Background: Appropriate antibiotic use has become an important issue. However, collecting data on the use of all antibiotics in a hospital is difficult without an advanced computerized system and dedicated staff. This paper examines if 1–3 antibiotics can satisfactorily represent the total antibiotic consumption at the hospital level. Methods: We collected antibiotic data from six large university hospitals in Korea for some years between 2004 and 2012. Since the total antibiotics consist of a few chosen representative antibiotics and the rest, we used those 2 3 chosen antibiotics along with additional variables constructed only with t (time) such as t, t , and t to capture the time trend and whether t belongs to each month or not to capture the monthly variations. The ordinary least squares method was used to explain the total antibiotic amount with these variables, and then the estimated model was employed to predict the use for 2013. To determine which antibiotics were the most representative in tracking general trends in antibiotic use over time, we tried various combinations of antibiotics to find the combination that best minimized the 2013 prediction error. Results: We found that fluoroquinolones and aminoglycosides were the most representative, followed by beta- lactam/beta-lactamase inhibitors and 4th-generation and 3rd-generation cephalosporins. The mean prediction error over 12 months in 2013 with these few antibiotics was only 1–3% of the monthly antibiotic consumption amount. Conclusions: The total antibiotic consumption amount at the hospital level can be represented sufficiently by a few antibiotics, such as fluoroquinolones and aminoglycosides, which means that hospitals can save resources by tracing only the usage of those few antibiotics instead of the entire inventory. Since the choice of fluoroquinolones and aminoglycosides is based solely on our Korean data, other hospitals may follow the same modelling methodology to find their own representative antibiotics. Keywords: Antibiotic consumption, antimicrobial stewardship, statistical model, fluoroquinolone, aminoglycoside Background As is well documented, the overuse/misuse of antibi- Antimicrobial resistance is a worldwide problem, which otics has been recognized as a key factor for the poses a serious threat to global public health . In 2011, emergence of antimicrobial-resistant organisms . In- to prevent further worsening of the problem, the World appropriate antibiotic use also causes extra medical ex- Health Organization (WHO) urged nations to be alert for penses: unnecessary or duplicative antibiotic use in US antimicrobial resistance and called for urgent action to de- hospitals led to an estimated $163 million in excess costs crease antimicrobial consumption . In accordance with . Hence, many experts have suggested establishing the initiative, the Korean government launched a national antimicrobial stewardship programmes in hospitals as action plan on antimicrobial resistance in 2016 . well as in communities . The first step to combat antibiotic abuse is finding out the severity of the prob- * Correspondence: firstname.lastname@example.org; email@example.com; lem. This calls for a proper measurement of antibiotic firstname.lastname@example.org † consumption , which helps to understand the epi- Bongyoung Kim and Hyeonjun Hwang contributed equally to this work. Department of Economics, College of Political Science & Economics, Korea demiology of antimicrobial resistance and provides hos- University, 145 Anam-ro, Sungbuk-gu, Seoul 02841, South Korea pitals with useful data to implement policies and Department of Internal medicine, College of Medicine, Hanyang University, guidelines about proper antibiotic usage . 