Prescribing Antibiotics in Public Primary Care Clinics in Singapore: A Retrospective Cohort Study

Background: Antibiotic prescription practices in primary care in Singapore have received little scholarly attention. In this study, we ascertained prescription prevalence and identified care gaps and predisposing factors. Methods: A retrospective study was conducted on adults (>21 years old) at six public primary care clinics in Singapore. Prescriptions >14 days were excluded. Descriptive statistics were used to showcase the prevalence data. We used chi-square and logistic regression analyses to identify the factors affecting care gaps. Results: A total of 141,944 (4.33%) oral and 108,357 (3.31%) topical antibiotics were prescribed for 3,278,562 visits from 2018 to 2021. There was a significant reduction in prescriptions (p < 0.01) before and after the pandemic, which was attributed to the 84% reduction in prescriptions for respiratory conditions. In 2020 to 2021, oral antibiotics were most prescribed for skin (37.7%), genitourinary (20.2%), and respiratory conditions (10.8%). Antibiotic use in the “Access” group (WHO AWaRe classification) improved from 85.6% (2018) to 92.1% (2021). Areas of improvement included a lack of documentation of reasons for antibiotic use, as well as inappropriate antibiotic prescription for skin conditions. Conclusion: There was a marked reduction in antibiotic prescriptions associated with the onset of the COVID-19 pandemic. Further studies could address the gaps identified here and evaluate private-sector primary care to inform antibiotic guidelines and the local development of stewardship programs.


Introduction
Antimicrobial resistance is widely recognized as a global public health threat [1]. No new classes of antibiotics have been discovered in the past 30 years, and prescription rates are at an all-time global high. This threatens our ability to respond effectively to the global and enduring threat of infectious diseases [2].
Singapore launched its National Strategic Action Plan on Antimicrobial Resistance in 2017 in response to the Global Action Plan for antimicrobial resistance developed by the World Health Assembly [2,3]. This led to the setup of antimicrobial stewardship programs across all public restructured hospitals. A similar initiative, however, is lacking in primary care [4]. Despite electronic prescribing being implemented in most healthcare settings in Singapore, prescription data remains difficult to access, monitor, and regulate [5]. While primary care accounts for 80% of antibiotic prescription in developed countries, 50% of these prescriptions are deemed inappropriate [6]. As healthcare in Singapore reforms towards a population health model called Healthier Singapore [7], this provides an excellent opportunity to launch a primary care antimicrobial stewardship program. Studying the existing data in public primary care institutions could shed light on the current practices, and act as a first step in this momentous push toward appropriate antimicrobial usage in the community.
During the COVID-19 pandemic, changes in patients' behavior in terms of seeking healthcare and physicians' prescription patterns may have affected community antibiotic prescription rates [8]. Several studies that have been performed in developed countries revealed a general trend of a reduction in antibiotic prescription in primary care during the pandemic [9,10]. A local study performed in an inpatient setting showed a reduction in antimicrobial prescriptions in 2020 compared to before the pandemic [11]. To date, no study has been conducted concerning antibiotic prescriptions in primary care in Singapore post-COVID-19 pandemic; hence, there is a need to replicate the abovementioned study in the outpatient primary care setting.
In our study, we aim to examine the current patterns of antibiotic prescriptions for adults in primary care, as well as to identify potential care gaps for improvement, and factors influencing these gaps. We hope that this will pave the way for the development of local antibiotic guidelines within primary care and improve governance and stewardship in the post COVID-19 era, toward the creation of a Healthier Singapore.

Data Source and Study Population
A retrospective observational study was conducted using data extracted from electronic health records (CPSS2 and EPIC) of patients from 6 public primary care clinics (National University Polyclinics) in Singapore, from 2018 to 2021. This included April 2020, which was the peak of the COVID-19 pandemic in Singapore [12]. De-identification was performed by a centralized, trusted third party (institution research office) before passing over to study team for analysis. The study included patients above 21 years of age who visited these 6 clinics and were prescribed an oral or topical antibiotic. Patients on long-term antibiotics for prophylaxis or treatment for more than 14 days were excluded.
Variables included patient demographics, visit diagnoses, presence of chronic diseases, such as diabetes mellitus and chronic kidney disease, antibiotic name and class, and prescriber information, such as place of practice, number of years of practice, training location, and family physician's accreditation status. Visit diagnoses in clinics were coded using the International Classification of Disease (ICD-10). Institution level data on the total number of visits for each visit diagnosis were also collected to determine the antibiotic prescription rate for each condition. For the purposes of this study, each antibiotic prescribed equates to 1 antibiotic prescription, regardless of number of visits.

