Changes in Antimicrobial Usage Patterns in Korea: 12-Year Analysis Based on Database of the National Health Insurance Service-National Sample Cohort

National antimicrobial usage and prescription patterns during the 12 years from 2002 to 2013 were analyzed using the National Health Insurance Service-National Sample Cohort. Antimicrobial usage was analyzed by major illness, sex, age, area of residence, income rank, diagnosis, and type of medical institution for each year. Total antimicrobial prescriptions increased from 15.943 daily defined dose (DDD)/1,000 inhabitants/day in 2002 to 24.219 in 2013. In 2013, 72% of total prescriptions were administered in clinics. Antimicrobials were most frequently prescribed to children younger than 10 years, followed by adults aged 70 years or older and those aged 60–69 years. Penicillins and cephems were the most popular classes of antimicrobial used. In 2013, 48% of total antibiotic usage (11.683 DDD/1,000 inhabitants/day) was due to respiratory diseases. After the Korean government has implemented a series of healthcare policies, antibiotic prescription decreased for the treatment of upper respiratory infection, the causative agents are mostly viruses.

This describes the number of people (per a population of 1,000) who use an antimicrobial every day. As an example, 10 DDD per 1,000 inhabitants per day means that 1% of the population, on average, might receive a certain drug or group of drugs daily. Antimicrobial usage was analyzed by sex, age, area of residence, income rank, diagnosis, and type of medical institution for each year.
Statistical analysis. Trend tests were performed to determine potential differences in antimicrobial usage by year, and linear mixed models using group, time, and group-by-time as fixed effects were used to identify potential differences in the yearly patterns of antimicrobial usage in relation to general characteristics. Data were organized and analyzed using SAS v9.4 software (SAS Institute Inc., Cary, NC, USA).
Regarding the types of medical institutions, the vast majority of antimicrobials were prescribed in clinics, followed by general hospitals, hospitals, and long-term care hospitals (Fig. 2). In 2013, 72% of total prescriptions occurred in clinics (Table 2). Among clinics, penicillins (6.378 DDD/1,000 inhabitants/day) were the most frequently prescribed antimicrobial, followed by cephems (4.796), macrolides (2.639), and fluoroquinolones (1.536). In general hospitals, cephems (1.443) were the most frequently prescribed antimicrobial, followed by fluoroquinolones (0.767), penicillins (0.635), and macrolides (0.488) ( Table 2). Antimicrobial prescriptions of all classes, except for that of penicillin in clinics (P = 0.111), increased every year regardless of medical institution ( Antimicrobial prescriptions by sociodemographic features. No difference was observed in the use of antimicrobials according to gender (Fig. 3, P = 0.54), with antimicrobial prescription increasing every year in both males and females (Fig. 3, P < 0.001). The use of antimicrobials varied with age group (Fig. 4, P < 0.001). Antimicrobials were most frequently prescribed to children younger than 10 years, followed by adults aged 70 years or older and those aged 60-69 years (Fig. 4). In 2013, penicillins were the most frequently prescribed antimicrobial in children younger than 10 years, followed by cephems and macrolides, whereas cephems were the most frequently prescribed antimicrobial in adults aged 70 years or older, followed by fluoroquinolones, penicillins, and macrolides (data not shown).
When antimicrobial prescription patterns were analyzed by stratifying the income levels of the sample into five groups (from the upper 20th percentile to the lower 20th percentile), no difference was found according to income level (Fig. 5, P = 0.65), with antimicrobial prescriptions increasing every year in all five income groups ( Fig. 5, P < 0.001). Areas of residence were classified into 17 cities and provinces using the subjects' addresses on resident registrations to examine differences in antimicrobial prescription patterns across regions. Antimicrobial prescriptions increased in all cities and provinces (P < 0.001), without regional differences (P = 0.47, data not shown). 7 Table 1. Antimicrobial prescription according to class of antimicrobials (DDD/1,000 inhabitants/days). All classes of antimicrobial were included in total amount. (% of total DDD). Antimicrobial prescriptions by major illness. For major illness prescribed for antimicrobials, more than half of all antimicrobial agents used were due to respiratory diseases (Fig. 6). Antimicrobial prescriptions for respiratory diseases increased every year (Table 3, P < 0.001). In 2013, 48% (11.683 DDD/1,000 inhabitants/day) of total antibiotic usage was due to respiratory diseases. Although not statistically significant, antimicrobial prescriptions for URIs decreased from 4.9 DDD/1,000 inhabitants/day in 2002 to 3.625 in 2013 (P = 0.109, Table 3). Antimicrobial prescriptions for genitourinary diseases, including urinary tract infections, increased every year ( Table 3, P < 0.001). By class, penicillins were the most frequently prescribed antimicrobial for respiratory diseases (5.281 DDD/1,000 inhabitants/day) in 2013, followed by cephems (3.113), macrolides (2.307), and fluoroquinolones (0.697), and similar trend was observed for URIs. For genitourinary diseases, fluoroquinolones were the most frequently prescribed antibiotics (0.796), followed by cephems (0.488), and a similar trend was observed for urinary tract infections.
