Differences in antibiotic use between COPD and non-COPD residents based on the health information system

Abstract

Objectives

To compare the differences in antibiotic use between COPD and non-COPD residents, and to explore the effect of COPD on antibiotic use.

Methods

Participants aged 40 years old or over from the Songjiang Adult Cohort were included. Information on prescription and baseline survey was collected based on the health information system. A logit-negative binomial Hurdle model was used to explore correlations between COPD and percentage of antibiotic use and average rate of antibiotic prescribing of different types of antibiotic. Multinomial logistic regression was used to assess the association between COPD and antimicrobial combination therapy and routes of administration.

Results

A total of 34576 individuals were included and 1594 (4.6%) were COPD patients. During the 6 years’ follow-up, the percentage of antibiotic use for COPD patients was 98.4%, which was 7.88 (95%CI: 5.24–11.85) times of that for non-COPD patients after adjusting for potential confounders. The prescribing rate was 3220 prescriptions (95%CI: 3063.6–3385.2) per 1000 person-years for COPD patients, which was 1.96 (95%CI: 1.87–2.06) times of that for non-COPD patients. Other beta-lactam antibacterials, Macrolides, lincosamides and streptogramins, and quinolone antibacterials were the most commonly used types of antibiotic. Except for aminoglycoside antibacterials, both percentage of antibiotic use and rate of antibiotic prescription were increased in COPD patients. COPD patients were more likely to be prescribed a maximum of two antibiotics (OR=1.34, 95%CI: 1.20–1.50); and were more likely to use antibiotics intravenously (OR=2.77, 95%CI: 2.47–3.11).

Conclusion

COPD patients were more likely to have increased antibiotic use in a large-scale population-based adult cohort, suggesting COPD patients are a high-priority group for the management of antibiotic use in communities.

Introduction

COPD is a prevalent, heterogeneous, preventable and treatable disease that is characterized by persistent respiratory symptoms and progressive airflow limitation.^1,2 According to the China Pulmonary Health research, 13.7% of the general Chinese population aged over 40 years had spirometry-defined COPD.³ The Global Burden of Disease Study (GBD) reported that COPD accounted for 41.85 (39.64–43.96) deaths per 100 000 individuals, classifying it as the seventh leading cause of years of life lost.^4,5

Studies reported that infections were responsible for 50%–75% of acute exacerbations of COPD (AECOPD), which are common manifestations of the disease.⁶ As a result, antibiotics are commonly used in COPD patients with a frequency ranging from 71.9% to 89% and have become a basic management strategy for AECOPD in primary care.^7,8 Overuse or even abuse of antibiotics can easily cause dysbiosis and diminished susceptibility to antibiotics, promoting antimicrobial resistance (AMR).^9,10 Antibiotic resistance is serious in China,¹¹ especially in COPD patients, and the rate of antibiotic resistance is increasing. Common airway colonizing ultra-broad-spectrum β-lactamase-producing strains in COPD patients are severely resistant to β-lactam antibiotics, which are widely used in the treatment of COPD, such as first/second generation cephalosporins and broad-spectrum penicillins, with resistance rates up to 65% or higher.¹² What is worse, Pseudomonas aeruginosa is almost 100% resistant to cefazolin and ampicillin.^13,14 Although there has not been exact evidence to correlate AMR with the frequency of AECOPD, quality of life or progression of COPD, studies have found that antibiotic-resistant patients with COPD require antibiotic combinations or additional antibiotics use,¹⁵ ultimately leading to a vicious circle of antibiotic abuse and AMR, and increased economic burden or mortality of COPD.^16,17

Previous studies on antibiotic use were based primarily on questionnaire investigations or electronic health records, which had limitations such as recall bias, small sample size, short-term access or a lack of covariates.^7,18 Large medical databases linked to demographic information have become a reliable source for obtaining information on long-term antibiotic use.^19,20 Few studies compared the differences of antibiotic use between people with COPD or not because most previous studies were conducted on hospital-based COPD or AECOPD patients. Besides, research that focused on the types of antibiotic neglected the rate of antibiotic prescribing, whereas research that focused on the rate neglected the routes of administration.^21,22

This study aimed to compare the differences in antibiotic use between COPD and non-COPD residents comprehensively, and to explore the effect of COPD on antibiotic use based on a health information system in the Songjiang Adult Cohort from the Shanghai Suburban Adult Cohort and Biobank.²³

Materials and methods

Ethics

This study was approved by the School of Public Health at Fudan University's Medical Research Ethics Committee (IB#2021-03-0883). Each participant signed an informed consent form and all their personal information was kept in strict confidence.

