Determinants of antibiotic prescription pattern and appropriateness in paediatric care in Ghana: a generalized estimating equation approach

Abstract

Background

Rising antibiotic resistance threatens to reverse health gains from antibiotic access, especially in low- and middle-income countries. Identifying factors influencing prescribing patterns and appropriateness is vital to optimising antibiotic use and strengthening antimicrobial stewardship. However, evidence on this in paediatric outpatient care in Ghana is limited.

Objectives

To identify patient-related and prescriber-related factors associated with antibiotic prescribing patterns and the appropriateness of antibiotic use among paediatric patients.

Methods

A retrospective review of outpatient records at the Princess Marie Louise Children’s Hospital in Ghana was conducted from 1 April to 30 June 2022. Logistic regression and generalised estimating equations analyses were used, where appropriate, to assess associations between patient age and sex, prescriber rank, prior antibiotic use, fever status and the appropriateness of antibiotic prescribing by indication, choice, dosage, duration, antibiotic tier and multiple-antibiotic use.

Results

Among 16 555 outpatient consultations, 3244 (20%) resulted in an antibiotic prescription, of which 2862 records were analysed. Most patients were male (54%). Male sex was associated with antibiotic prescribing when not indicated (p=0.007), whereas older children (p=0.030) and patients presenting with fever (p=0.036) were less frequently prescribed antibiotics when not indicated. Compared with house officers, specialised nurses and physician assistants were associated with inappropriate use of antibiotics (p<0.001 and p=0.005, respectively); however, they were less likely to prescribe multiple or higher-tier antibiotics. Medical officers and residents were associated with appropriate antibiotic dosing and duration (p=0.004 and p=0.005, respectively), and prior antibiotic use doubled the odds of receiving higher-tier antibiotics (p<0.001).

Conclusion

Patient and prescriber characteristics were significantly associated with antibiotic prescribing patterns and appropriateness. Targeted interventions leveraging these findings are needed to improve prescribing and strengthen antimicrobial stewardship.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• While patient and provider factors are known to influence prescribing patterns and appropriateness, their effects vary by context, and important evidence gaps remain in our setting. Addressing these gaps is essential for designing locally effective antimicrobial stewardship strategies.

WHAT THIS STUDY ADDS

• This study reveals that prescription patterns and appropriateness are influenced by prescriber and patient characteristics, including prescriber cadre, patient age and sex, presence of fever and prior antibiotic use.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This creates a foundation for future studies, such as comparative studies and those that explore causal pathways. It supports the need for targeted antimicrobial stewardship interventions. This study offers baseline evidence to inform policy changes and improve paediatric antibiotic use.

Introduction

Antibiotics are crucial for treating and preventing bacterial diseases, playing a significant role in modern medicine. While global access to antibiotics has improved due to factors like better health insurance and rising incomes, this has also contributed to a rise in antibiotic resistance. Resistance develops when bacteria adapt and become impervious to antibiotics, making infections increasingly difficult to treat.¹ Among the key factors driving resistance is inappropriate use in healthcare.² In 2019, antimicrobial resistance (AMR) caused approximately 4.95 million deaths, particularly in Western and sub-Saharan Africa, with projections estimating 10 million deaths annually by 2050 if no action is taken.³ In sub-Saharan Africa, resistance levels to common antibiotics are alarmingly high, driven largely by misuse in both human and veterinary contexts.⁴ In Ghana, research shows high antibiotic usage among patients, with hospitals serving as reservoirs for resistant strains, often due to inappropriate prescribing practices.⁵ Due to a lack of diagnostic support (especially in primary health facilities) or preference, prescribers usually rely on symptoms for diagnosis, leading to empiric prescriptions^{6 7} and high rates of antibiotic use reaching 66%.⁸

