Background
Mortality rates in children under 5 years old have fallen over the past 3 decades [
1,
2], but remain high in many resource-limited settings [
3]. Most child deaths are associated with infection-related illnesses that commonly present with fever and should be treatable with simple and affordable existing interventions [
1]. Hence, optimising case management of febrile illness is key to improving child survival [
4].
The expanded use of rapid diagnostic tests (RDTs) for malaria has improved case management and reduced inappropriate treatment [
5,
6], but there is an ongoing lack of diagnostic tools for non-malarial fever in resource-constrained settings. Integrated management guidelines developed for these settings are therefore largely dependent on clinical diagnosis with its well-recognised limitations [
7,
8]. Use of these guidelines has promoted early treatment of pneumonia [
9], improved patient outcomes [
10], and reduced healthcare cost [
11], but may also lead to both overprescription and missed treatment opportunities [
12‐
14].
Further to the inherent limitations of empiric guidelines, healthcare workers may make decisions that deviate from guidelines and further contribute to inappropriate management of febrile illnesses. For example, in the absence of severe clinical signs, antimicrobial therapy (antibacterial/antimalarial agents) is not always necessary for fever with no apparent source of infection, and planning follow-up monitoring without initiating these drugs are recommended [
7]. However antimicrobials may be prescribed unnecessarily due to various patient- and healthcare worker-related factors [
15,
16], particularly for patients with a negative malaria test result [
5]. In addition to generating adverse clinical outcomes, inappropriate management of febrile illness wastes healthcare resources, reduces confidence in the health system and enhances the development of drug resistance [
17]. More information is therefore needed to understand prescribing practice, particularly in settings where diagnostic facilities remain limited.
Ethiopia has seen substantial reductions in under-five year old mortality [
18], largely due to reductions in malaria [
19] and pneumonia [
20], and has set ambitious targets for further improvements [
21]. In this regard, understanding factors which determine outcome of major childhood illness is important to help identify those with a higher risk of poorer outcomes so that appropriate intervention measures can be considered. We recently reported on common pathogens and their antimicrobial susceptibility patterns in children with fever attending a large tertiary hospital in the south of the country [
22]. Within the same cohort, we conducted an evaluation, the first in Ethiopia and one of very few in Africa, of clinicians’ adherence to guidelines in prescribing antibacterials and antimalarials, and analysed clinical outcomes at two weeks follow-up. We evaluated whether withholding antimicrobials following guidelines is a safe management approach to reduce unnecessary use without compromising the clinical outcome. We also assessed predictors of hospitalization and outcomes among children with pneumonia, the leading cause of child mortality and hospital admissions [
20].
Discussion
In this comprehensive analysis of clinical management and outcome of febrile illnesses, we identified overprescription of antibacterial and antimalarial agents as a key issue. Specifically, antibacterial agents were overprescribed to 34.0% of children without an initial indication for treatment based on the national guidelines. We also found that 7.3% of children with negative malaria microscopy were overprescribed antimalarial drugs on initial management. Overprescribing antibacterial agents occurred more in younger children, and in those without tachypnea. Overprescribing antimalarials was independently predicted by older age, anaemia, absence of cough and higher fever at presentation. Independent predictors of persisting fever or death among children with pneumonia were age under 12 months, absence of vomiting, and being underweight. Among children whose conditions were determined not to have met guidelines for antibacterial or antimalarial treatment, there was no difference in outcome between those who did and did not receive these agents.
Our study has the strength of being the first in Ethiopia to assess clinical management of childhood febrile illnesses against guidelines and in light of clinical outcomes. Also, there have been very few studies from African countries that reported on management and outcomes of febrile illnesses. Results from this evaluation should inform management approaches and adherence of health workers to policy recommendations of prescribing and withholding antibacterial and antimalarial agents. However, interpretation of our findings needs to take account of several limitations to our study design. First, our goal of investigating fever aetiology in the same cohort [
22] introduced laboratory investigations which went beyond routine practice in the hospital, potentially influencing clinicians’ practice. Second, our laboratory and other diagnostic procedures did not cover all known infectious diseases, limiting our ability to inform on needed improvement of fever management guidelines. Third, missing data due to absence of specimens and loss to follow-up in some children might have introduced bias. Fourth, given that clinicians’ treatment decisions are influenced by various factors, including patients’ clinical presentation and treatment history, availability of drugs, known antimicrobial resistance patterns, and cost, we focussed our evaluation on broad categories of treatment. Last, there may have been recall and other errors in information gained from caregivers’ self-report.
Pneumonia remained the leading syndrome associated with fever presenting at this tertiary hospital despite a reduction in incidence and mortality due to implementation of interventions in Ethiopia including vaccination against
Streptococcus pneumoniae and
Haemophilus influenzae type b [
20]. Ongoing dependence on relatively non-specific clinical case definitions that prioritise sensitivity [
28] in the absence of applicable diagnostic tools and limited access to radiography hinder efforts to curb overuse of antibacterial agents. Blood culture is available to support clinical decision making but it involves a time delay and is generally considered to have a yield of bacteraemia in pneumonia of less than 10% [
33], a finding that we replicated (8.7%) as have others in Ethiopia (5.6%) [
34]. Urine-based pneumococcal tests are also of limited accuracy [
35,
36] in guiding treatment due to pneumococcal carriage, resulting in urinary excretion [
37], false positivity following resolved infection, and cross reaction with other closely related streptococci [
38,
39]. We also observed no significant difference in proportions of children positive for urine
S. pneumoniae antigen by pneumonia status, as other investigators found [
14]. There remains an urgent need for rapid and inexpensive tests that can differentiate bacterial from non-bacterial causes of respiratory infections, including pneumonia and tonsilopharyngitis to guide clinical decision-making and optimal use of antibacterial agents.