The Songjiang Adult Cohort

This study was based on a cohort in Songjiang District, Shanghai. The subjects’ enrolment and baseline investigation have been conducted since 2016 in four communities in Songjiang District, Shanghai, China. Nearly 40 000 residents aged 20 to 74 were included. A face-to-face questionnaire interview was conducted individually and structured questionnaires were used to collect the baseline information about demographics, lifestyle and physical health-related factors by self-designed software with an Android tablet. Each enrolled participant received a free physical examination to record their height and weight.²³

The health information system

The cohort database was linked to a local population-based information system that integrated the local electronic health records (EHR), electronic medical records (EMR), hospital information system, chronic disease management system and death registry system by using ID codes. EMR and hospital information system records contained information about inpatient and outpatient visits, diagnoses, examinations and treatment (including medication). The system can provide detailed records of the names and dates of diagnoses and medications.

Participants

Participants for the current analysis were selected according to the following criteria: (i) participants of the Songjiang Adult Cohort; (iii) had completed baseline records; and (iii) aged 40 years and above. According to the health information system, participants died within 1 year after the baseline investigation or had serious illness requiring hospitalization at the beginning of the cohort study were excluded.

Data filtering

Data filtering contains two parts: identification of COPD patients and antibiotic prescriptions. COPD was screened through Chinese keywords and the ICD-10: J44 according to EHR and the chronic disease management system.²⁴ Patients diagnosed in 2022 were excluded and duplicate items generated by database matching were removed.

Antibiotic prescriptions were also screened by keywords based on the linkage to EMR. According to Anatomical Therapeutic Chemical Classification System (ACT) code J01, antibiotics were classified into nine categories: J01A, tetracyclines; J01B, amphenicols; J01C, beta-lactam antibacterials, penicillins; J01D, other beta-lactam antibacterials; J01E, sulphonamides and trimethoprim; J01F, macrolides, lincosamides and streptogramins; J01G, aminoglycoside antibacterials; J01M, quinolone antibacterials and J01X, other antibacterials. Antimycobacterials (code J04) for the treatment of tuberculosis and lepra were not included in the current analysis.²⁵ Antibiotic prescriptions were obtained from 1 January 2016 to 31 of December 2021. Prescriptions were excluded when therapeutic schedule containing the keyword: ‘surgery’. A flow chart of the study is shown in Figure 1.

Figure 1.

Flow chart of the study.

Definition of antibiotic use and covariables

Percentage of antibiotic usage was the percentage of participants who used antibiotics from 1 January 2016 to 31 December 2021.^26,27 The average rate of antibiotic prescribing per 1000 person-years was calculated by counting the total number of antibiotic prescriptions per person from 1 January 2016 to 31 of December 2021 and the denominator was the total number of person-years contributed by the participant during this period.²¹

Antimicrobial combination therapy was two antibiotics added together for additional therapeutic effects.²⁸ It was defined as receiving multiple antibiotic prescriptions in one visit in the study. Records with the same prescription number represented one visit. For further analysis, the highest numbers of antibiotics in combination were recorded during the period.

Antibiotic administration routes were classified as oral, intravenous and other (including endotracheal, ophthalmic, percutaneous and so on). Besides, patterns of antibiotic use were classified as oral administration only, during the 6 years; had intravenous administration routes (including intravenous administration only, oral or other and injectable routes) and other routes only.

Except for COPD, comorbidities listed in the ‘Summary of Antimicrobial Prescribing Guidance’ as being relevant to the decisions to prescribe an antibiotic was taken into account, which included coronary heart disease (CHD), chronic kidney disease (CKD), diabetes, hypertension, stroke and other chronic respiratory diseases (including chronic bronchitis and asthma).²⁹

The BMI (kg/m²) was classified into four categories according to the Chinese standard.³⁰

Statistical analysis

R v.4.2 was used for statistical analysis. Age and BMI were divided into categorical variables. Categorical variables were described by numbers and proportion and were compared between COPD and non-COPD participants by Chi-square test. Overdispersion test in the qcc software package was used to test overdispersion; the R test was used to test zero-inflation (P < 0.05). Therefore, the Poisson model was not suitable for this study. The logit-negative binomial Hurdle model can work with count data of an excess of zeroes and can tackle overdispersion. It has two parts: the ‘zero-inflation model’ is a logistic regression to model the probability that a count is zero or a positive integer value; the ‘count model’ is a truncated-at-zero distribution to model the number of counts greater than zero. The Hurdle model was finally chosen to explore correlations between COPD and antibiotic use. Unordered multinomial logistic regression was used to evaluate the association between COPD and antimicrobial combination therapy and routes of administration.

