The impact of a multi-faceted intervention on non-prescription dispensing of antibiotics by urban community pharmacies in Indonesia: a mixed methods evaluation

Design

PINTAR (Protecting Indonesia from the Threat of Antibiotic Resistance) was a controlled, pre-post quasi-experimental study conducted from May 2021 to September 2022. Preintervention data collection took place in November 2021, which was after the second wave of the COVID-19 pandemic in mid-2021. The postintervention data was collected in September 2022. During the course of the intervention, the number of new COVID-19 cases increased slightly in February 2022 (see online supplemental appendix 1).

Setting

This study was conducted in Semarang, a city located in Central Java Province in Indonesia with a population of around 1.7 million28 (online supplemental appendix 2). In 2021, the city had 279 registered community pharmacies consisting of 214 independent pharmacies and 65 chain pharmacies. An independent pharmacy is a pharmacy which is privately owned and managed by a pharmacist or a small business owner.29 Chain pharmacies are usually operated under private or state-owned companies and follow a centralised management franchise system.30 In Indonesia, pharmacies can be attached to general practitioners (GPs) or specialist private practices and/or clinical laboratories located in the same compound.31

Participants and recruitment

The primary unit of intervention in this study was the community pharmacy with pharmacists-in-charge as the participants. All community pharmacies in Semarang were invited to participate in the intervention and identified either through the city health office registry or by staff in the city health office. Advertisements about the study were shared with pharmacists-in-charge of each outlet via a WhatsApp group managed by the Indonesian Pharmacist Association (IAI) of Semarang City. Online information sessions were also held with pharmacists, pharmacy owners and branch managers of chain pharmacies in Semarang to provide more detailed information about the study.

Recruitment of participants took place from May 2021 to January 2022. Invitation letters, study information sheets and consent forms were sent by the research team to all pharmacists-in-charge via the WhatsApp group. Those who responded to the invitation were telephoned or messaged by the research team who answered any remaining questions. Scanned copies of all consent forms were sent to the research team via WhatsApp and hard copies were collected 1 week before the intervention commenced. Consenting pharmacists from each community pharmacy were assigned to the ‘participating group’. Those who refused to participate or did not respond after two reminders, were assigned to the ‘non-participating group’. Flowchart of recruitment is shown in online supplemental appendix 3.

Sample size

A sample size of 80 community pharmacies, hence 80 community pharmacists, in each group was required to detect an overall reduction in non-prescription antibiotic dispensing from 50% to 27% among child diarrhoea cases, 74% to 51% among urinary tract infections (UTI) and from 56% to 33% among upper respiratory tract infections (URTI) (80% power, two-sided alpha=5%). These prevalence rates were reported in our previous SP surveys of non-prescription antibiotic dispensing in Indonesia.8

Intervention

The intervention was primarily delivered online via the WhatsApp messaging application, which is the most popular messaging application in Indonesia.32 Based on the location of the pharmacies, participants were assigned to WhatsApp groups, each with 8–9 members, facilitated by a trained senior pharmacist and one administrator from the research team. These groups remained the same throughout the intervention. The four intervention components are detailed in online supplemental appendix 4. Each component was delivered sequentially over a 7-month period (online supplemental appendix 5).

Assessment and outcome evaluation

Demographic data collected at baseline from each participating pharmacist included: age; sex; educational level; years of experience as a community pharmacist; date operational license was issued; size of the pharmacy (ie, number of staff); and whether a GP or specialist practice was linked to the pharmacy.

Data management and analysis

Data from the KOBO Toolbox were downloaded into Microsoft Excel and exported to STATA V.14 (StataCorp, College Station, TX, USA). Descriptive statistics were used to analyse rates of antibiotic dispensing and characteristics of participating pharmacies and pharmacists. Mean and medians were used to describe continuous variables depending on normality of the distribution. Differences in rates of non-prescription antibiotic dispensing were compared between non-participating and participating groups using χ² tests. Given the clinical scenarios are not independent observations (ie, applied to the same pharmacy), a random effects model was selected to evaluate the intervention.38 39 Multivariate random effects logistic regression was used to determine predictors of non-prescription antibiotic dispensing among all participants at postintervention. Predictors included the type of clinical case as well as pharmacist and pharmacy characteristics.27 40 41 All variables in the univariate analysis with a p value less than 0.05 were included in the multivariate analysis.