222-1 Wangsimni-ro, Seongdong-gu, Seoul 04763, South Korea Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Kim et al. BMC Infectious Diseases (2018) 18:247 Page 2 of 9 To this end, we collected data on antibiotic prescrip- All data were extracted from the electronic billing sys- tions from six large Korean university hospitals with good tem by the data processing department in each hospital. computerized systems. There were considerable difficul- ties in collecting data on antibiotics because there were Definitions too many different types of antibiotics but no experi- We define antibiotics as medications with class J01 in enced/dedicated staff to collect data on antibiotics. There Anatomical Therapeutic Chemical (ATC), which does was no problem obtaining data on the prescription depart- not include antifungal agents nor anti-tuberculosis ment, administration route and volume of drug, but a big agents. Systemic agents with oral or parenteral adminis- problem in the data collection was that antibiotics were tration routes are included, but topical agents are ex- recorded by the brand name, not by the ingredient names cluded. We convert each class of antibiotic amount to a nor by the antibiotic class name. This required extra effort defined daily dose (DDD) by using the ATC of the to convert the data into a suitable and consistent form. WHO and then standardize for 1000 patient days . The goal of this paper is to explore whether it is pos- We classify antibiotic agents into 19 classes: 1st- sible to look at only a couple of representative antibiotics generation cephalosporins (1st CEP), 2nd-generation to determine the total antibiotic consumption at the cephalosporins (2nd CEP), 3rd-generation cephalosporins hospital level. If yes, this means a considerable savings in (3rd CEP), 4th-generation cephalosporins (4th CEP), ami- terms of time and effort to keep track of all antibiotic noglycosides (AG), beta-lactam/beta-lactamase inhibitors use. To this goal, we built a statistical model, in which (BL-BLI), carbapenems, fluoroquinolones (FQ), glycopep- 1–3 representative antibiotics are chosen to predict the tides, lincosamide, macrolides, metronidazole, monobac- total antibiotic consumption at the hospital level, with tam, oxazolidinone, penicillins, polymyxin, tetracyclines, an acceptably small magnitude of prediction error. tigecycline and trimethoprim/sulfamethoxazole. Other an- tibiotics such as amphenicol, fosfomycin, and streptogra- min are excluded because they are rarely used. Methods Let n denote the patient days for hospital h =1 … 6 ht Study design and month t; t ranges over 1 … 12 (year 2004), 49 … 60 We build a simple linear statistical model, where the (year 2008) and 97 … 108 (year 2012) for the hospitals total antibiotic consumption at the hospital level is ex- other than HUS, and t ranges over t =1 … 108 for HUS. plained by 1–3 representative antibiotics along with time Let DDD denote the DDD for antibiotics a, hospital h, aht and month dummy variables–the time and month and time t. With ‘≡’ standing for “defined as”, let ‘all- dummy variables are “free”, as they depend on time hospital patient days at time t’ and ‘all-hospital DDD for index t only. Because the total amount consists of the antibiotics a at time t’ be representative antibiotics and the remaining (non-repre- sentative) ones, this modelling strategy amounts to n ≡ n þ … þ n and DDD ≡ DDD þ … þ DDD t 1t 6t at a1t a6t explaining the remaining antibiotics using their correla- tions with the representative antibiotics as well as the Then, the all-hospital DDD per 1000 patient days for time and month dummy variables. We estimate the antibiotics a at month t is model using the observations over 2004–2012 of six DDD at large university hospitals in Korea, one of which is the X ≡ 1000: at Hanyang University Seoul Hospital (HUS). Then, the t model prediction capability is evaluated for the ensuing Let m (‘m’ for main) be the number of the main (i.e., year (2013), using the data from HUS. representative) antibiotics; m = 1, 2 or 3 in this paper. Listing the main antibiotics first, the total antibiotic Data source amount at time t can be written as We collected data on the total antibiotic prescriptions Y ≡ the main antibiotic amount ð1Þ for inpatients and their total patient days in 2004, 2008 þ the others and 2012 from six university hospitals (4 tertiary and 2 m 19 X X secondary) in Korea: Hanyang University Seoul Hospital ¼ X þ X at at (758 beds), Chungbuk University Hospital (620 beds), a¼1 a¼mþ1 Chonnam University Hospital (970 beds), Gyeongsang University Hospital (889 beds), Hanyang University Guri Statistical Methodology Hospital (578 beds), and Korea University Ansan To achieve our goal of representing the total Y with the Hospital (543 beds). In addition, we collected data from t main