Results

Demographic characteristics of participants

A total of 34 576 individuals were included in the study and 1594 (4.6%) were COPD patients with documented diagnosis. Of the participants, 40.7% (14 063/34 576) were male. The median age of participants was 65 years. Among all participants, 34.6% had hypertension; 7.5% had chronic bronchitis and 2.1% had asthma, while the proportion in COPD patients was up to 35.4% and 14.2%, respectively. Compared with non-COPD participants, COPD patients were more likely to be male, elder, less educated, retired, in other marital status or have all types of comorbidity that require additional antibiotic use. Besides, a significantly higher proportion of COPD patients were obesity or underweight and had smoking history. (Table 1).

Table 1.

Characteristics of participants

	Total	Non-COPD	COPD	P value
n, %	34 576	32 982 (95.4)	1594 (4.6)
Gender				<0.001
male	14 063 (40.7)	13 220 (40.1)	843 (52.9)
female	20 513 (59.3)	19 762 (59.9)	751 (47.1)
Age				<0.001
40∼49	3070 (8.9)	3063 (9.3)	7 (0.4)
50∼59	8539 (24.7)	8403 (25.5)	136 (8.5)
60∼69	12 236 (35.4)	11 729 (35.6)	507 (31.8)
70∼	10 731 (31.0)	9787 (29.7)	944 (59.2)
BMI				<0.001
normal	15 410 (44.6)	14 784 (44.8)	626 (39.3)
underweight	738 (2.1)	678 (2.1)	60 (3.8)
overweight	13 785 (39.9)	13 154 (39.9)	631 (39.6)
obesity	4643 (13.4)	4366 (13.2)	277 (17.4)
Education				<0.001
primary or below	16 947 (49.0)	15 868 (48.1)	1079 (67.7)
junior high school	12 677 (36.7)	12 265 (37.2)	412 (25.8)
high school or above	4952 (14.3)	4849 (14.7)	103 (6.5)
Marriage				0.019
married	32 352 (93.6)	30 883 (93.6)	1469 (92.2)
other	2224 (6.4)	2099 (6.4)	125 (7.8)
Retirement				<0.001
No	13 404 (38.8)	13 106 (39.7)	298 (18.7)
Yes	21 172 (61.2)	19 876 (60.3)	1296 (81.3)
Health care				0.830
No	203 (0.6)	193 (0.6)	10 (0.6)
Yes	34 373 (99.4)	32 789 (99.4)	1584 (99.4)
Smoke				<0.001
Non-smokers	26 235 (75.9)	25 151 (76.3)	1084 (68.0)
Former smokers	1348 (3.9)	1202 (3.6)	146 (9.2)
Smokers	6993 (20.2)	6629 (20.1)	364 (22.8)
Exercise				0.365
No	23 610 (68.3)	22 538 (68.3)	1072 (67.3)
Yes	10 966 (31.7)	10 444 (31.7)	522 (32.7)
Hypertension				<0.001
No	22 618 (65.4)	21 782 (66.0)	836 (52.4)
Yes	11 958 (34.6)	11 200 (34.0)	758 (47.6)
CHD				<0.001
No	33 149 (95.9)	31 675 (96.0)	1474 (92.5)
Yes	1427 (4.1)	1307 (4.0)	120 (7.5)
Stroke				<0.001
No	33 670 (97.4)	32 161 (97.5)	1509 (94.7)
Yes	906 (2.6)	821 (2.5)	85 (5.3)
Diabetes				0.001
No	31 640 (91.5)	30 217 (91.6)	1423 (89.3)
Yes	2936 (8.5)	2765 (8.4)	171 (10.7)
CKD				0.001
No	34 286 (99.2)	32 717 (99.2)	1569 (98.4)
Yes	290 (0.8)	265 (0.8)	25 (1.6)
Other respiratory diseases				<0.001
No	31 639 (91.5)	30 702 (93.1)	937 (58.8)
Yes	2937 (8.5)	2280 (6.9)	657 (41.2)
Chronic bronchitis				<0.001
No	31 996 (92.5)	30 967 (93.9)	1029 (64.6)
Yes	2580 (7.5)	2915 (6.1)	565 (35.4)
Asthma				<0.001
No	33 841 (97.9)	32 473 (98.5)	1368 (85.8)
Yes	735 (2.1)	509 (1.5)	226 (14.2)