Diagnosis Categorization and Antibiotic Classifications
To analyze antibiotic prescription by diagnoses, visits prescribed with oral antibiotics were grouped into categories based on the indicated diagnosis. These categories consisted of respiratory, skin, genitourinary, gastrointestinal, infectious disease, and dental conditions. Prescriptions for miscellaneous or chronic disease diagnoses where indications were unable to ascertain were listed as 'Undefined'. The diagnosis categorization was conducted independently by two family physicians based on the World Health Organization (WHO) International Classification of Diseases 10th revision (ICD-10) [13], and split into conditions whereby antibiotics were often required versus not often required. Discrepancies in categorizations were de-conflicted afterwards. For antibiotic classification, we adopted the 2021 WHO AWaRe classification [14].
Often, there were oral antibiotics prescribed for visits with multiple diagnoses. We coded a tiered ranking logic system ( Figure A1) to select infective conditions over noninfective conditions, and prioritized ranking of conditions in terms of which antibiotics were often required, until each antibiotic prescription belonged to only one category (Table A1). For prescriptions that we were unable to determine the indication of from the listed diagnoses (multiple infective conditions or conditions where antibiotics were often required), they were grouped under 'multiple diagnoses'. For visits prescribed with topical antibiotics that were incongruent with the coded diagnosis, we reclassified the diagnosis such that they were prescribed for their indicated conditions and route (i.e., skin topical antibiotics prescribed for skin conditions). In the case of topical ciprofloxacin, which can be used as an eye or ear drop, we differentiated them by the prescribed dosage, duration, and route of application.
Data from 1 clinic were analyzed for the prescription rate of oral antibiotics, but was excluded from other analyses as the clinic was newly built and lacked data before 2021. All antibiotics prescribed by dentists were assumed to be for dental conditions. To ensure data validity and accuracy of ranking classification, 100 case notes were randomly selected and extracted for audits. All information was true and corresponded to the diagnoses and antibiotic characteristics that were extracted. This also ensured and validated the accuracy and robustness of the tiered ranking logic system in diagnosis selection, topical antibiotic diagnosis reclassification, and ciprofloxacin eye and ear drop dichotomization.

Statistical Analysis
Rstudio (R version 4.2.0), IBM SPSS Statistics Version 29.0 and Microsoft Excel 2010 were used in data cleaning and analysis. p-value of <0.05 in the two-sided test was considered statistically significant. Descriptive statistics were performed, and numerical variables were represented as mean with standard deviations, or n (%) for categorical variables. Antibiotic prescription rate was derived by dividing the number of prescriptions over the total number of patient visits. Segmented regression analysis was performed to describe antibiotic prescription trends before and after the peak of the pandemic. Chi-square tests were used for categorical variables (i.e., gender, race, and presence of chronic conditions) while logistic regression was performed for continuous variables (i.e., patient's age and physician's number of years of practice) to analyze antibiotic prescription for undefined conditions, "Watch" group antibiotic prescriptions, topical antibiotic prescriptions with irrelevant diagnoses, and dual antibiotic prescriptions for skin and soft tissue conditions. Subsequently, combined multivariate logistic regression was performed, considering all variables collected on the gaps identified.

Ethical Considerations
The research was conducted in accordance with the Declaration of Helsinki national and institutional standards and approved by the NHG Domain-Specific Review Board (DSRB) on June 2022 (2022/00319).

