Antibiotic resistance (ABR) is an established threat to global health. It is estimated that 700 000 people die annually as a result of ABR, and this figure is set to rise to 10 million by 2050 if no action is taken. [1] Inappropriate prescribing behaviours by clinicians have been implicated as a major contributing factor driving ABR worldwide, [2] while the judicious use of antibiotics may delay the emergence of resistant bacteria and possibly facilitate a return of susceptible strains. Antibiotic stewardship (AS), which includes measures to reduce inappropriate antibiotic prescribing, is a key intervention to improve prescribing practices in ambulatory care, [3] and AS programmes have been shown to reduce the incidence of infection and colonisation with antibiotic-resistant bacteria. [4]

Antibiotic prescription usually involves an individual consultation between a patient and a clinician, so individual clinician prescribing behaviour is critical. Psychological theories illustrate that clinicians’ knowledge and beliefs influence this process. [5] Specifically, disease knowledge, beliefs about the consequences of prescribing decisions and perceived patient expectations all influence decision-making at clinical consultations. [6]

Numerous studies have assessed clinicians’ knowledge of and beliefs about ABR, with a recent systematic review identifying 57 studies. [7] However, the majority of studies have been conducted in high-resource settings, and only three studies were identified from Africa and one from South Africa (SA).

Objectives

To determine knowledge of, attitudes to and perceptions of ABR among primary care prescribers in SA.

Methods

From October 2015 to December 2016, we conducted a cross- sectional survey of primary care prescribers in SA using a 14-point self-administered questionnaire. The questionnaire (Appendix 1) recorded basic demographic information, perceptions of antibiotic use and resistance and sources of current education about antibiotic use. Knowledge about antibiotic use was tested by seven multiple- choice questions. The structure and content of the survey were based on those used in previously conducted studies [7] and adapted for our setting after discussion with infectious disease specialists and general practitioners (GPs). Depending on the category, there were either single or multiple answer responses such as agree/disagree/unsure or often/sometimes/never. A convenience sampling approach was used. GPs were invited by Medicross (a primary care provider that manages 92 medical and dental centres across SA) management to participate in the survey following initial consultation with regional GP representatives. An email with a link to a SurveyMonkey (https://www.surveymonkey. com; SurveyMonkey Inc., USA) questionnaire was distributed to all Medicross facilities and two reminders were sent out during the course of the study. Paper questionnaires were also distributed to GPs at three conferences held between December 2015 and December 2016. Each conference was attended by ~250 GPs. Questionnaires were mostly completed at the conference, with the remainder returned by email within 2 weeks.

Ethics approval

Ethics approval for the study was obtained from the University of Cape Town Human Ethics Research Committee (ref. no. 722/2016). Data analysis

A descriptive analysis was completed, and knowledge scores were calculated out of the number of questions answered by each participant. Knowledge scores for prescribers were skewed, so medians and interquartile ranges (IQRs) are reported. Prescribers’ self-reported frequency of prescribing antibiotics was tested for association with knowledge scores using the Kruskal-Wallis test. Associations between knowledge scores and beliefs and behaviours were tested using Wilcoxon rank-sum tests and explored using logistic regression models. Analysis was conducted using Stata 14 (StataCorp, USA).

Results

Study population

A total of 264 prescribers completed the survey. Their demographics are shown in Table 1: 154/244 (63.1%) were aged >45 years (20.9% aged 45 - 55 and 42.2% aged >55), 238/242 (98.3%) were doctors, and the majority (207/244, 84.8%) practised exclusively in the private sector. Attitudes and perceptions

Fig. 1 illustrates specific prescriber beliefs. Of the prescribers, 97.1% (234/241) believed that antibiotics are overused in SA, 95.8% (230/240) believed that ABR is a big problem in SA, and 66.5% (157/236) felt pressure from patients to prescribe antibiotics; 29.6% (75/253) would prescribe antibiotics half of the time if a patient expected an antibiotic but the prescriber did not think it is necessary, while 48.0% (118/246) thought that other doctors often (over 70% of the time) prescribed antibiotics when they are not absolutely necessary. When respondents had prescribed antibiotics that were not necessary, the main reasons were pressure from patients ( n =108, 40.9%) and that the patients could not afford laboratory tests ( n =107, 40.5%).