Antibiotic use of participants

Characters of antibiotic use

During the 6-year study period, 15.1% of participants had never used antibiotics, while 98.4% of COPD patients had antibiotic prescriptions. The proportions of nine groups of antibiotics used in COPD patients were higher than that in non-COPD participants (P < 0.05).

Other beta-lactam antibacterials, Macrolides, lincosamides and streptogramins, and quinolone antibacterials were the most commonly used antibiotics for all participants, particularly other beta-lactam antibacterials, which was used by more than 90% of COPD patients.

Similarly, other beta-lactam antibacterials, macrolides, lincosamides and streptogramins, and quinolone antibacterials had the highest number of prescriptions. The proportions of beta-lactam antibacterials, penicillins and other beta-lactam antibacterials prescribed in COPD patients were higher than those in the non-COPD population, while the proportions of other antibiotic categories were lower.

The prescribing rate was 1289 (95%CI: 1277.2∼1311.9) per 1000 person-years. The overall prescribing rate per 1000 person-years for COPD patients was 3220 (95%CI: 3063.6∼3385.2) and 1197 (1185.5,1217.1) for non-COPD participants. Consistent with this, other beta-lactam antibacterials had the highest prescribing rate of 573 (95%CI: 566.9∼586.1) per 1000 person-years. (Table 2).

Table 2.

Antibiotic use of participants

	Percentage of antibiotic usage n (%)			Number of antibiotic prescriptions n (%)			Average rate of antibiotic prescribing per 1000 person-years (95%CI)
	Total	Non-COPD	COPD	Total	Non-COPD	COPD	Total	Non-COPD	COPD
J01A	3790 (11.0)	3503 (10.6)	287 (18.0)	7173 (2.8)	6497 (2.9)	676 (2.3)	36 (34.6, 38.2)	34 (32.8, 36.3)	75 (60.7, 88.8)
J01B	1171 (3.4)	1085 (3.3)	86 (5.4)	2062 (0.8)	1903 (0.8)	159 (0.5)	10 (9.6, 11.3)	10 (9.3, 11.0)	18 (12.4, 22.3)
J01C	2529 (7.3)	2279 (6.9)	250 (15.7)	6138 (2.4)	5080 (2.3)	1058 (3.6)	31 (29.3, 33.4)	27 (25.6, 29.0)	117(889.4, 143.1)
J01D	24 110 (69.7)	22 629 (68.6)	1481 (92.9)	113 298 (44.4)	98 206 (43.5)	15 092 (51.7)	573 (566.9, 586.1)	521 (515.1, 532.0)	1666 (1571.5, 1771.2)
J01E	255 (0.7)	228 (0.7)	27 (1.7)	297 (0.1)	261 (0.1)	36 (0.1)	2 (1.3, 1.7)	1 (1.2,1.6)	4 (2.3, 5.6)
J01F	16 845 (48.7)	15 756 (47.8)	1089 (68.3)	41 592 (16.3)	37 314 (16.5)	4278 (14.7)	210 (207.2, 214.8)	198 (194.7, 201.8)	472 (439.1, 510.0)
J01G	2166 (6.3)	2007 (6.1)	159 (10.0)	3552 (1.4)	3283 (1.5)	269 (0.9)	18 (16.9, 19.0)	17 (16.4, 18.4)	30 (23.7, 35.1)
J01M	17 968 (52.0)	16 747 (50.8)	1221 (76.6)	60 036 (23.5)	53 580 (23.7)	6456 (22.1)	304 (298.3, 310.5)	284 (278.9, 290.5)	713 (662.3, 761.1)
J01X	10 248 (29.6)	9721 (29.5)	527 (33.1)	20 794 (8.2)	19 649 (8.7)	1145 (3.9)	105 (102.7, 107.8)	104 (101.6, 106.9)	126 (112.9, 137.9)
Total	29 350 (84.9)	27 782 (84.2)	1568 (98.4)	254 942	225 773	29 169	1289 (1277.2, 1311.9)	1197 (1185.5, 1217.1)	3220 (3063.6, 3385.2)