Results
A total of 141,944 oral and 108,357 topical antibiotics were prescribed for 3,278,562 patient visits from 2018 to 2021, giving an overall prescription rate of 4.33% and 3.31%, respectively. For the purposes of analysis, the antibiotic prescriptions from Clinic F were removed due to its introduction in 2021; despite this, Clinic F's oral antibiotic prescription rate was consistent compared with the other clinics. There was a reduction in the oral antibiotic prescription rate from 5.11% to 3.38% from 2018 to 2021 (Table 1). In particular, we noted a significant reduction in 1926.8 prescriptions (p < 0.01) before and after the peak of the COVID-19 pandemic in Singapore in April 2020 ( Figure 1). The percentages displayed in the top row of Table 1 were achieved by dividing the total number of antibiotic prescriptions over the total number of patient visits for that year. We noted that this reduction in the antibiotic prescription rate was consistent across all age groups, genders, races, and clinics. The oral antibiotic prescription rates were the highest among younger age groups  and females. While the majority of antibiotics were prescribed for those of Chinese ethnicity, they had the lowest oral antibiotic prescription rate per clinic visit. The majority of antibiotics were prescribed by family physicians (58.3%) and overseas trained doctors (63.0%).  Oral antibiotics were most prescribed for respiratory conditions (29.6%), skin and soft tissue conditions (28.9%), and genitourinary conditions (15.2%) ( Table 2). In 2021, skin and soft tissue conditions (37.7%) and genitourinary conditions (20.2%) overtook respiratory conditions to become the top two most common conditions when oral antibiotics were prescribed. This was due to an 84% reduction in respiratory antibiotic prescriptions, with Oral antibiotics were most prescribed for respiratory conditions (29.6%), skin and soft tissue conditions (28.9%), and genitourinary conditions (15.2%) ( Table 2). In 2021, skin and soft tissue conditions (37.7%) and genitourinary conditions (20.2%) overtook respiratory conditions to become the top two most common conditions when oral antibiotics were prescribed. This was due to an 84% reduction in respiratory antibiotic prescriptions, with a 5.22% absolute reduction in respiratory condition visits prescribed with oral antibiotics (Figure 2).
Prescriptions for dental, skin and soft tissue, and ear, nose, and throat (ENT) conditions remained stable from 2018 to 2021 (Table 2). While the absolute number of prescriptions for dental conditions remained low, it had the highest percentage of visits that were prescribed with antibiotics (17.8%). The number of visits with multiple infectious conditions reduced from 3.69% in 2018 to 1.67% in 2021. The number of antibiotics prescribed for undefined conditions (diagnoses listed that were non-infectious in nature, such as chronic diseases) rose from 10.8% to 17.2% in terms of the total antibiotics prescribed across 2018 Antibiotics 2023, 12, 762 6 of 29 to 2021. While the patient's age (OR 1.005, 95% CI 1.004-1.006) was associated with antibiotic prescription for undefined conditions, the physician's years of practice (OR 0.993, 95% CI 0.991-0.995) was found to have an inverse relationship (Table A2). On a multivariate analysis after adjusting for the patient's age and physician's years of practice, the female gender (OR 1.12, 95% CI 1.08-1.15), race (p < 0.001), presence of diabetes mellitus (OR 1.34, 95% CI 1.29-1.40) and chronic kidney disease (OR 1.31, 95% CI 1.26-1.37), place of practice (p < 0.001), and having an accredited family physician (OR 1.16, 95% CI 1.12-1.20) were significantly associated with antibiotic prescriptions for undefined conditions (Table A2).    Prescriptions for dental, skin and soft tissue, and ear, nose, and throat (ENT) conditions remained stable from 2018 to 2021 (Table 2). While the absolute number of prescriptions for dental conditions remained low, it had the highest percentage of visits that were prescribed with antibiotics (17.8%). The number of visits with multiple infectious condi-   Figure 3 describes all the available oral antibiotics split into diagnoses and grouped according to the WHO AWaRE classification. The most common oral antibiotic prescribed from 2018 to 2021 was amoxicillin/clavulanate (58.8%). Skin and soft tissue infections had the highest percentage of antibiotic use in the Access group (98%). The overall increase in the use of antibiotics in the Access group from 85.6% (2018) to 92.1% (2021) was due to the reduction in clarithromycin use, particularly