When antibiotic use was compared between 2013 and 2002, cephem, macrolide and fluoroquinolone use increased, but penicillin use was not increased for respiratory diseases. For genitourinary diseases, prescription of cephem, macrolide, penicillin and fluoroquinolone increased (see Table 3).

Discussion
This study explored the changes of antimicrobial usage in Korea from 2002 to 2013 through analyzing the NHIS-NSC data using DDD developed by the WHO for drug statistics methodology.
Antimicrobial resistance in human pathogens shows substantial geographic differences across the globe. Although antimicrobial resistance remains low in Northern European countries, it is high in Southern European, Latin American and Asian countries. In Asia, extended spectrum β-lactamase (ESBL)-producing Enterobacteriaceae are of concern and are increasing. In 2009 and 2010, 28 percent of Enterobacteriaceae isolated from the patients with urinary tract infections in 11 countries were ESBL producers 18 . In China, in 2011, ESBL-producing E. coli accounted for 71 percent of E. coli isolates, and more than half of K. pneumoniae strains produced ESBL 19 . Infections with carbapenem-resistant Enterobacteriaceae are increasingly reported from healthcare facilities, not only in developed countries 20 , but also in low-and middle income countries. Treatment for carbapenem-resistant Enterobacteriaceae is much more complicated than ESBL-producing stains because of limited treatment options.
Between 2000 and 2010, global consumption of antibiotic drugs increased by 36% 21 . There is a high variability of antimicrobial consumption across countries. Among the Organisation for Economic Co-operation and Development (OECD) countries, Chile and Netherlands reported the lowest volume (9.4 and 11.4 DDD/1,000 inhabitants/day, respectively) and Greece and Italy reported 30.1 and 34.9 DDD/1,000 inhabitants/day, respectively in 2013 22 . Antimicrobial use in Japan was 14.7 in 2009 and 15.8 DDD/1,000 inhabitants/day 23 . Korea remains a country with heavy antimicrobial usage 16,24 ; its 2013 national average (28.4 DDD/1,000 inhabitants/ day) exceeds the OECD average (20.4 DDD/1,000 inhabitants/day) 22 . New systems were introduced in Korea upon the turn of the 21 st century to increase the quality of medicine and reduce the use of antimicrobials 25 Table 2. Antimicrobial prescription according to medical institution (DDD/1,000 inhabitants/days). All classes of antimicrobial were included in total amount. (% of subtotal DDD or % of total DDD).    after the separation of dispensing and prescribing drugs using health insurance found that usage was 23.62 in 2004 11 . A previous study, ' Annual Products of Medicine in 2008′, analyzed the cost and amounts of antibiotics produced and compared them with previous data. The cost and amounts of antibiotics are indirect parameters of drug utilization, and that study did not consider imported antibiotics. In this study, the DDD was used because it facilitates the presentation and comparison of drug consumption statistics at international, national and regional levels. Although drug consumption can be measured in cost, number of units, number of prescriptions or the physical quantity of drugs, these variables can vary between regions and countries over time. Because of such limitation in comparisons of drug consumption at an international level, a technical unit of measurement, DDD, is recommended by the WHO to measure drug utilization in countries 29 . Moreover, the NHIS-NSC data used in this study were claim data that could be used to estimate nationwide antibiotics usage directly 14 .
This study found that antimicrobial usage increased, from 15.943 DDD/1,000 inhabitants/day in 2002 to 24.219 in 2013. Analysis of usage by year showed a sharp increase from 2002 to 2010, but had been consistent since 2010. Since antimicrobial usage is increasing despite the enforcement of relevant policies, it can be suspected that these policies are not effective enough, or that the growth of the elderly population has increased antimicrobial usage. However, the rate at which antibiotics use increases is declining in recent years, and it seems that the antibiotic policy is beginning to show its effects.