Antibiotic combination therapy

During the 6-year study period, 32.3% (11 166/34 576) of participants had taken a maximum of two antibiotics in combination; 0.4% (674/34 576) had taken a combination therapy of three or more antibiotics; the proportion for COPD patients was 44.0% (701/1594) and 2.6%, respectively. The number of multiple antibiotic prescriptions for all participants and COPD patients was 46 938 and 3780, respectively. COPD patients were more likely to be prescribed antibiotics in combination. (Table 3).

Table 3.

Combination of antibiotics

	Percentage of participants prescribed multiple antibiotics n (%)				Number of antimicrobial combination therapies n (%)
	Total	Non-COPD	COPD	P	Total	Non-COPD	COPD	P
0	5226 (15.1)	5200 (15.8)	26 (1.6)	<0.001	5226 (2.0)	5200 (2.3)	26 (0.1)	<0.001
1	17 510 (50.6)	16 702 (50.5)	826 (51.8)		208 003 (79.9)	182 542 (79.0)	25 461 (87.2)
2	11 166 (32.3)	10 465 (31.7)	701 (44.0)		44 556 (17.1)	40 986 (17.7)	3570 (12.2)
≥3	674 (0.4)	633 (1.9)	41 (2.6)		2383 (0.9)	2245 (1.0)	138 (0.5)

Routes of administration for antibiotics

For non-COPD group and COPD patients, there were 124 875 (55.3%) and 14 557 (49.9%) prescriptions that were prescribed orally, respectively; 37 875 (16.8%) and 8430 (28.9%) prescriptions that were prescribed intravenously; 63 023 (27.9%) and 6182 (21.2%) prescriptions prescribed in other routes. In the 6 years, 15 519 (47.1%) non-COPD population and 464 (29.1%) COPD patients used antibiotics orally only; 10 620 (32.2%) and 68.1% participants had intravenous therapy; 1643 (5.0%) and 19 (1.2%) participants used in other routes only. In contrast to non-COPD patients, COPD patients used oral antibiotics at a lower rate during the study period, and a higher percentage had a history of injectable antibiotic use. (P < 0.001).

Effect of COPD on antibiotic use

Percentage and average rate of antibiotic use

According to the zero-inflation model, COPD patients were 7.88 (95%CI: 5.24–11.85) times more likely to use antibiotics. In addition, participants who were elderly, female, former smokers, married, retired and had a history of hypertension, CHD, and other respiratory diseases, and had health insurance were more likely to use antibiotics.

According to the count model, COPD patients tended to receive more antibiotic prescriptions. The average rate of antibiotic prescribing per 1000 person-years in COPD patients was 1.96 (95%CI: 1.87–2.06) times that in non-COPD participants. In addition, higher rate of antibiotic prescribing was associated with old age, female, having each of the comorbidities of overweight, obesity and retirement. (Table 4).

Table 4.

Effect of COPD on antibiotic use

Variables	Zero-inflation model		Count model
	RR	95%CI	RR	95%CI
COPD
No	1.00		1.00
Yes	7.88	5.24, 11.85	1.96	1.87, 2.06
Age
40–49	1.00		1.00
50–59	1.7	1.54, 1.89	1.27	1.22, 1.33
60–69	1.9	1.69, 2.14	1.35	1.28, 1.41
≥70	2.01	1.75, 2.30	1.48	1.40, 1.55
Gender
Male	1.00		1.00
Female	1.38	1.27, 1.51	1.08	1.05, 1.12
Hypertension
No	1.00		1.00
Yes	1.46	1.35, 1.58	1.05	1.03, 1.07
CHD
No	1.00		1.00
Yes	1.46	1.17, 1.82	1.11	1.05, 1.17
Stroke
No	1.00		1.00
Yes	1.00	0.79, 1.28	1.14	1.07, 1.22
Diabetes
No	1.00		1.00
Yes	1.12	0.98, 1.27	1.16	1.12, 1.20
CKD
No	1.00		1.00
Yes	0.89	0.61, 1.29	1.18	1.05, 1.32
Other respiratory diseases
No	1.00		1.00
Yes	1.71	1.46, 2.00	1.31	1.26, 1.36
BMI
Normal	1.00		1.00
Underweight	1.03	0.83, 1.27	1.06	0.98, 1.14
Overweight	1.04	0.97, 1.12	1.05	1.03, 1.08
Smoke
Non-smokers	1.00		1.00
Former smokers	1.34	1.10, 1.63	0.95	0.90, 1.01
Smokers	0.92	0.84, 1.02	0.86	0.83, 0.89
Obesity	1.08	0.97, 1.20	1.13	1.09, 1.17
Marriage
Married	1.00		1.00
Other	0.78	0.68, 0.89	1.00	0.95, 1.04
Retirement
Yes	1.00		1.00
No	0.82	0.75, 0.90	0.90	0.88, 0.93
Health insurance
No	1.00		1.00
Yes	1.51	1.05, 2.18	1.07	0.93, 1.23