Prescribing antibacterial agents based on presence of faecal pus cells/red blood cells, as suggested by the guideline for dysentery, may also lead to over-prescription in settings where enteric bacterial infections such as shigellosis and salmonellosis are rare [
22]. UTI also poses particular difficulties due to the non-specific nature of symptoms, and requires providing routine screening with urinalysis for febrile children as commonly practised in high income countries [
31].
The proportion of overprescribing antibacterial agents in the present study (34.0%) was lower than that reported from a study in Uganda (42.0%) [
15]. Further, our finding about overprescribing antimalarial drugs to patients with negative malaria test (7.1%) was comparable to a result in Tanzania (11.5%) [
40] although higher proportions of overprescribing (21.0–58.0%) [
41,
42] were reported from other African countries. A lower odds of overprescribing antibacterial treatment in older children compared to infants might be due to the likelihood of more conservative empiric antibacterial treatment for infants. The lower odds of overprescribing of antibacterial agents for those with tachypnoea may be because it is a sign of pneumonia. For antimalarials, overprescribing was more common in older children, as it was in children with anaemia, with higher fever, or without cough, clinical characteristics that clinicians may associate with malaria to support a decision to undertake empiric treatment. Notably, antibacterial and antimalarial agents were frequently overprescribed to children with undifferentiated fever. This may suggest that the lack of diagnostic tests for non-malarial fever influences the prescription of both types of drug, as long as the source of fever remains unexplained.
A better understanding of clinician non-adherence to treatment guidelines will benefit from qualitative research into motivating factors. Clinicians may be simply covering themselves against the possibility of missed diagnoses and consequent disease progression if children are sent home with no medication. We found that the proportion of participants without indications for antibacterial or antimalarial therapy who recovered at 7 days was not affected by whether or not they actually received therapy, consistent with earlier findings in relation to malaria treatment in the presence of a negative RDT result [
12,
43]. This finding might provide reassurance to clinicians facing this situation, and support a more rational approach based on ensuring adequate follow-up monitoring in circumstances where it is feasible. Uncertainty about the quality of malaria microscopy may have prompted empiric treatment, potentially suggesting a need for strengthening laboratory quality assurance program to enhance confidence of clinicians on laboratory findings. In this regard, re-testing with malaria RDTs may be helpful where microscopy is negative but there is a high index of suspicion, and provide further support for clinical decision-making within the guidelines. The adherence of caregivers to clinicians’ management advice, particularly in children sent home with no antimalarial or antibacterial prescription, is important to efforts of promoting rational use of antibacterial and antimalarial agents.
A higher proportion (38.1%) of febrile children were hospitalized in our study compared to previous reports from African countries (5.3–24.0%) [
44‐
47], possibly because our recruitment was from a tertiary facility which may have attracted patients with more severe disease than the primary care centres that were the sites for most earlier studies. Pneumonia as a main cause of hospitalization calls for improved access for timely management to minimise disease progression. We found that children with signs of severe respiratory distress (lower chest indrawing or retraction) were more likely to require admission [
8,
28]. We also saw increased odds of hospitalization among children with wasting, and poorer outcomes (persisting fever or death) among infants, who may have had incomplete immunization due to their age, and in those who were underweight. Our findings suggest the need for intensifying interventions, particularly enhancing nutritional status to reduce the risk of acquiring pneumonia and improving outcomes [
48].
Overall, we found that fever had resolved in 89.7% of participants by 7 days, consistent with earlier findings from African studies of children with uncomplicated febrile illnesses [
12,
14,
43]. The proportion of relapse of fever at day 14 in the present study (4.3%) was similar with a single earlier report from Tanzania (4.5%) [
14]. Our observed in-hospital fatality ratio of 5.9% was consistent with findings reported from Tanzania (5.7–7.3%) [
44,
49] although a lower result (1.4%) was also reported from the same country [
46].
Acknowledgements
We would like to acknowledge the study participants and their caregivers. We are thankful to study staff (Dr Bereketab T, Dr Zelalem G, Dr Abreham B, Sr Genet A, Sr Seble N, Sr Simegn T, Sr Tsion M, Mr Enqusillassie M, Ms Bereket T, Mrs Netsanet N, Mrs Elishaday A, Ms Berhane M, Mr Henok M, Ms Seblewongel T, Mrs Tihetina K, Mr Mesfin W, Mr Yidnekachew F) for their support with data collection. We also thank C. Escadafal for assistance with development of laboratory standard operating procedures. We are grateful to Setema Ltd, Ethiopia and Biomerieux Company, France for donating part of blood culture bottles. We extend our thanks to Hawassa University, College of Medicine and Health Sciences, Hawassa, Ethiopia, for allowing us to use the hospital and laboratory facilities and available resources required for this research work. TS received a PhD scholarship (University International Postgraduate Award) administered by UNSW Sydney. JMK is supported by a Fellowship from the NHMRC.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.