Educational resources

Many prescribers (210/240, 87.5%) wanted more education on the appropriate use of antibiotics, 38.7% (91/235) used guidelines to guide prescription behaviours, and 28.6% (67/234) relied on tablets and smartphones and 27.6% (66/239) on academic journals. The majority (226/235, 96.2%) requested data on local resistance patterns, and 90.4% (208/230) requested education resource aids for discussions on ABR with patients. Prescribers stated that they would value clearer guidelines to improve their antibiotic prescribing in hard copy (80.2%), on smart phone apps (79.7%) or on interactive internet platforms (74.7%).

Knowledge

The median knowledge score for prescribers was 5 (out of a maxi- mum of 7, IQR 4 - 6) (Fig. 2). All the questions were correctly answered by 53 prescribers (20.1%), and one prescriber (0.4%) answered all the questions incorrectly. Prescribers aged >55 years had a lower median knowledge score (4.6, IQR 3.5 - 5.8) than younger prescribers (5.8, IQR 4.6 - 7; p =0.0001) (Table 2).

Self-reported prescribing of antibiotics when not necessary was associated with lower knowledge scores ( p =0.005) (Table 3).

Several prescribing behaviours and beliefs were associated with increased knowledge scores. After adjusting for age and type of practitioner (nurse or doctor), a 1-unit increase in knowledge score was associated with an increase in odds of believing that narrow- spectrum antibiotics should be used instead of broad-spectrum antibiotics where possible (adjusted odds ratio (aOR) 1.29, 95% confi dence interval (CI) 1 - 1.65), that explaining to patients features that, if they develop, should prompt them to seek further medical assistance when antibiotics are not necessary (aOR 1.47, 95% CI 1.058 - 2.04), and that explaining what to expect with regard to duration of symptoms (aOR 122, 95% CI 0.98 - 1.52) were alternative strategies to use instead of prescribing. Those with higher knowledge scores were less likely to report that antibiotics cannot harm the patient if they are not needed, so they prescribe when not necessary (aOR 0.57, 95% CI 0.38 - 0.84; p =0.006) (Table 4).

Discussion

Our study shows that 97.1% of primary care prescribers believe that antibiotics are overused in SA, while 95.8% believe that ABR is a significant problem and 66.5% feel pressure from patients to prescribe antibiotics. We found that knowledge scores among prescribers were suboptimal (median 5/7) and were highest in younger respondents. A novel finding is that knowledge scores are associated with prescribing behaviours. Higher knowledge scores were associated with using narrow-spectrum antibiotics, understanding the potential harm antibiotics can cause, trying alternative strategies such as explaining to patients the disease features that should prompt follow-up, and explaining duration of symptoms.

Our results are consistent with the findings of a large systematic review of clinicians’ knowledge and beliefs about ABR. [7] Similarities include prescribers feeling that ABR is a significant problem (87% compared with 95.8%), and that prescribers feel ABR is driven by patient non-adherence (96% compared with 87.4%) and excessive use (99% compared with 97.1%). [7,8] Our study showed that 66.5% of prescribers felt pressure from patients to prescribe antibiotics, which is similar to an Australian-based study in which 37% of respondents stated they would prescribe antibiotics to meet a patient’s expectations, [9] and a qualitative study which showed that prescriptions were often given to maintain good relationships with patients. [10] An SA study showed that patients’ expectations of antibiotics are strongly associated with prescribing behaviours [11] and demonstrated that patients who are aware of the dangers of ABR expect antibiotics significantly less than those who are not.