For the qualitative component, all interviews were transcribed verbatim by the interviewers and then translated from Bahasa Indonesia into English for analysis. Coding and categorisation using an inductive approach42 were performed independently by two researchers (AF and MH) using OpenCode software (University of Umeå, Sweden). Coding schemes were compared, with any discrepancies in codes resolved through discussion until a consensus was reached. Periodic debriefing with the larger research team was also undertaken to ensure rigour in the qualitative analysis.43 When no new themes were appearing, it was concluded that data saturation had been reached.

Characteristics of pharmacies and pharmacists

Of the total 270 pharmacies in Semarang, 81 consented to take part in the intervention. The characteristics of the pharmacies are displayed in table 1, showing that there was a higher proportion of chain pharmacies and pharmacies attached to GP/specialist clinics in the participating group than in the non-participating group.

A total of 95 pharmacists were enrolled across the 81 participating pharmacies (results not shown here). One pharmacy and two pharmacists dropped out of the study due to time constraints. Therefore, a total of 93 pharmacists from 80 pharmacies completed the intervention (online supplemental appendix 2). Of these, 80 participating pharmacists were female and 13 male with a median age of 33 years (IQR=8). The median number of years working as a community pharmacist was 8 years (IQR=7) and the median number of years working in their current pharmacy was 4 years (IQR=6). One-third of pharmacists worked for a chain pharmacy.

At preintervention, the rate of antibiotic dispensing without a prescription in the participating group was significantly lower compared with the non-participating group (76.3% vs 84.6%, respectively, p value=0.02). At postintervention, the rate of antibiotic dispensed without a prescription in the participating group was 55.4% compared with 82.3% in the non-participating group, that is, a 26.9% difference (p value <0.001). Between preintervention and postintervention, the rate of antibiotic dispensed without a prescription in the participating group fell by 20.8% (76.3%–55.4%) compared with 2.3% (84.6%–82.3%) in the non-participating group (p value <0.001) (figure 1).

Request permission

For all clinical scenarios at postintervention, the rate of non-prescription antibiotic dispensing in the participating group was significantly lower than for the non-participating group (table 2). The largest difference between participating and non-participating groups in the rate of non-prescription antibiotic dispensing was 42.4% for the URTI scenario (41.3% in the participating group vs 83.7% in the non-participating group; p value <0.001), followed by 23.3% for the diarrhoea scenario (52.5% in the participating group vs 75.8% in the non-participating group; p value <0.001) and finally, 14.9% for the UTI scenario (72.5% in the participating group vs 87.4% in the non-participating group; p value=0.003); all differences were statistically significant.

Meanwhile, the reduction in the dispensing rate of non-prescription antibiotics for the participating and non-participating groups between preintervention and postintervention are as follows: URTI 38.7% versus 4.2% (p value <0.001); child diarrhoea 15.0% versus 1.6% (p value <0.001); and UTI 8.8% versus 1.0% (p value=0.003).

Predictors of dispensing antibiotics without a prescription at postintervention

Postintervention, the univariate analysis also showed that the likelihood of dispensing an antibiotic without a prescription was lower in the participating group and in pharmacies attached to a GP/specialist clinic. In contrast, dispensing an antibiotic without a prescription was more likely in pharmacies that dispensed antibiotics without a prescription at preintervention, in pharmacies attended by older pharmacy staff, where SPs were female, and for the UTI scenario. The multivariate analysis showed that the likelihood of dispensing an antibiotic without a prescription was less likely in the participating group than the non-participating group but more likely in pharmacies that dispensed antibiotics without a prescription at preintervention and for the UTI scenario (table 3). Subgroup analysis of chain and independent pharmacies also showed that being in the participant group resulted in lower odds of dispensing antibiotics without prescription after adjusting for other covariates (results are not shown).

Types of antibiotics dispensed

Figure 2 shows the change in the proportion of antibiotics dispensed between the participating group and non-participating group at preintervention and postintervention. Between preintervention and postintervention, an increase in the use of amoxicillin was seen in both groups which was probably influenced by the SP request, but the increase was larger in the non-participating group. The use of fradiomycin/gramicidin decreased in both groups while the use of pipemidic acid increased in the participating group but remained the same in the non-participating group. A decline in the use of nifuroxazide was evident in both groups.

