Problem

Upper respiratory tract infection (URTI) is a common presentation in primary care, accounting for up to 10% of presentations.1 The use of antibiotics in nasopharyngitis, common cold and non-specific URTI does not improve outcomes; in fact, patients report increased adverse events compared with placebo.2 Thus, the majority of patients with URTI do not require antibiotics. Overuse of antibiotics contributes to the emergence of antibiotic resistance.3 4 Despite the risk to individual and population health, inappropriate antibiotic use is widespread in URTI.5–7

The outreach clinics of a non-governmental organisation provide free healthcare to disadvantaged people living in Kolkata and West Bengal, serving homeless and slum communities. Anyone living below the poverty line is entitled to free medical care and may be entitled to benefits packages, for example, food supplies. The organisation employs 10 full-time doctors. Patients self-refer to clinics or are referred by government hospitals for extra support. Patients often present with complex medical and social needs, creating a pressured working environment. The outreach clinics are routinely staffed by a doctor, a pharmacist and support healthcare workers, with an average attendance of 30 patients per half-day session. URTI is a common presentation in the outreach clinics; on average, 100 patients present with URTI each month.

A baseline audit of 222 patients presenting with non-specific URTI over 8 weeks found antibiotics were prescribed in 62.6% of cases. This project aimed to reduce antibiotic prescribing in non-specific URTI from 62.6% to 30% in 4 months.

Background

Antibiotic resistance is a matter of global concern; increasing antibiotic use drives the emergence of resistance. Hence, as well as risking unwarranted side effects to the individual patient (such as drug side effects), inappropriate use of antibiotics also poses risks to community and public health.8 Patients treated with antibiotics for respiratory infections in primary care develop bacterial resistance to that antibiotic. This can persist for up to 12 months, increasing the prevalence of resistant organisms in the community.9 In developed countries, 80%–90% of antibiotics are prescribed in primary care settings; the majority of these are prescribed for respiratory tract infections.10 11 The rates of inappropriate antibiotic prescription are worryingly high; a major study of US ambulatory care visits estimated a third of all antibiotics prescribed were inappropriate.7

Antibiotic prescribing in primary care varies markedly within European countries. In 2015, antibiotic prescribing ranged from 10.7 defined daily dose per 1000 inhabitants per day in the Netherlands to 36.1 in Greece.12 In Dutch primary care, 19% of patients presenting with URTI in 2010 were prescribed antibiotics.13 In the UK, the median general practice in a large study prescribed antibiotics at 38% of consultations for URTI.14 Antibiotic prescribing for URTI in primary care in developing and transitional countries has increased from rates of approximately 44% in 1982 to 71% in 2006.15 In China and Thailand, over 80% were prescribed antibiotics.16 17 Pooled data for India shows antibiotic prescribing in URTI of over 70%.18

Why do doctors prescribe antibiotics inappropriately? Doctors’ knowledge of antibiotic prescribing guidelines and the risks of antimicrobial resistance may be inadequate.19–23 Older GPs are less likely to prescribe according to antibiotic guidelines, suggesting difficulty changing established practice in light of new evidence.24 Doctors may disagree with evidence-based guidelines, instead relying on their own clinical experience to guide them25; for example, doctors in India report using antibiotics as prevention in patients from poorer communities due to concerns about higher prevalence of bacterial infection.26 Barriers towards putting evidence into practice include time pressure, financial considerations and doctors’ perceptions of patients’ expectations.26–30

A number of randomised controlled trials in primary care in Europe have shown significant reductions in antibiotic prescribing (both all cause and specifically in URTI) through programmes comprising educational events and feedback on practice prescribing habits.31–33 For example, the Stemming the Tide of Antibiotic Resistance (STAR) programme, consisting of prescribing feedback, seminars and online education, led to a 4.2% reduction in all-cause antibiotic prescribing without increased hospital admissions or reconsultations in a Welsh primary care setting.33

Interventions that use a number of different approaches to improve antibiotic prescribing tend to be most successful. Audit and feedback alone generally lead to small improvements in professional practice.34 Interactive educational sessions are more effective than lectures.35

Measurement

Initial data collection reviewed all cases coded as upper respiratory tract infection in the outreach primary care clinics over 8 weeks in November and December 2016. The coding for URTI was a generic code with no subdivisions. Ear infections were coded differently and were not included. All notes are paper based and were reviewed manually.