for respiratory conditions. Ciprofloxacin constituted the largest proportion (68%) among the antibiotics used by the Watch group in 2021, of which the majority (70.6%) were prescribed for genitourinary conditions. Ciprofloxacin was 7 and 16 times more likely to be prescribed for genitourinary (OR 7.41, 95% CI 7.05-7.78) and gastrointestinal (OR 16.1, 95% CI 14.9-17.4) conditions, respectively, compared to other conditions. The changes in antibiotic prescription habits observed in 2020 and 2021 prompted us to assess the factors contributing to the prescription of Watch group antibiotics. This is showcased in Table A3. On a multivariate analysis (after adjusting for the patient s age), being male (OR 1.26, 95% CI 1.19-1.35) with gastrointestinal (OR 28.5, 95% CI 24.1-33.8), respiratory (OR 11.9, 95% CI 10.6-13.3), or genitourinary conditions (OR 10.2, 95% CI 9.07-11.4) made one significantly more likely to be prescribed a Watch group antibiotic. Factors such as the physician s years of experience, being local trained (OR 1.22, 95% CI 1.14-1.30), having an accredited family physician (OR 1.17, 95% CI 1.09-1.25), and the place of practice significantly contributed to the Watch group s antibiotic prescriptions (Table A3).
Topical antibiotic prescriptions were highest in the younger age groups (age 22-44), with gradual increments of ENT (0.463% to 0.59%) and skin (1.69% to 1.90%) topical antibiotic prescription rates from 2018 to 2021. This is described in Tables 3-5, respectively. Topical antibiotic prescriptions also differed between clinics. Topical antibiotics for skin conditions also saw the highest prescriptions among patients with diabetes and chronic kidney disease (Table 5). While the number of topical antibiotic prescriptions differed from clinic to clinic from 2018 to 2021, Clinic C had the highest topical eye and skin antibiotic prescription rate (Tables 4 and 5).  The changes in antibiotic prescription habits observed in 2020 and 2021 prompted us to assess the factors contributing to the prescription of Watch group antibiotics. This is showcased in Table A3. On a multivariate analysis (after adjusting for the patient's age), being male (OR 1.26, 95% CI 1.19-1.35) with gastrointestinal (OR 28.5, 95% CI 24.1-33.8), respiratory (OR 11.9, 95% CI 10.6-13.3), or genitourinary conditions (OR 10.2, 95% CI 9.07-11.4) made one significantly more likely to be prescribed a Watch group antibiotic. Factors such as the physician's years of experience, being local trained (OR 1.22, 95% CI 1.14-1.30), having an accredited family physician (OR 1.17, 95% CI 1.09-1.25), and the place of practice significantly contributed to the Watch group's antibiotic prescriptions (Table A3).
Topical antibiotic prescriptions were highest in the younger age groups (age 22-44), with gradual increments of ENT (0.463% to 0.59%) and skin (1.69% to 1.90%) topical antibiotic prescription rates from 2018 to 2021. This is described in Tables 3-5, respectively. Topical antibiotic prescriptions also differed between clinics. Topical antibiotics for skin conditions also saw the highest prescriptions among patients with diabetes and chronic kidney disease (Table 5). While the number of topical antibiotic prescriptions differed from clinic to clinic from 2018 to 2021, Clinic C had the highest topical eye and skin antibiotic prescription rate (Tables 4 and 5).       (Table A5). Factors such as the physician's years of experience, place of practice, being locally trained (OR 1.07, 95% CI 1.03-1.11), and having an accredited family physician (OR 1.10, 95% CI 1.06-1.14) were significantly associated with inappropriate diagnoses coding during topical antibiotic prescriptions (Table A5).
A significant percentage (35.8%) of same-visit prescriptions of oral and topical antibiotics for skin conditions was also observed compared to oral antibiotic prescriptions. Younger patient ages (OR 0.994, 95% CI 0.993-0.995) and higher years of experience of the physician (OR 1.017, 95% CI 1.015-1.019) were associated with dual antibiotic prescriptions (Table A5). In the multivariate analysis, the female gender (OR 1.13, 95% CI 1.08-1.18), diabetes mellitus (OR 1.06, 95% 1.001-1.11), and the absence of chronic kidney disease (OR 1.11, 95% CI 1.05-1.17) were significant predictors for dual antibiotic prescriptions (Table A6). Factors such as the physician's years of experience, being overseas trained (OR 1.18, 95% CI 1.13-1.24), having an family physician accredited (OR 1.16, 95% CI 1.11-1.21), and the place of practice were significantly associated with dual antibiotic prescriptions (Table A6).