antibiotics, it is necessary to account for the HUS for each year between 2004 and 2013 on the total antibiotic prescription records and the total patient days. remaining part X in (1) in a simple way. We at a¼mþ1 Kim et al. BMC Infectious Diseases (2018) 18:247 Page 3 of 9 achieved this by replacing the sum X with at a¼mþ1 “free variables”. If the free variables can represent Y −Y j ð2Þ τ τ τ¼109 X well enough, then we do not have to collect at a¼mþ1 data on those remaining antibiotics. In other words, (2) is the average of the monthly abso- We used three types of free variables to account for P lute deviations for 2013 between the actual and pre- X : (i) time index t to capture the trend, (ii) at a¼mþ1 dicted antibiotic uses in DDD/1000 patient days. The month dummies to capture the monthly variations, particular combination of two antibiotics minimizing (2) (iii) and some calendar time dummies to capture is the best choice. “structural breaks” (i.e., big events), if there are any. To explain why we consider different values for m, the Since all three types are determined by t,nodatacol- reason is that there is a trade-off in setting m large v. lection is needed for them. We illustrate these three small. If m is large, say 10, we can trace the overall anti- types next. Let 1[A]=1 if A holds, and 0 otherwise. biotic consumption better, but then the representative- Suppose we have t =1 … 17 monthly observations over ness will be worse; if m is small, say 1, then the opposite January 2004 to May 2005. First, use a polynomial func- happens. Between these extremes, 1–3 seem to be rea- P P p 2 q q tion such as α t (e.g., α t ¼ α þ α t q q 0 1 q¼0 q¼0 sonable values, and for each chosen value of m, we try þ α t ) to account for the trend, where α s are the pa- different combinations of antibiotics. rameters to be estimated using (1, t …t ). Second, The model for the ordinary least squares (OLS) esti- capture the monthly variations with the month mator, where the main antibiotic amount and the above dummies; e.g., the February dummy 1[t ∈ February]isto “free” t-based variables collectively explain Y , is shown capture the February effect relative to the baseline in Additional file 1: Tables S3 and S4; Additional file 1: January, where ‘∈’ means “belonging to”, and the March Table S3 uses all six hospitals’ data, whereas Additional dummy 1[t ∈ March] is to capture the March effect file 1: Table S4 uses only the HUS data for the model es- relative to January. Third, there might be a big policy timation. In each table, the OLS estimates and their change, say, a crackdown on antibiotic abuse at t = 6 and standard errors (SE) are provided. Dividing an estimate onwards, in which case 1[6 ≤ t] can be used to account by its SE gives the ‘t-value’ or ‘z-score’. It being above 2 for the crackdown that is a structural break. in absolute value indicates statistical significance at the Since the main antibiotics and t-based variables are 5% error level, i.e., we set statistical significance at P <0. in the model, whereas the other antibiotics are not, 05. R shows the proportion of the Y variation explained in essence, the omitted non-representative antibiotics by all “regressors” (i.e., explanatory variables) jointly. are explained by their correlations with the main anti- biotics and the t-based variables. After the model Results parameters are estimated using the time-series data Most prescribed antibiotics in the pooled data up to t = 108 (December 2012), we then construct the Pooling all time-series data of the six hospitals into one big predicted Y for t = 109~ 120 (2013) for HUS using data set, Table 1 provides descriptive statistics in all six hospi- b tals, including HUS, as well as HUS alone; the unit for all the estimated model; let Y denote the predicted numbers is DDD/1000 patient days. The average total anti- value. biotic consumption of the six hospitals plus/minus the stand- After model estimation using t = 1~ 108, predicting ard deviation (SD) is 864 ± 55.5, and that of HUS is 915 ± Y for t = 1~ 108 is “in-sample prediction”,and pre- 100. Overall, 3rd CEP was used most in all six hospitals (24. dicting Y for t = 109~ 120 is “out-sample prediction”. 