Effectiveness of Educational Interventions for Health Workers on Antibiotic Prescribing in Outpatient Settings in China: A Systematic Review and Meta-Analysis

Educational interventions are considered an important component of antibiotic stewardship, but their effect has not been systematically evaluated in outpatient settings in China. This research aims to evaluate the effectiveness of educational interventions for health workers on antibiotic prescribing rates in Chinese outpatient settings. Eight databases were searched for relevant randomized clinical trials, non-randomized trials, controlled before–after studies and interrupted time-series studies from January 2001 to July 2021. A total of 16 studies were included in the systematic review and 12 in the meta-analysis. The results showed that educational interventions overall reduced the antibiotic prescription rate significantly (relative risk, RR 0.72, 95% confidence interval, CI 0.61 to 0.84). Subgroup analysis demonstrated that certain features of education interventions had a significant effect on antibiotic prescription rate reduction: (1) combined with compulsory administrative regulations (RR With: 0.65 vs. Without: 0.78); (2) combined with financial incentives (RR With: 0.51 vs. Without: 0.77). Educational interventions can also significantly reduce antibiotic injection rates (RR 0.83, 95% CI 0.74 to 0.94) and the inappropriate use of antibiotics (RR 0.61, 95% CI 0.51 to 0.73). The limited number of high-quality studies limits the validity and reliability of the results. More high-quality educational interventions targeting the reduction of antibiotic prescribing rates are needed.


Introduction
Antimicrobial resistance (AMR) is one of the greatest challenges threatening global health, economy, and security today. It is estimated that by 2050, AMR will be responsible for 10 million deaths per year [1]. The World Bank has calculated that the healthcare costs of AMR will be as high as one trillion US dollars per year by 2050 [2]. Urgent action is needed to address this worsening situation [3].
China is the largest consumer of antibiotics across the medical and agricultural sectors: around half of the total 162,000 tons is used in medicine [4]. Concerns about the impact of massive antibiotic use on AMR have led the Chinese government to introduce a series of regulations on antibiotic use with a focus on rational prescribing in clinical practice [5]. Since 2011, several announcements relating to antimicrobial stewardship have been published by the National Health Commission. In 2016, the National Health Commission, Doctors are the prescribers at outpatient facilities, who were deemed as the primary participants in our study. We also included studies targeting both prescribers and other health workers such as caregivers, nursing staff and pharmacist. Though these personnel did not prescribe drugs directly, they may affect the prescribing behaviors of doctors to some extent.

Interventions
We included studies comprising educational interventions as individual programs or incorporated with other interventions. An educational intervention was defined as a program designed to encourage doctors to improve their practice performance through information or training strategies. Specifically, it includes delivery of print/audio-visual learning materials (printed matter, protocols, guidelines, self-instruction materials or manual) or interactive group learning or discussion (lectures, seminars, conferences, group sessions or tutorials) [14]. Non-educational interventions include: (1) prescription feedback (consisting of peer or expert review of prescriptions and feedback); (2) mandatory administrative regulations, including setting specific prescribing targets, implementing prescription audit and/or displaying ranking information); (3) financial incentives (reward or punishment).

Comparisons and Outcomes
Comparisons were made with usual practice. The primary outcome was antibiotic prescription rate. Secondary outcomes were specific prescription rates for a particular disease, types of antibiotics prescribed, that is, broad-spectrum or parenteral antibiotics, prescription rates for 2 or more kinds of antibiotics, and knowledge improvement.

Study Design
With reference to EPOC (Cochrane Effective Practice and Organization of Care), the following study designs were included [18]: randomized clinical trials (RCTs), nonrandomized trials (NRCTs), controlled before-after studies (CBA) and interrupted time series (ITS) studies.

Data Extracting
We extracted the following items according to the Cochrane Handbook for Systematic Reviews: first author, year of publication, study design, setting, location China, participants, intervention details, target illness, duration and outcomes measures.

Risk of Bias Assessment
The risk of bias was assessed using EPOC recommended criteria for studies with a separate control group (RCTs, NRCTs and CBA, with 9 criteria) and ITS (with 7 criteria) [19]. Common criteria include missing outcome data, selection of reported outcomes, expected interventions and contamination. ITS additionally included assessments for predetermined and independent effect traits, while other designs included random sequences, assignments, baseline measurements and characteristics. The risk of bias is low if all criteria were scored as 'low'; medium if one or two criteria were scored as 'unclear' or 'high', and high if three or more criteria were scored as 'unclear' or 'high' [12,20,21].