Discussion
This observational study is one of the first conducted on oral and topical antibiotic use within primary care clinics in Singapore, showcasing the prevalence of prescription practices and revealing the care gaps. All prescription data were included as they were extracted from a health records database, with no missing data. Diagnoses were mapped to prescription data, with prescription rates calculated based on the overall patient visits, which showcased the actual burden of antibiotic use. Some physician and patient factors affecting antibiotic prescription were also included and analyzed, with meaningful and applicable results. However, this was not an exhaustive list; future studies should focus on this area to help build a more complete picture.
The antibiotic prescription rate reduced with age, which was likely due to a higher proportion of older patients attending for chronic disease visits compared to younger patients. This could also be due to poorer knowledge associated with younger patients in Singapore, leading to more presentations and antibiotic requests [15]. Notably, overall, oral antibiotic prescriptions reduced at a greater proportion compared to visits for respiratory conditions from 2018 to 2021. This was also observed in many countries worldwide [8,10,16,17]. The segmented regression analysis performed showed a significant reduction in antibiotic prescriptions after the peak of the COVID-19 pandemic in April 2020, which was consistent with a previous inpatient local study [11]. This demonstrates that this was due to public health measures, which influenced both outpatient and inpatient antibiotic prescriptions. While previous local studies have suggested possible knowledge gaps among patients and physicians in terms of the variability of prescriptions for respiratory infections [18], the sustained reduction was largely due to increased public awareness and hygiene protocols during the pandemic [19], reduced visits due to altered patient health-seeking behavior, and increased referrals to hospitals for severe disease, which was not presented in primary care [20,21]. In 2020 and 2021, the increased accessibility of testing to the public and usage in primary care clinics for the diagnosis of COVID-19 [22,23], nationwide vaccination drives, and the implementation of vaccine-differentiated safe management measures may have amplified this phenomenon [24]. Future studies should be performed to assess the improvement in knowledge, attitudes, and practices of patients toward antibiotic use pertaining to respiratory infections to compared with the pre-COVID-19 pandemic studies [25].
Data from 2021 also showed that skin and genitourinary conditions accounted for the majority (57.9%) of total oral antibiotic prescriptions, highlighting shifts in antibiotic prescription habits among physicians and patient's antibiotic requests. The high percentage of antibiotic prescriptions for dental conditions could be attributed to our algorithm for diagnosis classification ( Figure A1). Further studies are needed to explore the accuracy of these gaps.
Within the clinics, the lack of prioritization in terms of ensuring the accurate coding of diagnosis for antibiotic prescriptions made the assessment and determination of the indications for antibiotic prescriptions challenging. This was evident in the two gaps that we identified: the increase in oral antibiotics prescribed for chronic condition diagnoses and topical skin antibiotic prescriptions with non-skin diagnoses. Similar gaps were discovered in the USA; the antibiotic prescribed was not listed as a diagnostic code in over 50% of cases [26]. We postulate that the similarity in the identified patient and physician factors could be due to a reduced prioritization of accurate coding with an increased consult complexity and diagnostic uncertainty, perceived demand and expectation from certain patient groups (older and female), and a laxity with regulation and dispensing of antibiotics, which differs from practice to practice and physician seniority [27]. Certainly, further research could be performed in this area to ascertain the accuracy and strength of Antibiotics 2023, 12, 762 13 of 29 these associations. The potential collinearity assessed between the factors is a limitation of our study, which we found difficult to adjust for.
While the increase in the "Access" group's antibiotic prescriptions was due to reduced clarithromycin use in respiratory conditions, the "Watch" group's antibiotic utility remained high in genitourinary and gastrointestinal conditions. As we discovered that more experienced, locally trained family physicians were more likely to prescribe "Watch" group antibiotics, this could stem from previous outdated local antimicrobial guidelines, in terms of encouraging ciprofloxacin use for urinary tract infections [28]. This local guideline also recommended ciprofloxacin as a first line for male urinary tract infections, which could possibly explain the factors that we identified (male patients and locally trained physicians). Updated antimicrobial guidelines based on latest antibiograms, whilst important, may not positively influence changes in prescription habits due to significant variability between clinics and physicians [29]. Future interventions such as academic detailing and decision support tools may be effective in monitoring "Watch" group antibiotic prescriptions [30].
Our study discovered an increasing use of dual oral and topical antibiotics for skin conditions, despite discouragement from international guidelines and studies due to a possible increased risk of antimicrobial resistance [31]. The factors identified as predisposing to dual antibiotic usage were as follows: demand and expectation from certain groups (younger patients), differences in co-morbidities that shaped their perceived severity (CKD), differing practices, and regulations in prescriptions (resulting in discrepancies in doctor experience, training location, seniority level, and place of practice) [27]. Further research could be performed to ascertain the burden of specific skin conditions and address the potential knowledge gaps.
A potential limitation in this study is selection bias (due to the use of one public primary care cluster in Singapore), which may not be representative of the whole primary care landscape. Our study found that oral antibiotics were prescribed in 3-5% of all patient visits; this was likely an under-estimate, given that a result of 10% was reported in the USA [32]. Studies conducted abroad were larger in scale and encompassed more data [31]. Compared with the national data, the antibiotic prescription patterns in our study cohort were found to be like the other healthcare clusters [33]. In Singapore, private clinics account for 80% of all primary care services, with 86% of consultations being acute consultations [34]. Most clinics adopted the same prescriber dispenser model, so we expect antibiotic prescriptions to be less regulated, and further studies in private practices to better reflect the antimicrobial gaps of care. Missing diagnoses coding and multiple visit diagnoses might affect the robustness of the tiered ranking system in diagnoses classification, resulting in a misrepresentation of antibiotics prescribed for certain diagnoses (such as dental conditions). An indication of antibiotic prescription might not necessarily equate to visit diagnoses; medical record reviews using machine learning could uncover this difference. The factors affecting the gaps identified should be further brainstormed, extracted, and evaluated to formulate a more accurate representation of antibiotic prescription patterns and habits, and these should be further triangulated in subsequent studies. Due to the lack of guidelines, antibiotic appropriateness could not be determined from this study.