Antimicrobial stewardship, the careful and responsible management of antimicrobials, is a key pillar in combating resistance.^9–12 In line with this, the WHO advocates for rational drug use policies like treatment guidelines to minimise resistance,¹³ but implementation is lacking in many low- and middle-income countries,^14–16 and in countries like Ghana, adherence remains voluntary and inconsistent.¹⁶ There is also a significant knowledge gap among healthcare workers regarding antibiotic prescription patterns and factors that influence nonadherence to guidelines (inappropriate use).⁷ Assessing antimicrobial prescriptions/utilisation patterns drives quality improvement initiatives in antimicrobial stewardship,^17–20 as post-prescription reviews with feedback have been shown to effectively reduce antibiotic use.¹² Several studies have reported on antibiotic prescription patterns both internationally^21–23 and locally,^{18 24} some in outpatient and inpatient settings, but there is a paucity of data evaluating antibiotic prescriptions/utilisation in purely paediatric outpatient settings. Using the Standard Treatment Guidelines (STGs),²⁵ this study examines factors associated with antibiotic prescribing patterns and appropriateness in paediatric outpatient care to provide feedback that can inform and strengthen antimicrobial stewardship programmes.

Materials and methods

Study design

This was a retrospective chart review involving the extraction of secondary data from patients’ electronic medical records from 1 April 2022 to 30 June 2022.

Study site

The study was conducted at Princess Marie Louise Children’s Hospital (PMLCH) in Accra, Ghana, a 110-bed specialist facility and the largest stand-alone children’s hospital in the country. It serves mostly children aged 0–18 years, with an average of 157 daily outpatient and emergency visits and about 440 admissions per month, primarily from Accra and surrounding areas.

Inclusion criteria

All children (aged 0–18 years) who visited the Outpatient Department at PMLCH between 1 April and 30 June 2022 and received only oral antibiotics were included in the study. The study focused on outpatient departments because drug use studies are generally recommended to be conducted within a single type of health facility to ensure meaningful interpretation of the findings. Although inpatient prescribing can also be evaluated, it is often more complex than outpatient prescribing. Given the specific objectives of this study, the outpatient setting was therefore more appropriate for evaluating antibiotic prescribing patterns and appropriateness.²⁶

Exclusion criteria

Patients with incomplete medical records, diagnoses not covered by the STGs,²⁵ or antibiotics (eg, linezolid) not listed in the STGs were excluded. Patients on anti-tuberculosis drugs and sickle cell patients on penicillin V prophylaxis were also excluded because these represent long-term treatment or prophylactic regimens rather than prescribing decisions for the presenting outpatient illness, which could bias estimates of antibiotic prescribing patterns and appropriateness.

Sampling method and sample size determination

The sampling method used was a census sampling. Thus, all Electronic Medical Records obtained within the period of the study (1 April 2022 to 30 June 2022) were considered.

Operational definition

Appropriate indication

Assessed whether the antibiotic prescribed was recommended by the STGs,²⁵ based on the differential or provisional diagnosis recorded during the patient consultation.

Appropriate choice

For cases where an antibiotic was indicated, assessed whether the prescribed drug matched the STG recommendation for the recorded diagnosis. For cases where there were multiple diagnoses, the diagnosis closely related to the antibiotic prescribed was considered.

Appropriate duration

Assessed whether the prescribed duration matched the STG recommendation (for cases where the antibiotic was indicated and the correct drug was chosen).

Appropriate dosage

Assessed whether the prescribed dosage matched the STG recommendation (for cases where the antibiotic was indicated and the correct drug was chosen), allowing for minor adjustments to fit available formulation strengths.

Prior antibiotic exposure

Any antibiotic taken by the patient within 28 days before the consultation.²⁷

First-line therapy

Antibiotic recommended as the first-line treatment for the recorded diagnosis according to the STG.

Higher-tier antibiotics

Antibiotics recommended as second-line or third-line treatment for the recorded diagnosis according to the STGs.

Data collection techniques and tools

Initially, a summary spreadsheet was obtained from the hospital’s Lightwave Health Information Management System. This spreadsheet had all patients who attended the outpatient department over the period under study. Additionally, a Google Forms-based data extraction tool was manually populated using data from patient folders, accessed one at a time. Data was extracted from each folder for which at least one antibiotic was prescribed (applying the inclusion and exclusion criteria). The Google Forms were divided into four sections. Section 1 collected demographic and clinical information (eg, age, sex, diagnosis, history of fever, antibiotic use, prescriber rank and indication for antibiotics). Sections 2, 3 and 4 focused on each antibiotic prescribed, including name, appropriateness in choice, duration and dosage. Where there was missing information for the history of fever and prior antibiotic use, missing data were recorded as a variable (information not available).