Data Analysis
A meta-analysis was performed on the primary outcome of antibiotic prescription rates, as well as secondary outcomes where enough studies were included. We extracted numbers of total prescriptions and prescriptions including antibiotics in both intervention and control groups pre-and post-intervention for RCTs or NRCTs. For CBA and ITS, we extracted numbers of total prescriptions and those where antibiotics were included pre-and post-intervention. Given the expected statistical heterogeneity, we estimated the pooled value with a random-effects model. The estimated effect size was shown in forest plots with risk ratio (RR) and 95% confidence intervals (95% CI).
Subgroup analysis was conducted in separated groups for education-only or educationplus interventions, with or without feedback interventions, compulsory administrative regulations, financial incentives and delivery of education online (learning materials, teaching or discussion via internet-, electronic-, and/or smart phone-based media) or offline (printed education materials, bulletin board in workplace or teaching or discussion face-to-face), to analyze the reduction of antibiotic prescription rate. To examine the heterogeneity, we applied a leave-one-out analysis. Cochrane's Q statistic and I 2 statistic, indicate the proportion of total variance due to heterogeneity. The possibility of publication bias was examined by Egger's test, as well as funnel plots. EndNote X9 was used to store bibliography and Microsoft Excel was for data management. R software version 4.1.2 with Review Manager version 5.4 were used for analyses and plots making.

Population
Many studies included caregivers, pharmacists and other medical staff, in addition to doctors, the prescribers. Six studies only targeted prescribers [25][26][27]29,30,32], while two studies included both prescribers and pharmacists [33,36] (without prescriptive authority in China), and another two studies included prescribers, pharmacists and caregivers as well [24,31]. Another six studies targeted all medical staff in outpatient settings [22,23,28,34,35,37], including two studies conducting universal interventions aimed to measure the implementation of an AMR regional policy [22,28], and the other four were set in primary care settings where other medical staff may easily influence doctors' prescribing behaviors [23,34,35,37]. Only two studies reported actual numbers of participants, including 977 health workers in 100 township health centers [23] and 820 doctors in a tertiary hospital in Beijing [28]. The other 14 studies did not report the number of individuals included in the intervention, but rather analyzed the impact of the intervention on prescribing through hospital-wide antimicrobial prescribing.
Of the 16 papers, 15 analyzed the effectiveness of interventions versus no intervention, using a control group or comparison of pre-and post-intervention. Chen [23] compared the effect of online text messages (intervention group) and offline training lectures (control group) specifically for the treatment of upper respiratory tract infection.

Antibiotic Prescription Rate
A total of 11 studies reported antibiotic prescription rates with a total number of prescriptions. Educational interventions were found to reduce antibiotic prescription rates significantly (RR 0.72, 95% CI 0.61 to 0.84) (Figure 3). However, no difference was found in antibiotic prescription rate between intervention and control group for education-only interventions (RR 1.00, 95% CI 0.93 to 1.08), while education-plus interventions showed a significant reduction in reducing antibiotic prescription rates (RR 0.67, 95% CI 0.57 to 0.79) (Figure 4). Educational interventions with (RR 0.73, 95% CI 0.61 to 0.86) or without (RR 0.70, 95% CI 0.51 to 0.97) feedback interventions showed similar significant effects on antibiotic prescription rate reduction (Figure 4). In addition, interventions including compulsory administrative regulations (RR 0.65, 95% CI 0.49 to 0.87) reduced antibiotic prescription rates more than those without, as measured by pre-and post-introduction of compulsory regulations (RR 0.78, 95% CI 0.65 to 0.92) (Figure 4). Educational interventions combined with financial incentives (RR 0.51, 95% CI 0.35 to 0.74) also showed a greater effect on reducing antibiotic prescription rate than those without (RR 0.77, 95% CI 0.67 to 0.89) (Figure 4). There was no significant difference between intervention and control groups when receiving online interventions (RR 0.86, 95% CI 0.71 to 1.05), while offline interventions showed a significant difference (RR 0.67, 95% CI 0.55 to 0.82 (Figure 4). Given the heterogeneity in our main results, we also conducted a leave-one-out analysis, and none of the omitted studies significantly influenced our results ( Figure S2).

Parenteral Use of Antibiotics
Seven education-only or education-plus interventions showed pooled results comparing parenteral use of antibiotics in control and intervention groups. The intervention group had a 17% reduction in parenteral use rate (RR 0.83, 95% CI 0.74 to 0.94, details in Figure S3).

Multiple Antibiotic Rates
The education-plus intervention did not reduce multiple antibiotic rates as reported in five studies (RR 0.73, 95% CI 0.72 to 1.29, Figure S4). Among those studies, two [36,37] reported on the use of two or three antibiotics respectively, while others reported an overall rate for two or more antibiotics.