In case notes audited, the predominant symptoms were cough, cold and runny nose. There were no positive examination findings apart from cervical lymphadenopathy and coryza. Some notes did not record symptoms but simply documented ‘URTI’; these were included. Case notes were excluded from analysis when the coding was judged to be incorrect (eg, patients with cough and fever with focal chest findings on examination). Inclusion criteria were developed to select a patient population in which antibiotics were not indicated. Therefore, cases where likely diagnosis might warrant antibiotic use (eg, tonsillitis, sinusitis) were excluded; these cases comprised 9 out of 231 patients coded as URTI. An Excel spreadsheet was used to store and analyse the data collected.

Subsequent audits used the same criteria but were carried out over 4-week periods. The coding system was updated in the second Plan, Do, Study, Act (PDSA) cycle. Following this, audit included patients coded as common cold or non-specific URTI.

In November and December 2016, 222 patients meeting inclusion criteria were seen. Antibiotics were prescribed in 62.6%. Two antibiotics were prescribed in 1.8%. Ten different antibiotics were prescribed; the most common being amoxicillin (31.7%), cefadroxil (23.7%) and co-trimoxazole (12.9%). Examination findings were documented in 47.3% of patients. Supportive treatment (such as paracetamol, cough syrup and antihistamines) was prescribed in 87% of patients.

Design

A multifaceted intervention to reduce inappropriate antibiotic prescribing was planned. This comprised audit prescribing feedback, educational seminars, one-on-one case-based discussion, guideline development and coding alterations.

Pooled audit feedback on URTI prescribing was presented at pre-established monthly doctors’ meetings, attended by the organisation’s doctors and chaired by the deputy CEO. Giving individual as well as pooled prescribing feedback was considered, but it was felt that this could alienate staff and reduce engagement. Furthermore, pooled feedback might motivate doctors who were prescribing at higher rates to modify their practice in line with colleagues. Doctors and medical management gathered around a table; PowerPoint was used for presentations.

Three seminars (1 hour) combining PowerPoint and round table discussion were planned for the doctors’ meetings to provide the evidence for antibiotic use in URTI and to encourage interactive discussion around the rationale for antibiotic prescribing. The hypothesis was that eliciting doctors’ views in a supportive environment would help to generate open debate. Allowing staff to share their prescribing motivations without censure made it possible to understand why inappropriate antibiotic prescribing was taking place. Meetings were not taped or transcribed; however, issues raised by staff was recorded during the meetings. Staff feedback was used to tailor training sessions around their specific issues and concerns. Staff engagement was critical to the long-term sustainability of this project.

One-on-one case-based discussions in the clinics were planned, as it was anticipated that these would yield more specific information on the reasons behind inappropriate antibiotic use and give an opportunity to suggest alternative management strategies. These were delivered in all outreach clinics on an informal basis (weekly–fortnightly) prompted by doctors’ queries or quality improvement team concerns around inappropriate prescribing identified in consultations.

The development of antibiotic guidelines for the organisation was planned; these were approved by management and frontline staff. This aimed to increase the sustainability of the process by providing agreed references for best practice. An update to the coding system was used to differentiate within URTI and encourage more focused diagnosis and thus decision making and management. This was done in collaboration with staff and management and subdivided URTI from a generic category into specific subsections.

Strategy

The PDSA model was used to trial interventions and assess their impact on reducing inappropriate antibiotic prescribing in URTI.