Data handling and analysis

Data were checked for completeness at the point of collection and analysed using Microsoft Excel and Stata V.15. Excel was used to estimate the prevalence of antibiotic use and to perform descriptive analyses, while Stata was used for inferential statistics. Patient characteristics, including sociodemographic and clinical variables, and the prescriber characteristics (prescriber rank) were summarised at the patient level. Variables, including age, sex, prescriber rank, history of fever and prior antibiotic use (independent variables), were selected a priori based on existing literature and clinical relevance. Logistic regression was used to assess the association between these and the number and highest tier of antibiotics prescribed because each consultation was treated as a single observation, irrespective of the number of antibiotics prescribed. However, when assessing the association between the independent variables and the appropriateness of antibiotic choice, dosage and duration, each prescribed antibiotic was treated as a separate observation. As a result, patients who received more than one antibiotic contributed multiple observations, resulting in repeated patient and prescriber characteristics for each antibiotic. These observations were therefore not independent. To account for this within-patient correlation, generalised estimating equations with an exchangeable working correlation structure and robust SEs, clustered by patient ID, were used. Statistical significance was set at p<0.05.

Patient and public involvement

Patients and the public were involved in the dissemination plans of this research.

Results

Sociodemographic and clinical characteristics of patients

Out of 16 555 OPD patient consultations during the study period, 3244 (20%) received antibiotics. Table 1 displays the sociodemographic and clinical characteristics of patients. Most antibiotic recipients were male (54%, n=1539) and aged 5 years or younger (78%, n=2239). The top diagnoses were pharyngitis/tonsillitis, otitis media, pneumonia, bacterial skin infections and diarrhoea/gastroenteritis, respectively. Patients who had used antibiotic(s) prior to consultation were 11.81% (n=338).

Table 1. Sociodemographic and clinical characteristics of paediatric outpatients receiving antibiotics.

Patient characteristics	Category	Frequency (n=2862)	Percentage
Age range
	0–27 days	5	0.17
	28 days to 12 months	432	15.09
	13–23 months	629	21.98
	2–5 years	1173	40.99
	6–11 years	511	17.85
	12–18 years	112	3.91
Sex
	Ambiguous	2	0.07
	Female	1321	46.16
	Male	1539	53.77
Presence of fever
	Yes	1916	66.95
	No	759	26.52
	Information not available	187	6.53
Prior antibiotic use
	No	1627	56.85
	Yes	338	11.81
	Information not available	897	31.34
Diagnosis*
	Pharyngitis/tonsillitis	1143	36.00
	Acute otitis media	530	16.69
	Pneumonia	419	13.20
	Bacterial skin infection	290	9.13
	Diarrhoea/acute gastroenteritis	289	9.10
	Urinary tract infection	255	8.03
	Mumps	43	1.35
	Nasopharyngitis	33	1.04
	Cellulitis	22	0.69
	Sinusitis	20	0.63
	Periodontal infection	16	0.50
	Severe acute malnutrition	15	0.47
	Allergic rhinitis	12	0.38
	Bronchiolitis	11	0.35
	Acute exacerbation of asthma	8	0.25
	Cough	8	0.25
	Typhoid fever	8	0.25
	Vomiting	7	0.22
	Atopic dermatitis	7	0.22
	Chicken pox	6	0.19
	scdx with voc	6	0.19
	Acute gastritis	5	0.16
	Others	22	0.69

^*Multiple diagnoses present in some cases.

Prescribing characteristics and nature of prescriptions

House officers attended to nearly half (48%) of all patients who received at least one antibiotic. First-line antibiotics were used at the highest level in 87% of consultations. Antibiotics were indicated in 95% of cases where at least one antibiotic was given. Among the prescriptions where antibiotics were indicated, the appropriate choice of antibiotic was prescribed 98% of the time, but the appropriate duration and dosage were prescribed only 40% and 85% of the time, respectively. Antibiotics used were mainly co-amoxiclav (45%), cefuroxime (16%) and amoxicillin (10%) (table 2).