Antibiotic Prescription Inappropriateness Rate
Four studies documented antibiotic prescription inappropriateness using similar evaluation criteria in Table S3 except one [35] with unspecified criteria. Results in Figure  S5 showed education-plus interventions did reduce antibiotic prescription inappropriateness rate (RR 0.61, 95% CI 0.51 to 0.73).

Parenteral Use of Antibiotics
Seven education-only or education-plus interventions showed pooled results comparing parenteral use of antibiotics in control and intervention groups. The intervention group had a 17% reduction in parenteral use rate (RR 0.83, 95% CI 0.74 to 0.94, details in Figure S3).

Multiple Antibiotic Rates
The education-plus intervention did not reduce multiple antibiotic rates as reported in five studies (RR 0.73, 95% CI 0.72 to 1.29, Figure S4). Among those studies, two [36,37] reported on the use of two or three antibiotics respectively, while others reported an overall rate for two or more antibiotics.

Antibiotic Prescription Inappropriateness Rate
Four studies documented antibiotic prescription inappropriateness using similar evaluation criteria in Table S3 except one [35] with unspecified criteria. Results in Figure S5 showed education-plus interventions did reduce antibiotic prescription inappropriateness rate (RR 0.61, 95% CI 0.51 to 0.73).

Changes in Types of Antibiotics or Target Diseases
Only two studies reported the specific types of antibiotics used ( Figure 5). Both studies reported Cephalosporins as the most frequently used antibiotics before and after the intervention. In Liu's study carried out in township health centers in Guangdong Province, cefuroxime tablets and ceftezole sodium for injection were the most widely used antibiotics in the included community health centers from 2012-2014 [26], and there was an increase in the proportion of Cephalosporins used from 66.8% to 79.0% following the intervention [26]. The other study by Li et al. reported a decrease from 38.3% to 34.6% of Cephalosporins in 2015, though it always ranked first, among 17 primary health centers in Zhejiang province [37].
intervention. In Liu's study carried out in township health centers in Guangdong Province, cefuroxime tablets and ceftezole sodium for injection were the most widely used antibiotics in the included community health centers from 2012-2014 [26], and there was an increase in the proportion of Cephalosporins used from 66.8% to 79.0% following the intervention [26]. The other study by Li et al. reported a decrease from 38.3% to 34.6% of Cephalosporins in 2015, though it always ranked first, among 17 primary health centers in Zhejiang province [37].
Liu's study also measured changes in antibiotic prescriptions for specific diseases [26]. While the antibiotic prescription rate showed an overall decline, the proportion of antibiotic prescriptions for acute upper respiratory tract infection increased from 46.7% to 56.0%, and numbers of antibiotic prescriptions and proportions for trauma, mouth ulcers, indigestion, acute gastroenteritis all declined [26].

Knowledge Improvement
One cluster RCT compared the effect of sending learning text messages with traditional training workshops [23]. Researchers used 10 multiple-choice questions on the appropriate treatment of the selected diseases and complications via a telephone survey. The knowledge score increased 16% (95% CI: 15.7-16.3%) in the intervention group, with no significant changes in the control group. Another study reported qualitative results on doctors' self-reported knowledge improvement. The participants reported better knowledge and improved confidence in the appropriate use of antibiotics and had increased their use of guidelines for prescribing following the intervention [29]. Liu's study also measured changes in antibiotic prescriptions for specific diseases [26]. While the antibiotic prescription rate showed an overall decline, the proportion of antibiotic prescriptions for acute upper respiratory tract infection increased from 46.7% to 56.0%, and numbers of antibiotic prescriptions and proportions for trauma, mouth ulcers, indigestion, acute gastroenteritis all declined [26].

Knowledge Improvement
One cluster RCT compared the effect of sending learning text messages with traditional training workshops [23]. Researchers used 10 multiple-choice questions on the appropriate treatment of the selected diseases and complications via a telephone survey. The knowledge score increased 16% (95% CI: 15.7-16.3%) in the intervention group, with no significant changes in the control group. Another study reported qualitative results on doctors' self-reported knowledge improvement. The participants reported better knowledge and improved confidence in the appropriate use of antibiotics and had increased their use of guidelines for prescribing following the intervention [29].