Lessons and limitations

This project aim was to reduce inappropriate antibiotic prescribing in URTI by approximately half, from 62.6% to 30%, through the use of a multifaceted intervention. The reduction in antibiotic prescribing was dramatic and better than anticipated, with rates of 7.2% in the postintervention audit.

A key strength of the project was the use of staff engagement to drive change. Following baseline audit feedback and discussion, the deputy CEO in charge of the medical projects became a strong advocate for reducing unnecessary antibiotic prescribing in URTI. Her support in conjunction with her authority within the organisation gave increased credibility to the project. The use of interactive sessions allowed staff to discuss their concerns and motivations for antibiotic prescribing in an open environment. This in turn allowed the development of strategies to address specific staff concerns. For example, some doctors said that they prescribed antibiotics as they were worried about missing an underlying bacterial infection. Examining the patient was suggested to help differentiate URTI and pneumonia. During the project, there was a marked rise in documentation of examinations, suggesting that this was a strategy that doctors found helpful.

The increase in examinations documented may have improved diagnostic accuracy (assuming the change is due to a real increase in examination rather than an increase in documentation). For example, a patient presenting with cough who had focal chest signs but was not examined in the baseline audit may have been diagnosed as URTI, whereas in the postintervention audit, the patient may have been diagnosed with pneumonia. This would be an unexpected advantage of the project; it is important that serious infections such as pneumonia receive appropriate management and follow-up.

The focus on common cold and non-specific URTI was both a strength and a limitation of the project. Following initial training sessions and audit feedback, it was possible to obtain group consensus that antibiotics were not appropriate in these patients. However, the majority of doctors felt that the use of antibiotics in sore throat and sinusitis was almost always indicated; they felt that the evidence put forward did not apply to their patient populations. The narrower focus in this intervention made it easier to deliver and reinforce clear messages. However, further work on improving antibiotic prescribing in tonsillitis and sinusitis would be beneficial.

Multiple interventions were trialled concurrently. Previous studies have shown multifaceted interventions to be effective. It is likely that the interventions had a synergistic impact and that the concurrent use of interventions led to the more rapid change seen in the second PDSA cycle. However, this design means it is not possible to analyse the impact of any one intervention.

A major potential limitation of the study is the impact of the Hawthorne effect. The very fact of observing prescribing behaviour can increase physician compliance with guidelines.36 To avoid this effect, a control group would also be needed. Since paper notes were analysed retrospectively, the audit process itself would not have affected the baseline data. After this, staff were aware of the audit process. Given that this was the first project looking at antibiotic prescribing in the organisation, this alone may have had a dramatic impact on the effect size of the interventions.

A new doctor was recruited between baseline data collection and the initial PDSA cycle. Their prescribing practice may have affected changes seen between the December–November baseline audit and the January–February audit. As prescribing data were pooled, it is not possible to determine the extent and direction of this effect.

The project relied on changing staff prescribing patterns, largely through training and audit feedback; this is a potential limitation. This process was not routine in the organisation before the project. Its importance has been recognised by medical management, who are driving future audit projects to improve service quality. An ongoing process of audit and feedback is likely to be crucial to making these changes sustainable long term.

In future, it would be helpful to think about introducing more system changes to support these changes, such as enabling the pharmacy team to directly query inappropriate antibiotic requests.

Seasonal variation may impact the results: the number of patients presenting with non-specific URTI decreased between the initial 8 weeks audit and the final audit. The time of year may influence doctors’ perceptions of risks of bacterial infection and thus impact prescribing practice.

It was not possible to collect all relevant data to assess the broader impact of the intervention. Although the organisation provides a primary care service, there is no notification by hospitals or other healthcare providers of patient attendances. Therefore, it was not possible to collect data on hospital presentations or admissions during this study. The ultimate goal of reducing inappropriate antibiotic use is to reduce the risk of antimicrobial resistance and to prevent unnecessary side effects. It was beyond the scope of this study to examine the impact on antimicrobial resistance in bacterial isolates.