8% from 213.82/862.94), followed by FQ (12.1%, 104.11/862. The out-sample prediction is to pick the representa- 94), 2nd CEP (11.4%, 98.17/862.94), 1st CEP (10.6%, 91.84/ tive antibiotics, and the in-sample prediction is just 862.94) and BL-BLI (10.5%, 90.27/862.94). Similarly, 3rd CEP to see how the chosen representative antibiotics (18.8%, 173.26/920.69) was used most in HUS, followed by perform in fitting the in-sample observations. Since 1st CEP (15.5%, 143.10/920.69), FQ (13.4%, 123.15/920.69), we put more emphasis on predicting the future than 2nd CEP (12.7%, 116.94/920.69), AG (9.8%, 90.57/920.69) on explaining the past, the out-sample prediction is and BL-BLI (8.0%, 74.11/920.69). In contrast, monobactam, our primary criterion to determine the representative oxazolidinoneand tigecyclinewererarelyused; therewas antibiotics, whereas the in-sample prediction is even no use at all of these antibiotics for some months. secondary. To explain how to select 1–3 main antibiotics, sup- Out-sample prediction, representative antibiotics, and pose m = 2. For each main antibiotic candidate, we in-sample fitness b b obtain Y ;:: Y for 2013 and its “mean prediction 109 120 Table 2 shows the main antibiotics minimizing the mean error”: prediction error (2). For example, the mean prediction Kim et al. BMC Infectious Diseases (2018) 18:247 Page 4 of 9 Table 1 Antibiotic consumption in all hospitals and in only Hanyang University Seoul hospital (unit: DDD/1,000 patient days) Six hospitals (2004, 2008, 2012 : 36 months) Hanyang University Seoul hospital (2004-2012 : 108 months) a a Mean (SD) Range Mean (SD) Range st 1 CEP 91.7 (8.8) 78.5-114.0 142.0 (23.4) 93.7-210.0 nd 2 CEP 98.4 (21.9) 60.8-138.0 115.0 (31.7) 77.9-204.0 rd 3 CEP 214.0 (15.4) 188.0-249.0 172.0 (19.6) 136.0-231.0 th 4 CEP 11.4 (5.8) 3.73-21.0 14.6 (9.7) 0-38.0 AG 71.2 (47.4) 19.5-146.0 86.6 (56.2) 18.3-184.0 BL-BLI 90.2 (7.4) 74.8-101.0 74.9 (15.8) 44.1-123.0 Carbapenems 14.7 (5.4) 6.4-24.6 9.9 (5.7) 0-22.1 FQ 104.0 (7.0) 90.3-121.0 123.0 (14.2) 88.6-172.0 Glycopeptides 23.3 (2.1) 18.3-27.2 18.7 (4.6) 9.3-32.4 Lincosamide 16.6 (3.8) 10.7-23.8 15.6 (4.9) 4.6-31.2 Macrolides 41.8 (8.1) 29.4-59.3 44.8 (14.6) 21.3-88.2 Metronidazole 41.0 (7.8) 30.6-61.9 43.6 (7.7) 21.3-61.2 Monobactam 0.6 (0.5) 0-2.0 0.8 (1.0) 0-4.5 Oxazolidinone 0.7 (0.6) 0-2.1 1.1 (1.1) 0-4.8 Penicillins 20.4 (6.4) 10.1-37.7 17.6 (7.5) 2.0-41.2 Polymyxin 2.8 (2.1) 0-6.3 2.3 (2.6) 0-11.6 Tetracyclines 9.7 (6.4) 3.5-30.9 4.7 (4.6) 0-21.7 Tigecycline 0.3 (0.5) 0-1.7 0.7 (1.2) 0-4.8 Trimethoprim/sulfamethoxazole 11.2 (1.9) 8.1-17.1 26.2 (10.1) 5.8-51.5 Total 864.0 (55.5) 766.0-975.0 915.0 (100.0) 770.0-1121.0 Monthly consumption averaged over the years st nd rd th Abbreviations: 1 CEP 1st-generation cephalosporins, 2 CEP 2nd-generation cephalosporins, 3 CEP 3rd-generation cephalosporins, 4 CEP 4th-generation cephalosporoins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors, FQ fluoroquinolones error is 26.2 DDD/1000 patient days using only AG for BL-BLI. When only the HUS data are used, FQ or BL- m = 1, and it is 17.2 using AG and 4th CEP for m =2, BLI does best, followed by AG and 3rd CEP. Combining where the predictors were obtained with all six hospitals’ these findings, we may state that FQ is the most repre- data. In contrast, the mean prediction error is 20.7 sentative, followed by AG, BL-BLI, 4th CEP and 3rd DDD/1000 patient days using only FQ for m = 1, and it CEP. Since the total number of observations is 180 is 18.3 using FQ and AG for m = 2, where the predictors (=36 months times 5 hospitals) plus 108 (12 months were obtained with only the HUS data. times 9 years from HUS) and HUS takes only 37.5% of In Table 2, when all six hospitals’ data are used in the the total observations 288 = 180 + 108, the result based left half, AG does best (with the mean prediction error on the entire data set that AG is the best changes when 26.2 when used alone), followed by FQ, 4th CEP, and only the HUS observations are used. The details on the Table 2 Representative antibiotics and mean prediction error (unit: DDD/1,000 patient days) Six hospitals Hanyang University Seoul hospital (HUS) # main antibiotics Representative antibiotics Prediction error Representative antibiotics Prediction error m = 1 AG 26.2 FQ 20.7 FQ 26.9 BL-BLI 22.0 th m =2 AG +4 CEP 17.2 AG + FQ 18.3 AG + penicillins 