Table 2. Prescribing characteristics and nature of prescriptions for paediatric outpatients who received antibiotics.

Prescribing characteristics	Category	Frequency (n=2862)	Percentage
Rank of prescriber
	House officer	1371	47.90
	Medical officer	902	31.52
	Nurse	252	8.81
	Physician assistant	259	9.05
	Resident	55	1.92
	Specialist (paediatrician)	23	0.80
Number of antibiotics
	One	2754	96.23
	Two	108	3.77
Highest antibiotic category prescribed
	First-line	2503	87.46
	Second-line	312	10.90
	Third-line	47	1.64
Indication for antibiotics*
	Yes	2719	95.00
	No	143	5.00
Prescriptions†		(n=2824)
	Right choice	2765	97.91
	Wrong choice	59	2.09
		(n=2765)
	Right duration	1097	39.67
	Wrong duration	1668	60.33
		(n=2765)
	Right dosage	2353	85.10
	Wrong dosage	412	14.90
Antibiotics prescribed†		(n=2967)
	Co-amoxiclav	1344	45.30
	Amoxicillin	307	10.35
	Cefuroxime	462	15.57
	Flucloxacillin	285	9.61
	Azithromycin	268	9.03
	Ciprofloxacin	133	4.48
	Ciprofloxacin+Tinidazole	1	0.03
	Erythromycin	75	2.53
	Metronidazole	59	1.99
	Clindamycin	28	0.94
	Doxycycline	3	0.10
	Nitrofurantoin	1	0.03
	Secnidazole	1	0.03

^*Patient-level characteristics used.

^†Some patients received multiple antibiotic prescriptions; therefore, each antibiotic was counted as a separate entry.

Patient and prescriber characteristics associated with appropriateness of antibiotic prescriptions

Male sex was associated with antibiotic prescribing when not indicated (adjusted OR (aOR) 1.64, 95% CI 1.14 to 2.34), whereas older age (12–18 years) (aOR 0.25, 95% CI 0.07 to 0.88) and a history of fever (aOR 0.76, 95% CI 0.51 to 1.15) were associated with fewer antibiotic prescriptions when not indicated. Compared with house officers, specialised nurses and physician assistants were associated with antibiotic prescribing when not indicated (specialised nurses: aOR 4.08, 95% CI 2.55 to 6.53; physician assistants: aOR 2.19, 95% CI 1.27 to 3.76) (table 3).

Table 3. Univariable and multivariable generalised estimating equations analyses of factors associated with antibiotic prescribing when not indicated.

Variable	OR (95% CI)	P value (OR)	aOR (95% CI)	P value (aOR)
Age
28 days to 12 months	Ref		Ref
0–27 days	Non-estimable		Non-estimable
13–23 months	0.57 (0.33 to 0.99)	0.045*	0.58 (0.33 to 1.02)	0.057
2–5 years	0.77 (0.49 to 1.21)	0.262	0.70 (0.44 to 1.12)	0.139
6–11 years	0.64 (0.36 to 1.12)	0.115	0.62 (0.35 to 1.11)	0.109
12–18 years	0.36 (0.11 to 1.21)	0.099	0.25 (0.07 to 0.88)	0.030*
Sex
Female	Ref		Ref
Male	1.64 (1.15 to 2.33)	0.006*	1.64 (1.14 to 2.34)	0.007*
Ambiguous	Non-estimable		Non-estimable
Rank of prescriber
House officer	Ref		Ref
Specialised nurse	4.75 (3.04 to 7.40)	<0.001*	4.08 (2.55 to 6.53)	<0.001*
Physician assistant	2.47 (1.47 to 4.14)	0.001	2.19 (1.27 to 3.76)	0.005*
Medical officer	0.90 (0.57 to 1.43)	0.666	0.87 (0.54 to 1.38)	0.546
Resident	0.48 (0.07 to 3.58)	0.478	0.46 (0.06 to 3.36)	0.459
Specialist (paediatricians)	1.23 (0.16 to 9.31)	0.841	1.18 (0.17 to 8.35)	0.868
History of fever
No	Ref		Ref
Yes	0.66 (0.46 to 0.95)	0.026*	0.76 (0.51 to 1.15)	0.036*
Information not available	0.98 (0.51 to 1.88)	0.943	0.84 (0.42 to 1.68)	0.661
Prior antibiotics given
No	Ref		Ref
Yes	1.16 (0.64 to 2.10)	0.631	1.05 (0.57 to 1.92)	0.791
Information not available	2.27 (1.59 to 3.24)	<0.001*	1.41 (0.93 to 2.13)	0.009*

^*P value <0.05.

aOR, adjusted OR.

