A defined daily dose (DDD) has been commonly used in previous studies to quantify antibiotic consumption. Reference Natsch, Hekster, de Jong, Heerdink, Herings and van der Meer1–Reference Stanic Benic, Milanic and Monnier4 DDDs are described by the World Health Organization as the maintenance dose per day for its main indication in 70 kg-adults. 5 DDD-based metrics allowed us to compare the use of antibiotics by standardization, even in different countries, areas, or hospitals. However, applying DDDs to the child population is unsuitable due to variations in body weight. Instead, days of therapy (DOT), defined as the total number of days that patients were administered antibiotics regardless of the dose, has been used. Reference Gerber, Newland and Coffin6–Reference Singh, Steurer, Cantey and Wattier8 Using the DOT metric has advantages and disadvantages; while it is unaffected by the changes and inconsistencies of DDDs, it cannot account for dose variations. Reference Polk, Fox, Mahoney, Letcavage and MacDougall9,Reference Gravatt and Pakyz10
DOT is the primary benchmark for monitoring the use of antibiotics and assessing antibiotic resistance in children. However, there are alternative metrics to quantify the dose of antibiotics, such as neonate DDDs, recommended daily doses per 100 kg days, and the prescribed daily dose (PDD). Reference Fortin, Fontela, Manges, Platt, Buckeridge and Quach2,Reference Liem, Heerdink, Egberts and Rademaker11–Reference Gagliotti, Ricchizzi, Buttazzi, Tumietto, Resi and Moro16 The PDD is based on data that reflect the prescribed dose. To deal with body weight variations, some studies have utilized PDDs as PDD (mg/kg/day) or have combined PDD (mg/day) with body weight stratification. Reference Amadeo, Zarb and Muller17–Reference Sánchez-Huesca, Lerma, Guzmán-Saldaña and Lerma20 PDDs are flexible and helpful metrics, even with weight variations similar to those in the child population. However, this metric must be developed in each country because of differences in indications or recommendations.
Several studies regarding the consumption of antibiotics have been conducted in the pediatric field since the antimicrobial resistance action plan was adopted in 2016. 21–Reference Okubo, Uda and Kinoshita27 These studies are important for comparing differences between area or hospitalization and evaluating associations between consumption and antimicrobial resistance. Most studies have used DOT-based metrics to assess antibiotics; however, data regarding additional metrics that should be used within the child population are insufficient in Japan. Therefore, to develop a PDD methodology in Japan, we examined PDDs and assessed the relationship between PDD- and DOT-based metrics using a national administrative claims database.
Methods
Study design and setting
In this observational study, we used the Japanese Diagnosis Procedure Combination (DPC) data from April 2018 to March 2019. The DPC is a case-mix patient classification system that was adopted by ∼1,700 hospitals by 2018. Reference Matsuda, Fujimori and Fushimi28–Reference Hayashida, Murakami, Matsuda and Fushimi30 The data consist of discharge abstracts; administrative claims data from acute-care hospitals and patient background information including age, sex, International Classification of Diseases, Tenth Revision diagnosis codes, procedures, and medications. The study included patients aged <16 years at admission. Neonates aged 28 days and younger were excluded because we focused on the assessment of oral antibiotics. We excluded admissions with body-weight outliers due to data-entry errors by using a mean ±3 standard deviation (SD) cutoff for every 1 year of age. To adjust for a large body-weight distribution, we categorized patient admissions into 4 age groups: 1 month to <1 year, 1–6 years, 7–12 years, and 13–15 years. We included all hospitalized patients throughout the year; thus, some patients were classified into 2 groups.
This study was approved by the Institutional Review Board of Tokyo Medical and Dental University. In addition, we used anonymized data for the studies; thus, the requirement for informed consent was waived.
Variable definition
Age and body weight were recorded at admission. Ages were summarized as days in the 1 month to <1 year group and as years in the remaining 3 age groups. We used the medical diagnostic categories (MDCs) to indicate the inpatient diagnoses. Reference Hayashida, Murakami, Matsuda and Fushimi30 The MDC system comprises 18 categories corresponding to a single organ system or etiology. Patients are assigned a 14-digit code related to the diagnosis of the most-used medical resources.
In Japan, the ratio of amoxicillin and clavulanic acid in the tablet form of amoxicillin-clavulanic acid (Augmentin) is 2:1, which is different from the international standard ratio of 7:1. Therefore, clinicians occasionally prescribe amoxicillin and amoxicillin-clavulanic acid simultaneously to adjust the ratio. We classified the medication as amoxicillin-clavulanic acid when amoxicillin and amoxicillin-clavulanic acid were administered on the same day. The PDD of amoxicillin-clavulanic acid was shown as an amoxicillin dose, and sulfamethoxazole and trimethoprim were shown as trimethoprim doses.