Model was adjusted for age, gender, CHD, CKD, diabetes, hypertension, stroke and other chronic respiratory diseases, BMI, smoke, marriage, retirement and health insurance.

In terms of the type of antibiotic, prescriptions of eight groups of antibiotics were increased in COPD patients both in percentage of antibiotic usage and rate of antibiotic prescribing, except for aminoglycoside antibacterials that had a higher percentage of antibiotic usage only in COPD patients. COPD patients were 4.33 (95%CI: 3.53–5.32) times more likely to be prescribed other beta-lactam antibacterials and the prescriptions of other beta-lactam antibacterials were 2.40 (95%CI: 2.22–2.58) times than non-COPD patients. (Table 5).

Table 5.

Effect of COPD on use of different types of antibiotic

		J01A	J01B	J01C	J01D	J01E	J01F	J01G	J01M	J01I
Zero-inflation model	Non-COPD	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
	COPD	1.62 (1.40, 1.88)	1.27 (1.14, 1.43)	2.58 (2.19, 3.03)	4.33 (3.53, 5.32)	1.74 (1.1, 2.75)	2.05 (1.82, 2.31)	1.40 (1.16, 1.70)	2.54 (2.23, 2.90)	1.28 (1.13, 1.44)
Count model	Non-COPD	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
	COPD	1.48 (1.12, 1.96)	1.19 (1.09, 1.30)	2.14 (1.67, 2.74)	2.40 (2.22, 2.58)	1.98 (1.75, 2.23)	1.91 (1.73, 2.10)	0.93 (0.64, 1.36)	1.78 (1.60, 1.97)	1.27 (1.05, 1.54)

Models were adjusted for age, gender, CHD, CKD, diabetes, hypertension, stroke and other chronic respiratory diseases, BMI, smoke, marriage, retirement and health insurance.

Antibiotic combination and administration

The unordered multinomial logistic regression model was adjusted for age, gender, CHD, CKD, diabetes, hypertension, stroke, chronic bronchitis, asthma, BMI, smoke, marriage, retirement and health insurance. In the adjusted model, compared to participants who were prescribed one antibiotic in a visit, the percentage of participants who were prescribed a maximum of two antibiotics was higher in COPD patients (OR = 1.34, 95%CI: 1.20–1.50); there was no clear association between combination of three or more antibiotics and COPD (OR = 1.26, 95%CI: 0.90–1.77).

Compared to participants used antibiotics orally only, COPD patients were 2.77 times more likely than non-COPD participants to use antibiotics intravenously (OR = 2.77, 95%CI: 2.47–3.11); and were less likely to only use antibiotics through other routes (OR = 0.36, 95% CI: 0.23,0.58).

Discussion

In this large population-based cohort study, we comprehensively investigated the status of antibiotic use including the percentage, common prescribed types, combination therapy and routes of administration, and the effect of COPD on antibiotic use.