No significant associations were observed between appropriate antibiotic choice and age, sex, prescriber rank, history of fever or prior antibiotic use (online supplemental table 1). However, medical officers were associated with correct antibiotic dosing compared with house officers (aOR 1.45, 95% CI 1.12 to 1.86). Additionally, a history of fever was associated with correct antibiotic dosing (aOR 1.28, 95% CI 1.01 to 1.62) (table 4).

Table 4. Univariable and multivariable generalised estimating equations analyses of factors associated with appropriate dosage of antibiotics.

Variable	OR (95% CI)	P value (OR)	aOR (95% CI)	P value (aOR)
Age
28 days to 12 months	Ref		Ref
0–27 days	0.87 (0.10 to 7.88)	0.900	1.02 (0.12 to 8.91)	0.985
13–23 months	1.37 (0.97 to 1.94)	0.072	1.36 (0.96 to 1.92)	0.082
2–5 years	1.47 (1.08 to 2.01)	0.015*	1.48 (1.08 to 2.02)	0.014*
6–11 years	0.95 (0.67 to 1.34)	0.754	0.98 (0.69 to 1.39)	0.906
12–18 years	1.40 (0.77 to 2.54)	0.271	1.41 (0.76 to 2.61)	0.279
Sex
Female	Ref		Ref
Male	0.93 (0.75 to 1.15)	0.504	0.92 (0.74 to 1.14)	0.433
Ambiguous	Non-estimable		Non-estimable
Rank of prescriber
House officer	Ref		Ref
Specialised nurse	1.34 (0.87 to 2.05)	0.185	1.30 (0.82 to 2.04)	0.264
Physician assistant	0.88 (0.61 to 1.26)	0.476	0.86 (0.60 to 1.24)	0.417
Medical officer	1.45 (1.13 to 1.86)	0.004*	1.45 (1.12 to 1.86)	0.004*
Resident	1.32 (0.59 to 2.95)	0.506	1.31 (0.58 to 2.99)	0.514
Specialist (paediatricians)	1.98 (0.46 to 8.52) 0.361		2.19 (0.52 to 9.13) 0.283
History of fever
No	Ref		Ref
Yes	1.31 (1.04 to 1.65)	0.025*	1.28 (1.01 to 1.62)	0.044*
Information not available	0.96 (0.62 to 1.50)	0.866	0.89 (0.57 to 1.39)	0.613
Prior antibiotics given
No	Ref		Ref
Yes	0.78 (0.56 to 1.07)	0.126	0.80 (0.58 to 1.11)	0.188
Information not available	0.98 (0.77 to 1.24)	0.844	1.01 (0.78 to 1.29)	0.965

^*P value <0.05. Analysis was restricted to prescriptions with an appropriate indication and antibiotic choice.

aOR, adjusted OR.

The odds of receiving the correct duration of antibiotic decreased among children aged 2–5 years and 6–11 years compared with those aged 28 days to 12 months. Relative to house officers, residents were associated with correct antibiotic duration (aOR 2.36, 95% CI 1.29 to 4.31), while specialised nurses (aOR 0.51, 95% CI 0.36 to 0.72) and physician assistants (aOR 0.72, 95% CI 0.53 to 0.97) were associated with incorrect antibiotic duration. Also, the presence of fever decreased the odds of prescribing the correct antibiotic duration (aOR 0.38, 95% CI 0.32 to 0.46) (table 5).

Table 5. Univariable and multivariable generalised estimating equations analyses of factors associated with appropriate duration of antibiotics.