A potential reason for older prescribers having lower knowledge scores could be lack of teaching on ABR at the time of their undergraduate medical training compared with more recent graduates. However, a recent study found that only 33% of final-year medical students in SA felt confident to prescribe antibiotics, and 95% stated they would like more information on appropriate use of antibiotics, suggesting that improvements in undergraduate training in ABR are required. [12]

The finding that knowledge scores are associated with positive prescribing behaviours suggests that improved education and training is likely to be an important intervention to address ABR in primary care prescribers. Many prescribers ( n =210, 87.5%) wished to have more education on the appropriate use of antibiotics and would value clearer guidelines to improve their antibiotic prescribing in hard copy (80.2%), on smart phone apps (79.7%) or on interactive internet platforms (74.7%). Respondents in a similar study who used guidelines scored higher on knowledge scores compared with those who did not use guidelines. [12] There are therefore multiple media that could be used to provide education. Our findings suggest that intervention strategies could focus on prescribers in the over-55 age group, and efforts should focus on patient education and patient- prescriber relationships to try to minimise the pressure placed on prescribers to prescribe when not necessary.

Interventions can include developing communication aids for prescriber/patient interactions, as 90.4% of prescribers requested education resource aids for discussions on ABR with patients. This would assist with issues of confusion when explaining the difference between viral and bacterial infections that have arisen in similar studies. [10] The majority (96.2%) of prescribers requested data on local resistance patterns, which opens up an opportunity to improve prescribing by introducing systems to supply up-to-date data on these. Educational programmes for prescribers have been shown to be successful in a range of settings [13] and could be an effective intervention in SA. Further research on the most effective tools should be conducted, and proven methods should then be implemented. These suggestions show useful pathways that could be used to educate both patients and prescribers on the damage that can be caused by overprescription, and on the ABR crisis. Interventions should focus on increasing knowledge about ABR, which could then have an impact on the attitudes, perceptions and behaviours of prescribers.

Study limitations

Our study has several limitations, some of which may limit its generalisability. The use of convenience sampling may have introduced selection bias in favour of prescribers who were more interested in ABR and felt more confident in their knowledge. This is likely to have led to an over-estimate of true knowledge scores. Our sample was almost exclusively of prescribers working in the private sector, which may have important differences from the public sector in SA; in particular, remuneration of private prescribers is often determined by patient numbers, which may introduce a motivation to adhere to patient expectations in order to encourage return visits. A large proportion of antibiotic prescribers in the public sector are nurses, and our sample size of nurse prescribers is too small to draw accurate conclusions about this group. Social desirability bias could be introduced, as the accuracy of answers could not be assessed because they are self-reported, but this was minimised as the surveys were anonymous. Further research using alternative sampling methods is needed to remove the effects of these biases and confounders.

Conclusions

Our study demonstrates gaps in prescriber knowledge that are associated with potentially harmful perceptions and prescribing behaviours. These associations, together with our finding that prescribers would like more education on ABR, suggest that educational tools and patient-provider communication tools could promote rational antibiotic use.

Acknowledgements. The authors thank the Federation of Infectious Diseases Societies of Southern Africa (FIDSSA) and GlaxoSmithKline (GSK) for their support of this project.

Author contributions. EF: first author, substantial contribution to analysis and interpretation of data and the drafting of important scientific content; AS: supervisor, substantial contribution to the analysis and interpretation of data and the drafting of important scientific content; M-AD: supervisor, substantial contribution to analysis and interpretation of data and the drafting of important scientific content; MG: substantial contribution to data collection and approved the final manuscript; DvdB: substantial contribution to data collection and approved the final manuscript; THB: supervisor, substantial contribution to conceptualisation, design, analysis and interpretation of data and the drafting of important scientific content.

Funding.  This work was supported by an unrestricted grant from the FIDSSA and GSK. The funders played no role in the design, conduct or reporting of the study.

Conflicts of interest.  None.