Statistical analysis
We assessed patient age (days or years), sex, body weight, length of stay (LOS), and MDC in each age category. Continuous variables are shown as the median and interquartile range (IQR) or mean and SD, as appropriate. Categorical variables are presented as frequencies and percentages. We assessed the PDD metrics in 3 steps based on a previous study. Reference Porta, Hsia and Doerholt18 First, we summarized the proportion of each antibiotic among patients who were prescribed oral antibiotics. We did not assess parenteral antibiotics because the data would reflect the dose for payment rather than the administered dose based on body weight. Second, we calculated median PDDs (mg/day and mg/kg/day) and interquartile ranges for frequently prescribed antibiotics, to deal with outliers. Third, we standardized the antibiotic use in each hospital using the median PDDs (mg/day and mg/kg/day) stratified by age group, as follows: (1) PDD per 1,000 patient days based on the median PDD (mg/day) and (2) PDD per 1,000 patient days based on the median PDD (mg/kg/day). In the latter metric, PDD per 1,000 patient days standardized using PDD (mg/day) was calculated as the median PDD (mg/kg/day) × body weight (kg); thus, both metrics of PDD per 1,000 patient days were unified to compare them with each other. In addition, we showed the differences between the 2 metrics, since PDD per 1,000 patient days by median PDD (mg/day) may not reflect the dosing variation associated with body weight, compared with PDD per 1,000 patient days by median PDD (mg/kg/day). Fourth, we compared hospital-level metrics, estimated as PDD per 1,000 patient days by median PDD (mg/kg/day) and DOT per 1,000 patient days, from hospitals that showed >30 days of therapy for each antibiotic. In addition, we confirmed the relationship between the PDD- and DOT-based methods using a scatter plot and Spearman correlation coefficients. Statistical significance was set at P < .05.
In this study, we evaluated several antibiotics that were highly ranked according to the first step. We excluded the data of doses >10 times the median PDD for each antibiotic when calculating the PDD per 1,000 patient days based on the median PDD (mg/kg/day) and DOT per 1,000 patient days. All analyses were performed using R version 4.0.5 software (The R Foundation for Statistical Computing, Vienna, Austria).
Results
Among 465,407 patients and 579,417 admissions between April 2018 and March 2019, we included 453,001 patients and 564,326 admissions. In total, 12,406 patients (2.7%) and 1,5091admissions (2.6%) were excluded by the body weight outlier or mission data. Table 1 shows the characteristics of admitted patients in each age group, and the mean age and mean weight were 5 years (SD, 4) and 19 kg (SD, 14), respectively. The most common MDCs in the 1 month to <1 year group and the 1–6 years group were MDC04 (diseases and disorders of the respiratory system, mainly lower respiratory infections, influenza, and pneumonia; 40%) and MDC04 (diseases and disorders of the respiratory system, such as respiratory infections and asthma; 28%), respectively. In the 7–12 and 13–15 years old groups, the common diseases were MDC06 (diseases and disorders of the digestive system, hepatobiliary system, and pancreas, especially appendicitis, viral, and other specified intestinal infections, and inguinal hernia; 16%) and MDC16 (trauma, burns, and poisoning, including any type of fracture and injury; 22%), respectively. For the 86,389 patients (19.1%) that were prescribed oral antibiotics, the commonly prescribed oral antibiotics ranked as follows: amoxicillin (21.8%), clarithromycin (16.8%), cefditoren (12.7%), cefcapene (10.2%), cefaclor (9.4%), cefdinir (6.9%), sulfamethoxazole and trimethoprim (6.7%), amoxicillin-clavulanic acid (4.7%), and azithromycin (4.4%) (Supplementary Table 1). We assessed the PDD for the following 8 antibiotics: amoxicillin, amoxicillin-clavulanic acid, cefalexin, cefaclor, cefditoren, clarithromycin, tosufloxacin, and sulfamethoxazole and trimethoprim. The median PDDs (mg/day) were increased in the 1 month to <1 year group and the 7–12 years group, and it did not increase in the 7–12 years group or the 13–15 years group for all antibiotics (Table 2). Regarding PDDs (mg/kg/day), there were no significant differences between the 1 month to <1 year group and the 1–6 years group; however, they were gradually decreased in the 7–12 years group. After the use of antibiotic was standardized using the median PDD (mg/day) and median PDD (mg/kg/day), no differences were observed between PDD per 1,000 patient days based on the median PDD (mg/day) and PDD per 1,000 patient days based on the median PDD (mg/kg/day) (Table 3). The correlation coefficients were high for most antibiotics; however, the relationship between PDD per 1,000 patient days based on the median PDD (mg/kg/day) and DOT per 1,000 patient days differed between antibiotics (Fig. 1 and Supplementary Material).
