medication for the elderly.

This study revealed that some substances approved in Japan were omitted from the ATC classification system, which was also reported by Groot et al. [8] in reference to the STOPP/START. This may occur when a drug is marketed in Japan only, and the substance or combination is not registered with the WHO Collaborating Center for Drug Statistics. When other countries have the same situation, it would also be necessary to set up the framework to ask the WHO Collaboration Center to include medicinal substances limited to them. This would enhance ATC completeness. To support ATC users, the Uppsala Monitoring Center/WHO Collaborating Centre for International Drug Monitoring does provide the WHO Drug Global with drug information, including Japanese approved drugs and referencing ATC codes at the 5th level, for global pharmacovigilance [21]. Their service supports linking Japanese substances with the ATC, and major global companies use this service for internal databases. It is important to make ATC codes useful in pharmacovigilance and pharmacoepidemiology studies for all Japanese and worldwide drugs and create an official framework to register new substance as soon as possible. This would facilitate drug safety monitoring by pharmaceutical companies and the review of drugs at the class and substance levels. We excluded some medicinal products from the first listing step of the drug indication categories. This paradoxically suggests that researchers run the risk of including appropriately prescribed drugs when extracting data from the drug classification systems.

When the Beers criteria were applied to studies on Japanese elderly patients, hospitalization risk was higher in potentially inappropriate medication users [12] and, in contrast, no association was observed between potentially inappropriate medication use and adverse outcomes [13]. The study using STOPP and START addressed the notion that potentially inappropriate prescribing increased healthcare utilization [22]. Although some drug utilization studies have been reported on the STOPP-J, the future applications of our results to pharmacoepidemiologic clinical studies are worth considering in Japan, similar to a previous study using the Beers criteria in Japan [12, 13]. The use of large databases has become more sophisticated, and 13 Japanese healthcare databases are acknowledged by other entities [23] (e.g., JMDC Claims Database® [24], which provides the names and ATC codes of drugs prescribed from 2005). Some unlisted domestic databases also exist, including the National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB), which maintain data records from April 2013 provided by the ministry [25]. Currently, no ATC codes are available in the NDB, but the National Health Insurance Drug Price List codes are provided and, therefore, our proposed codes can be used. Another database is the Japanese Adverse Drug Event Report database (JADER), which records spontaneous reports of adverse events to the regulatory agency and lists drug names in plain text, without codes [26]. MID-NET is another prospective database, which was launched in April 2018 by the regulatory agency [27]. It is noteworthy that global comparisons based on the guidelines for medication in the elderly would be complicated or difficult to analyze because substances and their corresponding codes vary.

Currently, there are many therapeutic guidelines and principles for the proper use of medicines, and different definitions are presented worldwide or even in certain countries. Since those guidelines are to be updated periodically in several years, the guidelines propose their philosophies and examples, without identification of drugs. Therefore, interpretation and practice tend to vary by users. When adopting the guidelines, it is important to first define drugs of interest at the component molecule level; however, papers that do not identify the studied drug names might exist. In this research, with reference to the research method of Groot et al. [8] of Ireland, we presented concrete pharmaceutical molecules intended by the STOPP-J proposed by the JGS and proposed corresponding drug codes to be widely used in Japan. The results of this research are expected to be helpful in designing research and validating the actual condition of medical service at a clinical institute. Another important application is to import the drug code list into electronic prescription systems and health information systems so that the system can aid physicians in prescribing cautiously. This application is expected to be used in practice in the near future.

Limitations of using the list

This study was limited to Japanese drugs for internal use, except insulin, because the JGS guidelines focus on the long-term use of drugs to promote appropriate medications and avoid systemic adverse events in the elderly. The study also excluded drugs mainly used for short-term treatments of less than 1 month, e.g., antipyretics. In addition, based on the JGS, the target population in our list comprised patients older than 75 years who are with or without frailty, which is quite different from other guidelines. The drug list would be useful in research to understand the status of drug prescribing or hypothesize about the trends in total drug use and polypharmacy. However, more information such as dosage regimens and comorbidities is normally required to answer clinical questions. Users also need to consider how to interpret the output. For example, the alerted drug should be able to be monitored or stopped for individual patients. Because the JGS tool is not meant to be a prescription rule, but rather provides information to support physicians’ judgment when prescribing, the dosage regimen and underlying diseases should be mentioned. Lastly, a periodic update of the list is critical for efficient use in practice.

This was the first challenge to identify the STOPP-J substances to be coded. Some difficulties were found through the work in the interpretation of the STOPP-J, for example, insulins, and healthcare data users may misunderstand what the guidelines really proposed. In addition, new medicines need to be timely evaluated to determine whether they should be prescribed with special caution or considered for medication.

Conclusion

The STOPP-J drug list is proposed in this study as a starting point for discussion for researchers. Our consolidated lists can be used for pharmacoepidemiological database studies. Some WHO ATC codes were omitted owing to regionalized drug availability or combination drugs, which must be considered when using or interpreting the present data.