18.1 BL-BLI + FQ 18.7 th rd m = 3 AG + BL-BLI + 4 CEP 12.3 BL-BLI + FQ + 3 CEP 15.0 th rd AG + 4 CEP + monobactam 14.0 AG + BL-BLI + 3 CEP 15.4 rd th Abbreviations: 3 CEP 3rd-generation cephalosporins, 4 CEP 4th-generation cephalosporins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors, FQ fluoroquinolones Kim et al. BMC Infectious Diseases (2018) 18:247 Page 5 of 9 OLS used for prediction are provided in the Additional looks good; Fig. 4 does thesamefor HUS. Noticea file 1. large drop at t =52 in Fig. 4, which prompted using Define the “mean prediction error multiplied by 100 1[52 ≤ t] in the OLS for the HUS data. and divided by the monthly antibiotics consumption amount in Table 1” as the “relative mean prediction The most prescribed antibiotics are not necessarily the error”. Using the six hospitals’ data, the relative mean most representative prediction error for m = 1, 2, and 3 is, respectively, Taking the most prescribed antibiotics (FQ, 1st CEP and 3rd CEP) as the three representative antibiotics, we redrew 26:2 17:2 100 ¼ 3:0%; 100 Figs. 1, 2, 3 and 4 to present the result in Figs. 5, 6, 7. 864 864 Comparing Figs. 1 and 2 to 5, 3 to 6 and 4 to 7,it is clear 12:3 ¼ 2:0%; 100 ¼ 1:4% that the most prescribed antibiotics do not constitute the most representative antibiotics. Specifically, the mean pre- Using the HUS data, the relative mean prediction error diction errors with all six hospitals in Fig. 5 (solid line) for m=1,2,and 3is and with only HUS (double line) when the three most pre- scribed antibiotics are used are 41.7 and 33.0, respectively, 20:7 18:3 whereas the mean prediction errors with m = 3 are 100 ¼ 2:3%; 100 915 915 approximately 12–15 in Table 2. 15:0 ¼ 2:0%; 100 ¼ 1:6% Discussion Judging from the m = 2 and 3 cases here and the rows Collecting data on all antibiotics is a tedious and pains- for m = 2 and 3 in Table 2, although the predicted Y is taking task in Korea, and this may be the case in other for HUS, using all the hospital data is preferable to using countries as well. We showed that it is possible to use only the HUS data to minimize the prediction error. only 1–3 representative antibiotics to track the total Figure 1 shows the out-sample prediction time-series antibiotic consumption at the hospital level. Our main plot and the difference between the observed and pre- finding is that FQ and AG are the most representative, dicted values using all hospital data, and Fig. 2 shows followed by BL-BLI, 4th CEP and 3rd CEP. Our mean the same using only the HUS data. The figures show prediction error is only 1–3% of the monthly antibiotic that the predicted lines match the actual line (dotted) consumption amount, which is the average across the well, and the 95% confidence interval for the prediction hospitals and years in our data. Whether or not these error includes zero in almost all cases. levels of prediction error are tolerable depends on how Figure 3 presents the in-sample actual and fitted much we save in terms of time and money by not col- (m = 1,2,3) values for the six hospitals, and the fitness lecting data on the other antibiotics. Fig. 1 Out-sample prediction for Hanyang University Seoul Hospital, 2013, using six hospital data sets. Abbreviations: 4th CEP 4th-generation cephalosporins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors, CI confidence interval Kim et al. BMC Infectious Diseases (2018) 18:247 Page 6 of 9 Fig. 2 Out-sample prediction for Hanyang University Seoul Hospital, 2013, using Hanyang University Seoul Hospital data only. Abbreviations: 3rd CEP 3rd-generation cephalosporins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors, FQ fluoroquinolones, CI confidence interval Although our representative antibiotics were selected, for inpatients in Korea, followed by AG, 1st CEP and FQ not based on their medical effectiveness, but based on how . These studies suggest that we might have found almost well they collectively represented the total antibiotics usage, the same representative antibiotics had we analysed other the representative antibiotics also happened to be the most Korean hospitals’ data that are not in our data set. commonly prescribed for inpatients, except for 4th CEP. The antibiotic usage pattern is different at various To better monitor antibiotic