Table 1. Admission Characteristics for Pediatric Patients in Each Age Category
Note. IQR, interquartile range; LOS, length of stay; MDC, major diagnosis category; SD, standard deviation.
a The age presented as days in 1 month to <1 year, others presented as years.
Table 2. Prescribed Daily Dose (PDD, mg/day and mg/kg/day) in 4 Age Groups
Note. IQR, interquartile range.
a Amoxicillin-clavulanic acid presented as the amoxicillin dose.
b Sulfamethoxazole and trimethoprim presented as the trimethoprim dose.
Table 3. Comparison of Prescribed Daily Dose (PDD) per 1,000 Patient Days Based on Median PDD (mg/day and mg/kg/day)
a The groups included >30 days of therapy per age group.
Fig. 1. The relationship between DOT per 1,000 patient days and PDD per 1,000 patient days per hospital for amoxicillin and amoxicillin-clavulanic acid. The solid line represents a slope of 1; the upper dot line represents slope of 2, and the upper dot line represents slope of 0.5. PDD per 1,000 patient days based on the median PDD (mg/kg/day).
Discussion
In this study, we developed a PDD-based methodology to assess the use of antibiotics in children. We showed median PDDs (mg/day and mg/kg/day) and the relationship between DOT per 1,000 patient days and PDD per 1,000 patient days for pediatric patients using the nationwide inpatient database in Japan. In addition, we characterized the relationship between PDD (mg/day) and DOT metrics and used scatter plots to describe the variation between the hospitals. We found that PDD (mg/day) metrics-stratified age groups could be an alternative to DDD-based metrics in the pediatric population.
Our results for oral antibiotic prescriptions for pediatric inpatients showed a similar tendency to that of pediatric outpatients in Japan. Reference Yamasaki, Tanabe, Muraki, Kato, Ohmagari and Yagi22–Reference Uda, Okubo, Kinoshita, Morisaki, Kasai, Horikoshi and Miyairi24 These studies also reported that penicillin with an extended-spectrum (anatomical therapeutic chemical code: J01CA), third-generation cephalosporins (J01DD), and macrolides (J01FA) were commonly prescribed for several years. In addition to the prevalence of antibiotics, we calculated PDDs for each antibiotic. The median PDD (mg/day) increased gradually with increasing age, reaching a plateau at 7–12 years for each antibiotic. In contrast, PDDs (mg/kg/day) decreased for each antibiotic because the patients were more likely to reach an adult body weight; thus, the median PDDs (mg/day) gradually decreased with increasing age.
For amoxicillin and amoxicillin-clavulanic acid, the PDDs showed a significant difference in the amoxicillin dose (Table 2). The PDDs of amoxicillin-clavulanic acid were more than double that of the PDDs of amoxicillin in all groups. These doses coincide with the Japanese guidelines for respiratory disease and labeled doses in Japan, and these variations may reflect the indications for each antibiotic. Reference Mikasa, Aoki and Aoki31 In our study, we indicated the PDDs as an amoxicillin dose instead of as a unit dose because the ratios of amoxicillin-clavulanic acid differed depending on the product used Reference Huttner, Bielicki, Clements, Frimodt-Møller, Muller, Paccaud and Mouton32 ; some are at a ratio of 2:1 and others are 14:1. In addition, clinicians occasionally prescribe amoxicillin-clavulanic acid products with amoxicillin products to adjust the ratio. Porta et al Reference Porta, Hsia and Doerholt18 reported that the PDDs (mg/kg/day) of amoxicillin plus clavulanic acid varied widely between centers, even though it was commonly prescribed in each hospital. These variations are explained mainly by differences in the disease or indications, which characterize the use of antibiotics.
The relationship between PDD per 1,000 patient days and DOT per 1,000 patient days varied with the antibiotics. For example, for tosufloxacin, the 2 metrics were almost identical (Supplementary Fig. 1). The study also reported a similar tendency for the difference between DDD per 1,000 patient days and DOT per 1,000 patient days for levofloxacin. Reference Polk, Fox, Mahoney, Letcavage and MacDougall9 In contrast, clarithromycin and sulfamethoxazole and trimethoprim were widely distributed. The results were associated with the wide range of doses. Sulfamethoxazole and trimethoprim are used to prevent urinary tract infections in children with vesicoureteral reflux or Pneumocystis pneumonia in children with immune suppression. Reference Wang, Gbadegesin, Foreman, Nagaraj, Wigfall, Wiener and Routh33,Reference Stern, Green, Paul, Vidal and Leibovici34 Macrolides also have low-dose treatment options for chronic airway diseases. Reference Sun and Long-term35 Thus, PDD-based metrics are more likely to be beneficial than DOT-based metrics for assessing the consumption of antibiotics when the range of doses is wide due to indications. The results imply that the PDD and DOT methodologies have different characteristics as metrics, and combining the 2 metrics was beneficial to audit child antibiotic use.