Discussion
Our study shows that education interventions can significantly reduce the antibiotic prescription rate when combined with other types of interventions, especially with compulsory administrative regulations or financial incentives. Offline interventions also had a greater effect on reducing antibiotic prescription rates than online ones. However, due to the high heterogeneity, the results should be interpreted with caution.
Our findings are consistent with previous systematic reviews worldwide [12], The results in the primary analyses indicating that educational interventions can achieve significant reductions in antibiotic prescribing, antibiotic injection rates and inappropriate prescribing, by combining with other strategies (Figure 3). As has been reported, one of the main driving factors for excessive antibiotic prescription is the inappropriateness of antibiotic use for the condition. It derives from inadequate knowledge of guideline recommendations on antibiotic indications [38], as well as habit, the use of the same antibiotics over many years [39]. Thus, comprehensive educational interventions are necessary to improve antibiotic prescription behaviors.
Though the results in subgroup analyses showed that educational interventions conducted offline seemed to be more effective than those online, it should be taken with caution since only two studies were conducted online. A meta-analysis of prescribing to children for upper respiratory infections also demonstrated the positive effect of face-to-face training for appropriate antibiotic prescribing in high-income countries (RR 0.77, 95% CI, 0.65 to 0.92) [40]. Online interventions such as sending educational text messages only may have marginal effects in reducing antibiotic prescription rates [23], while online training seemed to be effective when combined with offline educational materials [24,36]. The effect was also influenced by the types and duration of the intervention [41]. An RCT showed that short-term offline sessions failed to reduce the antibiotic prescribing rate. The authors attributed this to the short duration and limited effect of a single educational intervention [21]. This suggests that interventions should be sustained over the long term until habits are formed. Holding conferences is a widely-used educational method, which has shown promising effects in many studies as a primary component of continuing education [42]. Future studies are needed to clarify the optimal types of educational interventions in China.
Concerning the two education-only interventions, we found that the single intervention with only brief reading materials was not effective. As a previous review suggested, a single intervention has little or no effects on prescribing behavior [43]. Nevertheless, a combination of educational and other interventions is effective (Figure 4). In our study, the three strategies most often combined with educational interventions were prescription feedback, mandatory regulations and financial rewards and sanctions, respectively. Prescription feedback has been widely recommended [44] and is well validated in other studies [45][46][47]. Feedback and audit as well as comparisons of behaviors with peers can help to maximize the impact and improve the acceptability of stewardship interventions [38]. However, we found no difference in antibiotic prescription rates with or without additional prescription feedback. This may provide additional evidence for intervention models in combination with antibiotic education.
In addition, our study indicated a stronger effect on reducing antibiotic prescribing rates among educational interventions which were combined with compulsory administrative regulations or financial incentives. This suggests that a combination of educational and mandatory interventions could have a greater impact on reducing antimicrobial prescribing rates in China. In addition to the exams included in the educational interventions, other mandatory administrative regulations such as specific prescribing targets or displaying ranking information of prescribing behavior of doctors are also helpful. This is the first review of the effect of educational interventions conducted in Chinese outpatient settings. Our work has provided an evidence base for future studies conducted in China. Nevertheless, there are limitations. Firstly, we did not include unpublished studies in the search and identification. Some of these were probably unpublished because of non-significant changes in outcomes, introducing considerable publication bias. Secondly, due to the limited number of articles, we cannot state which educational intervention is more effective. This needs to be explored in future studies. Thirdly, since interventions were grouped when conducted, we cannot obtain the effect of a single measure, for example, the automatic prescription screening system combined with other interventions in one study [28]. Finally, since outcomes were antibiotic prescribing rates measured at the facility rather than individual level, and staffing changes were not reported, the results may be biased by inclusion, at follow-up, of doctors who had not received the intervention. Thus, future studies should explore the changes in individual doctor prescribing behavior rather than just changes in the overall antimicrobial prescribing rate [48].
Our review demonstrated that education-plus interventions can significantly reduce the antibiotic prescription rate in Chinese outpatient settings. A comprehensive approach, including education, is needed to reduce antibiotic prescribing rates. Further high-quality studies are needed to identify effective interventions to improve AMS and reduce misuse of antibiotics in China.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/antibiotics11060791/s1. Table S1: PRISMA checklist of the systematic reviews and meta-analysis; Table S2: Search terms and search strategies in databases; Table S3: Criteria for the inappropriateness of antibiotic prescription; Figure S1: Risk of bias graph for studies included; Figure S2: Leave one out analysis of the effect of educational interventions on antibiotic prescription rates; Figure S3: Effect of the educational intervention on the antibiotic injection rates; Figure S4: Effect of the educational intervention on the multiple antibiotic rates; Figure S5: Effect of the educational intervention on the inappropriateness of antibiotic prescription.  The funding sources had no role in the study design; the collection, analysis, and interpretation of data; the writing of the report; or the decision to submit the article for publication.