Variable	OR (95% CI)	P value (OR)	aOR (95% CI)	P value (aOR)
Age
28 days to 12 months	Ref		Ref
0–27 days	5.10 (0.57 to 46.10)	0.147	2.83 (0.30 to 26.57)	0.363
13–23 months	0.96 (0.74 to 1.24)	0.747	1.00 (0.77 to 1.29)	0.978
2–5 years	0.69 (0.55 to 0.88)	0.002*	0.69 (0.54 to 0.87)	0.002*
6–11 years	0.83 (0.63 to 1.09)	0.181	0.73 (0.55 to 0.97)	0.029*
12–18 years	1.31 (0.85 to 2.01)	0.217	1.14 (0.72 to 1.79)	0.584
Sex
Female	Ref		Ref
Male	0.84 (0.72 to 0.98)	0.026*	0.85 (0.73 to 1.00)	0.052
Ambiguous	Non-estimable		Non-estimable
Rank of prescriber
House officer	Ref		Ref
Specialised nurse	0.59 (0.43 to 0.81)	0.001*	0.51 (0.36 to 0.72)	<0.001*
Physician assistant	0.74 (0.55 to 0.99)	0.043*	0.72 (0.53 to 0.97)	0.031*
Medical officer	1.00 (0.84 to 1.19)	0.972	0.97 (0.81 to 1.16)	0.746
Resident	2.26 (1.29 to 3.97)	0.004*	2.36 (1.29 to 4.31)	0.005*
Specialist (paediatricians)	2.57 (1.07 to 6.16)	0.035*	2.15 (0.88 to 5.24)	0.093
History of fever
No	Ref		Ref
Yes	0.38 (0.32 to 0.46)	<0.001*	0.38 (0.32 to 0.46)	<0.001*
Information not available	1.00 (0.71 to 1.40)	1.000	0.99 (0.70 to 1.40)	0.954
Prior antibiotics given
No	Ref		Ref
Yes	1.61 (1.26 to 2.05)	<0.001*	1.56 (1.20 to 2.02)	0.001*
Information not available	1.14 (0.96 to 1.36)	0.136	1.14 (0.94 to 1.37)	0.179

^*P value <0.05. Analysis was restricted to prescriptions with an appropriate indication and antibiotic choice.

aOR, adjusted OR.

Older age was associated with receiving multiple antibiotics. Compared with children aged 28 days to 12 months, the odds of receiving multiple antibiotics increased progressively with age: aOR 2.46 (95% CI 1.15 to 5.26) for children aged 2–5 years, 2.58 (95% CI 1.14 to 5.81) for those aged 6–11 years and 4.58 (95% CI 1.69 to 12.43) for those aged 12–18 years. Relative to house officers, the odds of prescribing multiple antibiotics were reduced among specialised nurses (aOR 0.12, 95% CI 0.03 to 0.52), physician assistants (aOR 0.39, 95% CI 0.15 to 0.98), and medical officers (aOR 0.55, 95% CI 0.35 to 0.87) (online supplemental table 2).

Compared with house officers, the odds of prescribing higher-tier antibiotics were reduced among specialised nurses (aOR 0.35, 95% CI 0.20 to 0.61), physician assistants (aOR 0.26, 95% CI 0.14 to 0.47) and medical officers (aOR 0.65, 95% CI 0.51 to 0.85). Additionally, prior antibiotic use was associated with increased odds of prescribing higher-tier antibiotics (aOR 2.12, 95% CI 1.56 to 2.86) (online supplemental table 3).

Discussion

The overall prevalence of antibiotic use in our outpatient department (20%) was well within the local and WHO targets of 20% and 30%,²⁸ respectively, possibly due to recent antimicrobial stewardship campaigns and awareness.^29–31 However, the prevalence was substantially lower than that reported in China (63.8%)³² and India (36.8%),³³ but was similar to that reported in Madagascar, Senegal and Cambodia (29.3%).³⁴

While a meta-analysis from high-income countries found that females generally received more antibiotics,³⁵ the opposite pattern was observed in this low-income and middle-income paediatric setting. Males received more antibiotics, consistent with findings by Mandal et al³³ and Young et al,³⁶ and were 1.6 times more likely to receive antibiotics when not indicated, as also reported in Uganda.³⁷ Limited diagnostic access in these settings may encourage reliance on symptom severity,³⁸ which may be greater in males because of biological and hormonal differences.³⁹ Stewardship interventions addressing symptom-based prescribing and promoting the appropriate use of diagnostic investigations could therefore reduce unnecessary antibiotic use, particularly among boys.