We used the median PDDs (mg/day) with stratification by age instead of the median PDDs (mg/kg/day) with body weight to adjust the dosing variation due to body weight. The stratification by age were valuable to control variations because the differences between the 2 metrics of PDD per 1,000 patient days may not be significant. Using the median PDDs (mg/day) in the stratified cohort is acceptable, given that DDDs are defined by the dose in 70-kg adults. Furthermore, there is a benefit to using the PDDs (mg/day) stratified by age because there are limitations to obtaining information on patient body weight in Japan. The Japanese claims database, which is mainly used to evaluate oral antibiotics in an outpatient setting, does not generally contain body-weight data, but it includes age. Thus, using PDD (mg/day) after stratifying by age would be one of the most practical approaches to assess the consumption of antibiotics rather than PDD (mg/kg/day).
The Manual of Antimicrobial Stewardship and changes in the reimbursement system regarding the regulation of antimicrobial stewardship fees after the national action plan for antimicrobial resistance were adopted in Japan. 21,36 Studies have reported that after the new antimicrobial stewardship fee was implemented, the frequency of antibiotic prescriptions decreased. Reference Muraki, Kusama and Tanabe37–Reference Kimura, Fukuda and Hayakawa39 These changes were made to promote appropriate antimicrobial therapy and affect decision making regarding the necessity of treatment, antibiotic selection, treatment period, and dose. Monitoring PDDs and the use of PDD benchmarking allows us to assess the effects in detail.
This study has notable strengths. It is the largest reported study on this subject based on a national administrative database. In addition, we have developed metrics for the use of antibiotics in Japanese pediatric settings, and we have assessed hospital variations for PDD- and DOT-based metrics. Our results indicate that the use of PDD- and DOT-based metrics, instead of DDD-based metrics, is helpful in pediatric settings and informs future interventions to improve the use of antibiotics in children. The metrics can also be used for benchmarking hospitals and further longitudinal analyses based on the database. Moreover, DDD-based metrics are not suitable for patients with renal dysfunction and children. Reference Polk, Fox, Mahoney, Letcavage and MacDougall9 Therefore, the potential application of PDD-based metrics should be examined by utilizing more detailed data, such as electronic health records, in further studies including patients and populations that reported large deviations between PDD and DDD. Reference Nunes19
This study had several limitations. First, these metrics cannot be applied to children in other countries because the PDDs indicate the dose prescribed for pediatric patients in Japan. In addition, our findings may not be generalizable to other populations, although we used the nationwide database in Japan. We could not assess the validity in outpatients due to the use of inpatient data. Second, we excluded admissions with an outlier or missing values of body weight due to data-entry errors. However, median PDD (mg/day) and PDD (mg/kg/day) values could be reasonable under the approved dosage and indications for infectious disease in Japan. Third, we could not show the PDD for parenteral antibiotics because the data more likely reflected the dose for payments and not the dose received by patients. Thus, these data were not suitable for PDD, and we alternative methods or data are needed to assess the dose of parenteral antibiotics in addition to DOT metrics in pediatric patients.
In conclusion, we assessed PDD (mg/day and mg/kg/day) and oral antibiotic metrics among children admitted to hospitals in Japan. The PDDs (mg/day) and PDDs (mg/kg/day) changed gradually, increasing with age. In addition, the relationship between the PDD per 1,000 patient days and DOT per 1,000 patient days was antibiotic dependent. Therefore, PDD could be useful for auditing intra- and interhospital changes and differences in antibiotic usage in Japan.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ash.2023.2.
Acknowledgments
Financial support
This study was supported by a Grant-in-Aid for Research on Policy Planning and Evaluation from the Ministry of Health, Labour and Welfare, Japan (grant no. 22AA2003 to K.F.) and a grant-in-aid for scientific research from the Japan Society for the Promotion of Science (grant no. JSPS KAKENHI, 20H03921 to D.S.).
Conflicts of interest
K.S. has received lecture fees from Sumitomo Dainippon Pharma, Astellas, Mitsubishi Tanabe Pharma, Gilead, BioMérieux Japan Ltd, and Abbie GK. All other authors have no conflicts to declare.
Open access
