This study was conducted in 17 LTC facilities in Japan located throughout the country. We collected the MDS assessment data on 1883 patients aged 65 years and over who were assessed between January and July 2002. Because data on medication prescription for 214 patients were missing, they were excluded. As a result, the database was constructed from the data for the 1669 patients whose data were complete (477 in 8 NHs, 374 in 5 HFEs, and 818 in 4 LTCHs). There were no differences in demographic characteristics (gender, age) between the 1669 subjects of this study and the 214 who were excluded.
The MDS instrument provides individual level data on the following: background information, such as age, gender, payment source; patient status such as cognitive patterns, physical functioning; and the care provided. Trained staff in each facility filled in the MDS form for each patient by using information obtained through interviews, observations and chart reviews. The MDS also includes detailed information on the medication prescribed during the last 7 days, including the names and doses of the drugs prescribed, their route of administration, and total dosage. A database that included scheduled medication, non-scheduled medication, and PRN medication used at the assessment reference date was constructed. It also included oral medication, external preparations, and injections, but over-the-counter medications were excluded because the data were incomplete.
We also used the MDS ADL Self-Performance Hierarchy in the MDS assessment database, to obtain a composite score for ADL functional status [52
]. The scale ranges from 0 (independent) to 6 (total dependence). In this study, a score of 2 (limited impairment) or more were classified as having an ADL disability. Cognitive impairment was assessed by the Cognitive Performance Scale (CPS) [53
], which ranges from 0 (intact) to 6 (very severe impairment), and a score of 2 (mild impairment) or more were classified as cognitively impaired. Depression was scored by the Depression Rating Scale (DRS) [54
], which ranges from 0 to 14, and a score of 3 or more were classified as depressed as defined by the developers of the scale.
We used the 3rd version of the Beers criteria [8
] to identify prescription of potentially inappropriate medication, which are more useful for screening prescriptions that include potentially inappropriate medication than others. We applied the 2003 Beers criteria in this study even though the data were collected in 2002, before the publication of the 2003 version, because we concluded that the differences between the versions would have little impact in Japan since very few physicians are familiar with the Beers criteria and the later version was more comprehensive. We thought that the 2003 version served our purpose of estimating the current prevalence of inappropriate medication use in Japanese LTC facilities based on the current guidelines.
The 2003 Beers criteria consist of 2 lists. One is a list of 49 individual medications or medication classes that are inappropriate for patients 65 years or older regardless of their disease or condition. The other is a list of 56 medications or medication classes in 19 diseases or conditions for which they should be avoided.
In this study, we focused on the 30 medications or medication classes and 15 diseases or conditions. All the medications, medication classes, and diseases or conditions which were excluded from the analysis are indicated in Table .
Inappropriate medication criteria excluded from the analysis
In addition, Beers criteria include medications, such as indomethacin and diphenhydramine, that are frequently used as external preparations in Japan. If limited to external use, the risk of systemic adverse effect should be low. Therefore, we decided to exclude external preparations.
A multiple logistic regression analysis was performed to identify predictors of potentially inappropriate medication use in the patients treated with at least 1 medication. The dependent variable was inappropriate medication use independent of disease or condition. Independent variables were divided into 2 groups. The first group consisted of patient variables, such as age, gender, abnormal laboratory test results in the last 90 days (which were defined as laboratory values that were abnormal when compared to standard values), physical restraint for the last 7 days, ADL disability, cognitive impairment, depression, length of stay, number of diseases, number of medications used per day, medication cost per day, and psychotropic drug use (as defined by the Narcotics and Psychotropics Control Law in Japan). The second group consisted of facility variables and were facility type, method of reimbursement for the cost of medication, and the number of beds in the facility. Medication cost per day was converted to natural logs because it had a long-tail distribution. All variables were entered into the multiple logistic regression model by the backward stepwise method. Data were analyzed by using SPSS 12.0J software for Windows.