Conclusions
This study showcases the prevalence of antibiotic prescriptions within public primary care in Singapore and their significant reduction, particularly for respiratory conditions during the COVID-19 pandemic. The gaps identified include inaccurate diagnosis coding for oral and topical antibiotic prescriptions, and dual antibiotic use for skin and soft tissue infections. These are associated with certain patient and physician factors. While the usage of "Watch" group antibiotics has decreased, greater emphasis can be placed on antibiotics prescribed for certain diagnoses. This paves the way for further studies in primary care in Singapore and lays the foundation for updated antimicrobial guidelines and stewardship programs in primary care in Singapore. Informed Consent Statement: Patient consent was waived as this study involved a de-identified retrospective extraction of antibiotic prescription records from an internal database. The risks to participants were minimal, with no interventions or procedures performed. Informed consent would not have been possible in this case as participants were not contacted. The information collected was not sensitive in nature, and data obtained were derived from institutional protocols.

Data Availability Statement:
The data presented in the study are available on request from the corresponding author.

Acknowledgments:
We would like to thank Chang Yang Yi and Chew Hui Shan (Family Medicine Development, National University Polyclinics) for the data extraction, and Waseemah Begum (National University Health System Pharmacy) for antibiotic coding and advice.

Conflicts of Interest:
The authors have no conflict of interest to declare.