Children aged 5 years or younger accounted for most antibiotic prescriptions, consistent with findings from Finland⁴⁰ and India,³³ likely because of their greater burden of infectious diseases.⁴¹ However, many of these infections are viral and do not require antibiotics.⁴² These children were also more likely to receive antibiotics when not indicated, as reported in the USA.⁴³ This may reflect clinicians’ perception that younger children are more vulnerable or caregiver pressure to prescribe antibiotics.⁴⁴ Antimicrobial stewardship programmes that educate caregivers of younger children about the signs and symptoms of viral infections and equip clinicians to manage caregiver pressure for unnecessary antibiotic prescribing could reduce inappropriate antibiotic use in this age group.⁴⁵ Children aged 6–11 years were less likely to receive the correct antibiotic duration than infants, consistent with findings from Ethiopia, where inappropriate antibiotic use was also common in this age group.⁴⁶ This may be explained by the higher frequency of conditions such as acute pharyngitis/tonsillitis and acute otitis media (AOM) among them,⁴⁷ for which penicillin was often prescribed for shorter durations than recommended. For example, although Ghana’s STGs recommend a 10-day course of antibiotics for AOM across all age groups, nearly three-quarters of patients received only 5–7 days. This suggests that prescribing practices may not fully align with national guidelines and may instead reflect international trends, as similar deviations have been reported in the USA.⁴⁸ Qualitative studies exploring prescribers’ views and reasons for deviating from recommended treatment durations could provide further insight into this issue and inform improvements in treatment guidelines, antibiotic formulations and prescribing practices. Our study also found that older children were more likely to be prescribed multiple antibiotics than infants, despite infants being more vulnerable to severe infections.⁴⁹ This is worrisome, as Obakiro et al found that health workers were less likely to adhere to guidelines when prescribing antibiotics in combination.³⁷ This pattern may stem from diagnostic uncertainty or overestimation of infection complexity in older children.⁵⁰ Addressing this issue will require further investigation into the underlying causes of these prescribing practices and a stronger focus on evidence-based guidelines to ensure appropriate antibiotic use across all age groups.

The majority (67%) of patients who received antibiotics in our study complained of fever, which aligns closely with findings from Opoku et al, who reported a 70.1% prevalence of antibiotic prescriptions among febrile patients in outpatient departments in Ghana.⁴¹ This suggests fever is a common trigger for antibiotic use, often due to concerns about underlying infections. However, they were less likely to receive the correct duration of antibiotics. This may indicate that fever, while often seen as a key indicator for antibiotic treatment, does not always result in appropriate or guideline-adherent prescribing practices. Symptom-based antimicrobial stewardship interventions targeting fever, including guidelines,⁵¹ clinical flowcharts and diagnostic tools such as C reactive protein testing⁵² to differentiate viral from bacterial causes of fever, could reduce inappropriate antibiotic use.

Patients who had received antibiotics before presentation had double the odds of being prescribed a higher-tier antibiotic than those who had not. Higher-tier antibiotics are broad-spectrum and often prescribed when initial treatments fail or when there are concerns about resistant infections.⁵³ Inappropriate use of higher-tier antibiotics is dangerous, as these antibiotics are our only defence against resistant strains. Lack of clinical improvement following initial first-line antibiotic therapy should prompt prescribers to consider alternative explanations, including substandard or falsified medicines, poor adherence, viral infections or an incorrect diagnosis, rather than automatically escalating to higher-tier antibiotics. Antimicrobial stewardship programmes should therefore train prescribers to systematically evaluate these factors before prescribing higher-tier antibiotics.