consumption in hospitals, one levels. For instance, in Italy and the UK, AG are not of the broad-spectrum antibiotics or antibiotics against used as frequently as in Korea [10, 11], which illustrates multi-drug resistant pathogens (such as carbapenems) country-level differences; also, there are large differences could be co-monitored with our representative antibiotics. in the consumption profiles for treatments of the same The overall antibiotic usage patterns in our data differ lit- bacterial infection among European countries . Even tle from other studies in Korea. A single-centre study found among the hospitals in the same country, large differ- that 3rd CEP was the most commonly prescribed antibiotic ences in antibiotic usage patterns exist; e.g., medium- for hospitalized patients in Korea, followed by FQ, BL-BLI sized, private and university hospitals use more and 1st CEP . Additionally, a population-based study antibiotics ; additionally, antibiotic usage patterns showed that 3rd CEP was the most prescribed antibiotic differ between small and large community hospitals in Fig. 3 In-sample prediction (2004, 2008, 2012) for six hospitals, using six hospitals’ data. Abbreviations: 4th CEP 4th-generation cephalosporins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors Kim et al. BMC Infectious Diseases (2018) 18:247 Page 7 of 9 Fig. 4 In-sample prediction for Hanyang University Seoul Hospital, using Hanyang University Seoul Hospital data only. Abbreviations: 3rd CEP 3rd-generation cephalosporins, AG aminoglycosides, BL-BLI beta-lactam/beta-lactamase inhibitors, FQ fluoroquinolones Korea . Possible reasons for these differences are vari- in choosing antibiotics to consider first (it would be FQ, ations in bacterial epidemiology at hospital level, the AG, BL-BLI, 4th CEP and 3rd CEP); of course, the most medical staff’s attitude towards prescribing antibiotics, commonly prescribed antibiotics in the hospital would also antimicrobial stewardship programme effectiveness, etc. make good candidates. Hence, if possible, it would be ideal for each hospital to We attribute the structural break in Fig. 4 at HUS to the conduct a study of its own (as was done in this paper) to pre-authorization of an antibiotic use programme that find its own representative antibiotics. started in 2008. The programme put restrictions on pre- The methodology we presented used basic statistics for scribing broad-spectrum antibiotics such as carbapenems, predicting future time-series variables. It should not be too glycopeptides, oxazolidinone, polymyxin and tigecycline difficult for hospitals to tailor the methodology to meet by requiring an extra approval step from the infectious their needs, finding a few representative antibiotics by disease department . Additionally, the programme re- using, e.g., different functions of t and different structural inforced educating physicians on the appropriate use of breaks at different times. Once the methodology is set, the antibiotics and collecting feedback after drug use. hospital would then address the problem of selecting a few As the HUS time-series data plot illustrates in Fig. 4,a representative antibiotics, which is in fact more difficult structural break can move the intercept substantially, than it looks; e.g., if three are to be chosen out of 20 antibi- the ignorance of which would result in large biases in otics in total, there are 1140 possible combinations. In this the other estimates because the other estimates would case, despite many differences across countries and hospi- be adjusted downward to account for the large drop in tals within the same country, our findings should be helpful the intercept. Detecting structural breaks is relatively Fig. 5 Out-sample prediction for Hanyang University Seoul Hospital, 2013, with three most commonly used antibiotics (FQ, 3rd CEP, and 1st CEP). Abbreviations: FQ fluoroquinolones, 3rd CEP 3rd-generation cephalosporins, 1st CEP 1st-generation cephalosporins Kim et al. BMC Infectious Diseases (2018) 18:247 Page 8 of 9 Fig. 6 In-sample prediction (2004, 2008, 2012) for six hospitals with the most commonly used antibiotics (FQ, 3rd CEP, and 1st CEP). Abbreviations: FQ, fluoroquinolones 3rd CEP 3rd-generation cephalosporins, 1st CEP 1st-generation cephalosporins