Request permission

Reasons for dispensing antibiotics without a prescription

Despite the significant decline in the rate of non-prescription antibiotic dispensing, our SP survey showed that around 50% of participating pharmacists did not change their behaviour. Our interviews revealed several reasons for this. First, pharmacists were fearful of losing customers who would then choose an alternative pharmacy. One owner mentioned, ‘We have experienced that if we refuse to give antibiotics without prescription, we lost customers’ (205-3, pharmacy owner).

For this reason, requests for non-prescription antibiotics were not refused outright but often partly met by staff offering a smaller dosage than recommended. One pharmacy staff member mentioned, ‘If I stop (dispensing antibiotics) then it will affect sales, so sometimes I offer small doses, for example, 1–2 blisters’ (386-2, pharmacy staff).

While many pharmacists realised that giving non-prescription antibiotics was prohibited, they admitted to succumbing to pressure exerted by pharmacy owners to maximise sales: ‘Well … I always say [to customers] they must come with a prescription, but my boss [the owner] objects. One time there was a customer who insisted and debated with me. Eventually, I got a warning—[the owner] sent me a warning letter’ (278-1, pharmacist). Another interviewee said the owner gave them a particularly harsh warning: ‘… if you don’t give antibiotics, you will receive no salary!’ (403-1, pharmacist).

Interviewees reported that it was common for customers who were initially refused an antibiotic to push pharmacy staff to give them antibiotics. ‘Sometimes I would offer them other alternative, but if they say, ‘I want amoxicillin’, they would refuse other drugs’ (396-2, pharmacy staff). Interviewees also noted that some customers would bargain for lower dosages of antibiotics, send another family member to try and buy antibiotics without a prescription, or ask for antibiotic lozenges as a ‘milder’ form of antibiotic.

Some interviewees stated they were more inclined to give non-prescription antibiotics to customers showing signs of infection, needing antibiotics for children or older people, with a history of using the same antibiotic, employed as a health worker or claiming to have an old prescription at home. ‘If they said that they have been prescribed before, I would give them [antibiotics]. But I would also mention that next time they should see a doctor first. Usually they argue that it is too far or closed on Sundays’ (338-1, pharmacist).

Another reason given for dispensing non-prescription antibiotics was a lack of oversight from a qualified pharmacist. Despite regulations stipulating that antibiotics can only be dispensed by a pharmacist on presentation of a prescription, this was not always the case. A representative from the Ministry of Health said: ‘For pharmacists, the big problem [in the pharmacy] is attendance. The attendance of pharmacists is not that good. There is an impression that pharmacists are not always present’ (Ministry of Health staff).

Finally, our intervention focused primarily on pharmacists-in-charge who were subsequently responsible for sharing materials and learnings from the educational sessions with colleagues in their workplace. Some pharmacy technicians and other attending staff admitted they had a limited knowledge of guidelines and regulations on antibiotic use. One interviewee stated, ‘I did not know any regulation on antibiotics. Or maybe I was not up to date. Because the regulation about antibiotics is not really clear, unlike [the regulation on] psychotropic agents’ (386-2, pharmacy staff).

Strengths and limitations

A major strength of the intervention and the study design was the way it was embedded into the real-world activities of pharmacies and city health office teams responsible for supervising and monitoring community pharmacies. For example, the accreditation component of the intervention was linked to the continuing professional education scheme for community pharmacists in Indonesia.56 The use of the SP methodology was another strength of the study. This approach is widely referred to as the ‘gold standard’ for measuring the quality of healthcare due to its ability to avoid typical biases or confounding issues often associated with alternative measures such as interviews and vignettes.27 57 58 In contrast, a key limitation of this study was that only around one-third of all pharmacies in Semarang agreed to take part in the intervention. This may have partly been due to a tradition of limited engagement with community pharmacies by local health authorities,59 exacerbated during the recent COVID-19 pandemic.31 From a methodological viewpoint, it could also be argued that the 7-month evaluation period and the one-month interval between completion of the intervention and the SP survey was too short to determine whether the improvements observed are sustainable. Spill-over effects from participating pharmacists to non-participating pharmacists cannot be ruled out - these could result from the unintentional dissemination of new knowledge through existing Whatsapp groups or local events run by pharmacist associations. To address this concern, rules for sharing materials were established at the start of the intervention when it was also noted that all participants would have access to the intervention materials at the end of the study. Measurement of spill-over effects was not within the scope of the study and could be a topic for future research. Finally, while quasi-experimental designs used in real-world settings tend to have higher external validity,60 the risk of selection bias and its impact on internal validity cannot be excluded.