We found that COPD patients had higher percentage of antibiotic usage and rate of antibiotic prescribing especially in beta-lactam antibacterials, penicillins; other beta-lactam antibacterials; macrolides, lincosamides and streptogramins; and quinolone antibacterials. It was consistent with GOLD guidelines recommending the use of amoxicillin, cephalosporins, azithromycin and levofloxacin for COPD patients.² Percentage of antibiotic usage in this study was higher than the ideal antibiotic prescribing proportions.³¹ A study in the UK reported that the rate of antibiotic prescribing to COPD patients was 3-fold greater than that for the general population by using negative binomial regression model.²¹ Our result suggested that the average rate of antibiotic prescribing was 3220 prescriptions per 1000 person-years for COPD patients, nearly 2.5 times greater than that for the general population (1289 prescriptions per 1000 person-years), which was similar to the study in the UK. However, the negative binomial model cannot tackle an excess of zeroes. The multivariate analysis result of Hurdle model in our study revealed that COPD patients were 7.88 times more likely to use antibiotics and the rate of antibiotic prescribing was 1.96 times greater.

Consistent with the results of previous studies and the Summary of Antimicrobial Prescribing Guidance, participants with comorbidities and overweight or obesity had more antibiotic use.^21,29 Meanwhile, previous findings revealed that age, gender, marriage, health insurance and retirement status were related to healthcare-seeking times,^32,33 resulting in effects on antibiotic prescriptions. Contrarily, current smokers tended to use fewer antibiotics in our study,²¹ which might be attributed to the fact that the diagnosed COPD patients had somehow changed their smoking status.

The percentage of two antibiotics combination for COPD patients was higher in this study (44.0%) than in the other two hospital-based studies, which had respective percentages of 31.3% and 19.2%.^34,35 The finding that COPD patients were more likely to be prescribed combined antibiotics and used antibiotics intravenously was consistent with previous studies.^15,36,37

The high rate of antibiotic prescriptions, as well as combined and intravenous therapy among COPD patients indicate that COPD patients suffer more infections.³⁸ However, antibiotic use for COPD remains controversial.^39,40 COPD patients’ frequent inflammatory symptoms caused inappropriate antibiotic use.⁴¹ According to the clinical criteria, antibiotic treatment for AECOPD should be judged by sputum, dyspnoea and mechanical ventilation,⁴² while in primary care and some hospitals, the dilemma of whether to prescribe antibiotics is exacerbated because of the difficulty to distinguish between chronic respiratory disease and infection.⁴³ Considering that reduction in antibiotic use had cost-effective effect on curbing AMR,⁴⁴ antibiotic use in COPD patients, a high-antibiotic-exposure group, needs to be further emphasized. It is critical to instruct clinicians to recognize circumstances that warrant antibiotic use in COPD patients strictly adhere to existing guidelines, and develop more intense management strategies.

The strengths of this study include the large sample, the assessment of long-term antibiotic prescriptions information at individual-level, the consideration of various types of antibiotic and rate of antibiotic prescribing, and the ability to adjust for SES, lifestyle habits, comorbidities and other potential confounding variables. Moreover, the study used the concept of ‘person-time’ rather than merely the number of prescriptions to investigate the associations and exemplified the high number of antibiotic prescriptions in COPD patients with specific data that can act as an important cautionary warning worthy of clinical consideration when prescribing antibiotics for COPD patients.

Similar to other pharmacoepidemiologic studies, one of the limitations is that antibiotic prescriptions cannot represent the residents’ actual usage. However, considering that most antibiotic prescriptions were based on positive symptoms caused by infection,⁴⁵ it does suggest that COPD patients are a high-priority population for antibiotic use. In addition, some COPD patients were diagnosed during the study period. A cohort study on antibiotic prescribing before and after the diagnosis of COPD found that antibiotic use in COPD patients was already at a high level before the diagnosis of COPD and increased rapidly when respiratory symptoms onset.⁴³ As a result, effect of COPD on antibiotic use might be underestimated in this study. We were unable to detect the effect of severity or acute exacerbations of COPD on antibiotic use due to limitations of the health information system. Further studies are required to confirm the association between COPD, antibiotic use and AMR.

Conclusion

In conclusion, our study found that COPD was associated with increased antibiotic use in a large-scale population-based Chinese adult cohort, suggesting COPD patients are a high-priority group for the management of antibiotic use in communities.

Acknowledgements

We are grateful to all for their assistance and cooperation. The authors are grateful to each participant.

Funding

This work was supported by the National Natural Science Foundation of China (grant number 82073634), the Shanghai Three-year Action Plan for Public Health from Shanghai government (grant number GWVI-11.1-23), and the National Key Research and Development Program from Ministry of Science and Technology of China (grant number 2017YFC0907000).

Transparency declarations

None to declare.

Author notes

Xin Yin and Yonggen Jiang made an equal contribution to the article.