Study design
We retrospectively analysed aggregated monthly surveillance data on antibiotic sales to 586 hospitals from 28 provinces in China from January 2011 to December 2018 (Table 1).
Table 1 Distribution of sample hospitals
Region a
|
Tertiary b
|
Secondary c
|
Eastern
|
256/982 (26.1)
|
73/1991 (3.7)
|
Middle
|
132/602 (21.9)
|
36/2132 (1.4)
|
Western
|
64/484 (13.2)
|
25/1763 (1.7)
|
Total
|
452/2068 (21.9)
|
134/5886 (2.3)
|
a: Classification of the regions was obtained from the China Health Statistics Yearbook. Eastern region: Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Shandong, and Guangdong; Middle region: Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, and Hunan; Western region: Inner Mongolia, Chongqing, Guangxi, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Ningxia, and Xinjiang.
b: Percentage in parenthesis was calculated by dividing the number of sampled tertiary hospitals by the total number of tertiary hospitals in the region.
c: Percentage in parenthesis was calculated by dividing the number of sampled secondary hospitals by the total number of secondary hospitals in the region.
Data source
All the data was obtained from China Medicine Economic Information (CMEI), an observational database containing information of drug sales records (i.e. acquisition data, not the usage data) in medical institutions from 28 provinces (out of 34) across the country (Qinghai, Tibet, Hainan, Hongkong, Macau, and Taiwan excluded). The database covers more than 1,000 city-level public hospitals, including outpatient and inpatient, across mainland China. The sales of participating hospitals, which were all public hospitals, account for approximately 40% of total drug sales at city level public hospitals in China. Hospitals were sampled hierarchically based on geographical and socio-economic factors[14]. Hospitals were selected on the basis that they each had full records of antibiotic sales during the study period of 8 years. Based on this criterion, the selected 452 tertiary hospitals accounted for 21.9% of the total tertiary hospitals and the selected 134 secondary hospitals accounted for 2.3% of the total secondary hospitals in the study regions.
Data collection and management
We extracted monthly antibiotic sales records data from the CMEI electronic database. Information including the generic name, sales amount, dosage form, strength, the route of administration, and geographical data were collected. Hospital names were concealed to protect confidentiality.
Sales data were categorized according to Anatomical Therapeutic and Chemical (ATC) classification J01 (i.e. antibacterial for systemic use) expressed in defined daily dose (DDD) as a measurement unit, following the recommendation of the WHO Collaborating Center for Drug Statistic Methodology [15]. The DDD of the drugs which could not be coded in the ATC system was calculated as the recommended daily amount for each study medication based on the dosage regimen recommended in the manufacturers’ instructions, as approved by China Food and Drug Administration. A total of 186 unique chemical substance names were identified in single or combination antibiotics. These antibiotics were aggregated into 32 ATC-4 classes and then into 9 ATC-3 groups. Data were managed and analysed in Microsoft Excel 2013 and STATA 14.0 (StataCorp LLC, Texas, USA).
Data analysis
To make the antibiotic sales data consistent with the international standards, the data were converted into DDD per 1,000 inhabitants per day (DID) at the level of the active substance. Based on the following two assumptions, equation1 was adopted to calculate the weighted population as a proxy for the population our sample hospitals had covered. The first assumption is that there was no significant difference in the distribution of the sample hospitals across the provinces; second, there was no significant difference in the distribution of the population which was covered by the sample hospitals across the provinces. To avoid bias in calculating inhabitants, the inhabitants’ coverage were calculated together for outpatient and inpatient, instead of separate calculation as we did before [14]. The inhabitants’ coverage for secondary and tertiary hospitals were calculated separately.
[Please see supplementary files section to access the equation] (1)
Yi: Coverage inhabitants in a given year;
Pi: Total population in a given year in province i;
ni: Number of sample hospitals in province i;
Ni: Number of total hospitals in province i;
mi: Number of inpatients and outpatients in sample hospitals in province i;
Mi: Number of inpatients and outpatients in all hospitals in province i.
In addition to ATC classification, we adopted ‘Access, Watch, Reserve’ (AWaRe) categorization established by WHO as part of the update of the WHO Model List of Essential Medicines in 2017 to analyse the antibiotic sales [16].
To derive a comparable metric of antibiotic sales across time, we calculated the compound annual growth rate (CAGR) of antibiotic sales [4].
[See supplementary files] (2)
C2018: Total antibiotic sales for the year 2018 (expressed in DID).
C2011: Total antibiotic sales for the year 2011 (expressed in DID).
All the relevant census data for calculating inhabitants were collected from China Health Statistics Yearbook and China Statistics Year Book [17]. Linear regression analysis was adopted to determine the trends in antibiotic use with time. A difference with p < 0.05 was considered to indicate statistical significance.