straightforward and can be accomplished by plotting the other hospital data or a longer time span of data may time-series data, as in Figs. 3 and 4. Of course, if the alter/qualify our findings. Third, we adopted a relatively break magnitudes are small, then they are hard to detect simple ordinary least squares estimator to find the time with the naked eye, but then they would not be called trend and monthly variations; more statistically sophisti- “breaks”. Structural breaks might have to be incorpo- cated models and approaches may refine and improve rated using outside information such as announced law/ the prediction capability. Finally, we measured antibiotic regulation changes. consumption by DDD instead of days of therapy (DOT). There are some notable limitations in our study. First, According to a recent guideline for antibiotic steward- the six university hospitals were selected, not by any ship programmes, DOT is preferred to DDD as a meas- sampling principle, but by ease in data collection, in ure of antibiotic consumption . However, we could which sense our data may not be representative of the not use DOT because only the total amount of antibiotic large university hospitals in Korea that would be our consumption per patient was available in five of the six study population of interest. For five hospitals, we could hospitals. gather only three years of data, which resulted in rela- As far as we are aware, our study is the first of its kind tively larger standard errors than we would have liked. to look at the possibility of using only a few antibiotics Second, the prediction performance was gauged using to track the total antibiotic consumption at the hospital only one hospital’s single-year data, and thus, using level. Hopefully, more studies will be done to save Fig. 7 In-sample prediction for Hanyang University Seoul Hospital with the most commonly used antibiotics (FQ, 3rd CEP, and 1st CEP). Abbreviations: FQ fluoroquinolones, 3rd CEP 3rd-generation cephalosporins, 1st CEP 1st-generation cephalosporins Kim et al. BMC Infectious Diseases (2018) 18:247 Page 9 of 9 medical personnels’ time and effort surrounding non- Publisher’sNote Springer Nature remains neutral with regard to jurisdictional claims in essential data collection, so that they can concentrate on published maps and institutional affiliations. more important healthcare activities. Author details Department of Internal medicine, College of Medicine, Hanyang University, Conclusions 222-1 Wangsimni-ro, Seongdong-gu, Seoul 04763, South Korea. School of This study showed that the total antibiotic consumption Economic Sciences, Washington State University, Pullman, USA. Department of Economics, College of Political Science & Economics, Korea University, 145 at the hospital level can be represented sufficiently well Anam-ro, Sungbuk-gu, Seoul 02841, South Korea. by a few antibiotics. FQ and AG were the most repre- sentative in the sense of minimizing the mean prediction Received: 19 May 2017 Accepted: 4 May 2018 error, followed by BL-BLI, 4th CEP and 3rd CEP; the mean prediction error is only 1–3% of the monthly anti- References biotic consumption amount. Despite this positive find- 1. Spellberg B, Guidos R, Gilbert D, Bradley J, Boucher HW, Scheld WM, et al. The epidemic of antibiotic-resistant infections: a call to action for the ing, because our analysis is based solely on Korean data medical community from the Infectious Diseases Society of America. Clin and because the medical environment/practice of each Infect Dis. 2008;46(2):155–64. country and each hospital differs, other hospitals may 2. World Health Organization (WHO). World Health Day 2011. Combat drug resistance: no action today means no cure tomorrow. http://www.who.int/ follow a similar modelling strategy to find their own rep- dg/speeches/2011/WHD_20110407/en/. Accessed 19 Oct 2016. resentative antibiotics instead of readily adopting the 3. Ryu S. The new Korean action plan for containment of antimicrobial aforementioned antibiotics as the most representative. resistance. J Glob Antimicrob Resist. 2017;8:70–3. 4. Ventola CL. The antibiotic resistance crisis: part 1: causes and threats. P T. 2015;40(4):277–83. Additional file 5. Schultz L, Lowe TJ, Srinivasan A, Neilson D, Pugliese G. Economic impact of redundant antimicrobial therapy in US hospitals. Infect Control Hosp Epidemiol. 