Our study found that 87% of antibiotic prescriptions were first-line, exceeding the recommended 80% benchmark.⁵⁴ This is encouraging, as unnecessary broad-spectrum antibiotic use promotes AMR.⁵³ Antibiotics most commonly prescribed included co-amoxiclav, cefuroxime and amoxicillin, consistent with other findings.^{21 55} Despite this positive trend, antibiotic overuse was evident. Antibiotics were prescribed at the first visit for 75% of children with pharyngitis and 74% of children with AOM, although the vast majority of cases are caused by viruses.⁵⁶ Also, 18% of children with acute diarrhoea received antibiotics despite WHO recommendations for ORS and zinc as mainstays and antibiotics only for bloody diarrhoea, suspected cholera with severe dehydration and serious non-intestinal infections.⁵⁷ Since bacterial causes like bloody diarrhoea are uncommon (2.2%)⁵⁸ and no cholera cases were reported, this suggests unnecessary prescribing, consistent with prior reports of up to 40% antibiotic use for non-bacterial cases.⁵⁹ This may reflect the lack of WHO-recommended clinical decision-support tools in Ghana’s STGs, such as clinical severity scoring systems and guidance on watchful waiting for mild cases.^{25 56} Updating the guidelines to incorporate these approaches and align more closely with the WHO AWaRe guidelines⁵⁶ could therefore reduce unnecessary antibiotic use.⁶⁰

Prescribers’ years of experience have been found to have an effect on treatment decisions,⁶¹ and house officers (who have <3 years of experience) accounted for the majority of antibiotic prescriptions. This was inconsistent with findings by Opoku et al that found prescribers with more years of experience (>5 years) being more likely to prescribe antibiotics compared with those with fewer years (<3 years).⁴¹ This was likely because house officers constituted the largest proportion of prescribers at PMLCH. House officers were also more likely to prescribe multiple or select higher-tier antibiotics than specialised nurses, physician assistants and medical officers, possibly reflecting more cautious prescribing practices due to their limited clinical experience.⁶² Compared with house officers, medical officers and residents were more likely to prescribe the correct antibiotic dosage and duration, respectively, probably reflecting greater clinical experience and familiarity with treatment guidelines. House officer orientation should therefore include training on STGs, exposure to antimicrobial stewardship initiatives and mentorship from senior clinicians (residents and medical officers). In contrast, specialised nurses and physician assistants were more likely to prescribe antibiotics when not indicated, consistent with findings from Uganda, where clinical officers with similar roles also tended to prescribe antibiotics unnecessarily.⁶³ Similar observations have been reported elsewhere, with less-qualified healthcare providers more likely to prescribe antibiotics,⁶⁴ possibly because of professional culture, diagnostic uncertainty⁵⁰ or limited familiarity with clinical guidelines. Strengthening formal and continuous medical training for physician assistants and specialised nurses may improve diagnostic confidence, encourage consideration of alternative management strategies and reduce inappropriate antibiotic prescribing.⁶⁵

Limitations

This study was limited to enteral antibiotics, excluding parenteral formulations commonly used in inpatient settings and for certain conditions such as STIs in older adolescents. Most diagnoses were provisional rather than microbiologically confirmed, and cases outside the STGs were excluded. As with all retrospective studies, the findings were subject to incomplete or inaccurate clinical documentation, which may lead to misclassification of antibiotic appropriateness, and also does not allow for causal inference. In addition, the observed association between sex and antibiotic indication may reflect unmeasured differences in underlying diagnoses. Only prescriber rank was available as a prescriber-related characteristic; other characteristics, such as sex, age, years of experience and AMS training, could not be assessed because they were not routinely recorded in the electronic medical records.

Recommendations

Policymakers and clinicians should adopt a multifaceted approach to antibiotic use surveillance and antimicrobial stewardship, as inappropriate prescribing may persist despite achieving current antibiotic prescribing targets. Future research should standardise this surveillance approach and evaluate whether WHO targets should be expanded to include measures of prescribing quality (eg, indication, antibiotic choice, dosage and duration).

Conclusion

The prevalence of antibiotic prescribing among paediatric outpatients was within national and WHO targets, with high adherence to first-line antibiotic prescribing. However, inappropriate prescribing remained evident, particularly among males, younger children, patients presenting with fever and certain diagnoses. One prescriber group was more likely to prescribe multiple and higher-tier antibiotics, practices that could promote AMR. Strengthening antimicrobial stewardship through improved diagnostic support, updated clinical guidelines, targeted prescriber training and mentorship, and caregiver education could further optimise antibiotic use and help mitigate AMR.

Data availability statement

Data are available upon reasonable request.