URI, a disease that encompasses simple colds, allergic rhinitis, middle ear infections, and pharyngitis, is viral in most cases, and antimicrobials are rarely required, suggesting that antimicrobial usage for this particular disease should be managed closely 30 . Although changing prescribing behaviors can be difficult, guidelines exist for acute rhinosinusitis 31 , acute uncomplicated bronchitis 32 , and pharyngitis 33 , and proven evidence-based methods can be used to optimize antibiotic therapy in outpatient settings [34][35][36] . Penicillin 0.080 (14) 0.093 (14) 0.089 (13) 0.086 (12) 0.084 (12) 0.076 (9) 0.072 (7) 0.064 (7) 0.065 (6) 0.065 (6) 0.063  Table 3. Antimicrobial prescription according to illness (DDD/1,000 inhabitants/days). All classes of antimicrobial were included in total amount. (% of subtotal DDD or % of total DDD). Over the last decade the Korean government has implemented a series of healthcare policies directed to URI consisting of drug utilization review in the outpatient service in 2001 and public reporting of the rate of antibiotics use for URI in 2006. As a result of government policy for URI, antimicrobial usage for URIs tended to decline, though the trend was not statistically significant, from 4.9 DDD/1,000 inhabitants/day in 2002 to 3.625 in 2013. The proportion of total antimicrobial usage for acute URIs decreased from 30% in 2002 to 14% in 2013. In particular, antimicrobials prescribed for URI decreased after public reporting of the rate of antibiotics use for URI in 2006 (4.982 in 2005 and 4.000 in 2006).
Perception change of primary physician and a public campaign for 'appropriate antibiotic use' could contribute to decrease of antimicrobial usage. Studies using clinical scenarios involving the common cold or standard patients showed that physicians' responses in favor of antibiotic prescription in acute URTIs fell from 75.0% in 1991 to 54.7% in 2003 and to 27.2% in 2010 although the clinical cases and study subjects differed between the studies [37][38][39][40] . A public campaign for 'appropriate antibiotic use' organized by the Korean Society of Infectious Diseases and the Korean Society for Chemotherapy, and sponsored by the Korea CDC, was launched in Korea in 2011 41 .
In 2008, the use of antimicrobials increased significantly. After the implementation of public reporting in 2006, diagnostic coding which has not been designated for antibiotic use evaluation were increased significantly. For example, the most common respiratory tract infections was J03 acute tonsillitis in 2005 but changed to J20 acute bronchitis in 2009. Increased use of antimicrobials in 2008 may be explained in part by diagnostic shift in the outpatient setting although the cause of increasing trends in total antimicrobial usage needs to be determined 27 .
The use of aminoglycosides decreased from 1.494 in 2002 (10% of total antimicrobial usage) to 0.807 in 2013 (3% of total antimicrobial usage). Although amoxicillin was the most popular penicillin in 2002, amoxicillin-clavulanate was more popular in 2013, suggesting a preference for combined drugs to overcome bacterial resistance. Whereas the use of penicillins did not increase, the use of cephems increased more than two-fold over our study period. Cephem use was only half of that of penicillins in 2002, but prescriptions increased to a level comparable with that of penicillins in 2013. The most frequently used cephem was cefaclor; its usage in 2013 was 3.222 DDD/1000 inhabitants/day, which accounted for half of the total cephem usage (7.211). The fact that the use of cefaclor, a second-generation cephem, exceeded that of oral third-generation cephems such as cefixime and cefpodoxime (though their use increased as well) could be related to insurance policies that do not cover primary treatment with third-generation cephems. Furthermore, the side effects (e.g., diarrhea) of amoxicillin-clavulanate, which is used to overcome resistance, could contribute to the preference for cefaclor. Macrolide and fluoroquinolone prescriptions also increased, and prescriptions for clarithromycin, levofloxacin, and ciprofloxacin increased by particularly high degrees. Use of sulfamethoxazole-trimethoprim, a drug typically used to treat urinary tract infections, declined, possibly due to concerns of resistance of Escherichia coli-the most common pathogen in urinary tract infections.
Regarding age, antimicrobials were most heavily used in children under the age of 10, followed by adults aged 70 years or older and those aged 60-69 years. Penicillins were the most popular antimicrobial among children. Fluoroquinolones are not used in children due to the potential side effects. Cephems were preferred over penicillins among adults aged 20 years or older, and cephems and fluoroquinolones were preferred over penicillins among adults aged 70 years or older in 2013.
In this study analyzed all antimicrobial used, there was no difference was found according to income level. Recent study for antimicrobial used for acute URIs found that income level was negatively correlated with prescription rate and suggested demand factors as well as supply factors contributed antibiotic overuse 15 . We hypothesize that demand factors have a greater impact on mild illnesses such as acute URIs.
In 2013, 72% of total prescriptions occurred in clinics. This is attributable to the fact that a greater number of patients are treated in clinics than in other types of medical institutions. Thus, antimicrobial usage in clinics should be closely monitored.
In conclusion, antimicrobial usage increased in Korea from 2002 to 2013, with 70% of the prescriptions made in clinics, and penicillins and cephems being the most popular classes of antimicrobials. In 2013, 48% of all antimicrobials were prescribed for respiratory diseases, and antimicrobial prescriptions for URIs are declining. On the cusp of a post-antibiotics era, monitoring antimicrobial use can be an important strategy to prolong the activities of existing antimicrobials. Data availability. The datasets generated during the current study are available from the corresponding author on reasonable request.