2014;35(10):1229–35. Additional file 1: Supplementary materials. The model for ordinary 6. Gravatt LA, Pakyz AL. Challenges in measuring antibiotic consumption. least squares (OLS) estimator. (DOCX 97 kb) Curr Infect Dis Rep. 2013;15(6):559–63. 7. World Health Organization (WHO). Definition and general considerations of Abbreviations Defined Daily Dose (DDD). http://www.whocc.no/ddd/definition_and_ 1st CEP: 1st-generation cephalosporins; 2nd CEP: 2nd-generation general_considera/. Accessed 28 Oct 2016. cephalosporins; 3rd CEP: 3rd-generation cephalosporins; 4th CEP: 4th- 8. Jun KI, Koo HL, Kim MK, Kang CK, Kim MJ, Chun SH, Song JS, Kim HS, Kim generation cephalosporins; AG: Aminoglycosides; ATC: Anatomic Therapeutic NJ, Kim EC, Oh MD. Trends in antibiotic use in a single university hospital. Chemical; BL-BLI: Beta-lactam/beta-lactamase inhibitors; DDD: Defined Daily Korean J Nosocomial Infect Control 2013; 18(2): 44–50 (in Korean). Dose; DOT: Days of therapy; FQ: Fluoroquinolones; HUS: Hanyang University 9. Yoon YK, Park GC, An H, Chun BC, Sohn JW, Kim MJ. Trends of Seoul Hospital; OLS: Ordinary least squares; SD: Standard deviation; antibiotic consumption in Korea according to national reimbursement SE: Standard errors; WHO: World Health Organization data (2008-2012): a population-based epidemiologic study. Medicine (Baltimore). 2015;94(46):e2100. Acknowledgements 10. Mascarello M, Simonetti O, Knezevich A, Carniel LI, Monticellin J, Busetti M, The authors thank Yeonjae Kim, Shin-Woo Kim, In-Gyu Bae, Won Suk Choi, et al. Correlation between antibiotic consumption and resistance of Sook-In Jung and Hye Won Jeong for their help in collecting data. The authors bloodstream bacteria in a University Hospital in North Eastern Italy, 2008- are also grateful to the Editor, Nathaniel J. Rhodes, Kimberly C. Claeys and Kelly 2014. Infection. 2017. https://doi.org/10.1007/s15010-017-0998-z. R. Reveles for their detailed helpful comments. 11. Cooke J, Stephens P, Ashiru-Oredope D, Charani E, Dryden M, Fry C, et al. Longitudinal tends and cross-sectional analysis of English national hospital antibacterial use over 5 years (2008-13): working towards hospital Authors’ contribution prescribing quality measures. J Antimicrob Chemother. 2015;70(1):279–85. ML and HP are co-corresponding authors, who conceived and designed the 12. Kahlmeter G, Menday P, Cars O. Non-hospital antimicrobial usage and research. BK and HH are co-first authors, who wrote and revised the resistance in community-acquried Escherichia coli urinary tract infection. manuscript. JK provided critical review. BK and HH collected data and J Antimicrob Chemother. 2003;52(6):1005–10. prepared figures. All authors read and approved the final manuscript. 13. Bitterman R, Hussein K, Leibovici L, Carmeli Y, Paul M. Systematic review of antibiotic consumption in acute care hospitals. Clin Microbiol Inect 2016; Ethical approval and consent to participate 22(6): 561. e7–561. e19. This study was approved by the institutional review board of the Hanyang 14. Kim B, Kim J, Kim SW, Pai H. A survey of antimicrobial stewardship programs University Seoul Hospital (IRB number 2013–07-016). The requirements for in Korea, 2015. J Korean Med Sci. 2016;31(10):1553–9. informed consent were waived by the IRB. 15. Barlam TF, Cosgrove SE, Abbo LM, MacDougall C, Schuetz AN, Septimus EJ, et al. Implementing an antibiotic stewardship program: guidelines by the Funding infectious diseases society of America and the society for healthcare This work was supported by a grant from the Korea Healthcare Technology epidemiology of America. Clin Infect Dis. 2016;62(10):e51–77. R&D Project, Nationwide surveillance system of multidrug-resistant pathogens for prevention and control of antimicrobial resistance in Korea (HI12C0756), Ministry of Health and Welfare, Republic of Korea. Availability of data and materials The data used in this paper are available from the corresponding authors upon request. Competing interests The authors declare that they have no competing interests regarding the manuscript.
BMC Infectious Diseases – Springer Journals
Published: May 31, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera