Antibiotic-Prescribing Practices for Management of Childhood Diarrhea in 3 Sub-Saharan African Countries: Findings From the Vaccine Impact on Diarrhea in Africa (VIDA) Study, 2015–2018

Abstract Background Despite antibiotic prescription being recommended for dysentery and suspected cholera only, diarrhea still triggers unwarranted antibiotic prescription. We evaluated antibiotic-prescribing practices and their predictors among children aged 2–59 months in the Vaccine Impact on Diarrhea in Africa (VIDA) Study performed in The Gambia, Mali, and Kenya. Methods VIDA was a prospective case-control study (May 2015–July 2018) among children presenting for care with moderate-to-severe diarrhea (MSD). We defined inappropriate antibiotic use as prescription or use of antibiotics when not indicated by World Health Organization (WHO) guidelines. We used logistic regression to assess factors associated with antibiotic prescription for MSD cases who had no indication for an antibiotic, at each site. Results VIDA enrolled 4840 cases. Among 1757 (36.3%) who had no apparent indication for antibiotic treatment, 1358 (77.3%) were prescribed antibiotics. In The Gambia, children who presented with a cough (adjusted odds ratio [aOR]: 2.05; 95% confidence interval [95% CI]: 1.21–3.48) were more likely to be prescribed an antibiotic. In Mali, those who presented with dry mouth (aOR: 3.16; 95% CI: 1.02–9.73) were more likely to be prescribed antibiotics. In Kenya, those who presented with a cough (aOR: 2.18; 95% CI: 1.01–4.70), decreased skin turgor (aOR: 2.06; 95% CI: 1.02–4.16), and were very thirsty (aOR: 4.15; 95% CI: 1.78–9.68) were more likely to be prescribed antibiotics. Conclusions Antibiotic prescription was associated with signs and symptoms inconsistent with WHO guidelines, suggesting the need for antibiotic stewardship and clinician awareness of diarrhea case-management recommendations in these settings.

The overuse or misuse of antibiotics when deemed unnecessary or not recommended by the international and national treatment guidelines is a major global public health threat [1]. Diarrheal diseases remain a common clinical syndrome in which inappropriate antibiotics are used, even though most of these illnesses are caused by viruses or self-limited bacterial infections. Accordingly, the World Health Organization (WHO) Integrated Management of Childhood Infections (IMCI) handbook does not recommend antibiotics for the majority of diarrheal episodes, with the exception of bloody diarrhea (with the aim of treating shigellosis, the most common cause), suspected cholera with severe dehydration, or coexisting severe acute malnutrition (SAM) [2]. Inappropriate use of antibiotics for diarrheal diseases or other conditions can promote antimicrobial resistance (AMR) to both diarrheal and nondiarrheal pathogens [3], increase healthcare costs to both providers and families [4], and increase morbidity and risks of adverse reactions, including prolonged hospitalization [5][6][7][8][9]. Antibiotic-resistant strains of Shigella, for example, have emerged, leaving few options for effective, affordable therapy [10][11][12].
Although AMR patterns can differ by geographic region [13], resistant strains have spread globally [14] and exert the greatest impact in low-income countries [15,16]. Resistance to essential antibiotics continues to increase in sub-Saharan Africa [17]-a setting where knowledge of factors driving antibiotic prescription remains poorly understood. This was exemplified by a study of AMR among atypical enteropathogenic Escherichia coli isolated from stool samples from children younger than 5 years old with and without diarrhea in 7 developing countries in sub-Saharan Africa and South Asia participating in the Global Enteric Multicenter Study (GEMS) [18]. Using a combination of phenotyping and genomics, these investigators found that 65% of isolates displayed resistance to 3 or more drug classes over a 3-year period. To better understand the drivers of antibiotic prescribing for diarrhea in sub-Saharan Africa, we examined findings from the Vaccine Impact on Diarrhea in Africa (VIDA) Study, a 3-year follow-on study to GEMS that prospectively collected data on the clinical management of MSD among children younger than 5 years of age attending healthcare facilities at 3 sites in sub-Saharan Africa [19].

Study Design
The aim of this analysis was to assess the prevalence of inappropriate antibiotic-prescribing practices and their predictors among cases with MSD aged 2-59 months nested within VIDA, a case-control study designed to elucidate the incidence, etiology, and adverse clinical consequences of MSD in children aged 0-59 months residing in censused populations in Basse and Bansang, The Gambia; Bamako, Mali; and Siaya County, Kenya following rotavirus vaccine introduction. VIDA used comparable methods to GEMS [20]. The study methods are described in detail previously [21] and more recently [20].

Case Definition and Recruitment Methods
In brief, during a 36-month period from May 2015 to July 2018, each site enrolled children 0-59 months of age belonging to a censused population when they sought care at local Sentinel Health Centers (SHCs). This analysis included a subset of these children in 3 age strata (2-11 months, 12-23 months, and 24-59 months) with an acute, new episode of MSD, defined as 3 abnormally loose or watery stools in the previous 24 hours, accompanied by at least 1 of the following: sunken eyes, skin tenting, intravenous rehydration, or hospitalization. To study antibiotic use in those children with MSD for whom antibiotics were not recommended, children enrolled in the main VIDA study were excluded from this analysis if they were younger than 2 months of age (because WHO recommendations do not address this age group [2]) or who had a WHO indication for antibiotics based on clinician diagnosis, including dysentery, pneumonia, cholera, meningitis, or other invasive bacterial infections; otitis media, tonsillitis, or pharyngitis; or SAM [2]. The list of the definitions for the exclusion criteria are shown in Supplementary Table 1. Note that, per IMCI, cough or difficulty breathing in the absence of chest in-drawing or tachypnea was not considered an indication for antibiotics.

Data Collection, Study Definitions, and Statistical Analysis Methods
Information on demographic, epidemiological, and clinical characteristics prior to presentation at the SHC was collected from the child's primary caretaker at enrollment on standardized forms. We recorded the study clinician's assessment of the child at enrollment and the medical management during the child's stay at the SHC, including antibiotic prescription data. Antibiotic prescription was defined as either having been given an antibiotic at the facility or receiving a prescription for home use. We defined probable unnecessary antibiotic use as prescription or use of antibiotics when not indicated by WHO guidelines. Undernutrition was defined as children who were wasted/very thin based on clinician judgement but who did not meet criteria for SAM. Clinical fever was defined as a temperature 37.5°C or higher. Tachypnea was defined as respiratory rate in breaths/minute greater than 50 in infants aged 2-12 months and greater than 40 in children aged 1-5 years.
Using either chi-square or Fisher`s exact test, as appropriate, we compared proportions of MSD cases with various clinical and demographic characteristics among those who were or were not given a prescription for antibiotics. Medians for continuous variables were compared using a Wilcoxon rank-sum test. We performed logistic regression with antibiotic prescription as the outcome and all demographic and clinical symptoms as potential covariates separately for each site to allow for site heterogeneity (clinical presentation, demographic, and national guidelines). For all analyses, a P value less than .05 was considered statistically significant. Data analysis was performed using Stata/SE 16.0 (StataCorp, College Station, TX, USA).

Ethical Approval
This study was approved by the ethical review committees at the University of Maryland, Baltimore (HP-00062472); the Centers for Disease Control and Prevention (CDC) (reliance agreement 6729); The Gambia Government/Medical Research Council/Gambia at the London School of Hygiene . Written informed consent was obtained from the parent(s) or primary caretaker(s) of each child who met eligibility criteria before any research activities were performed.

Participants
VIDA enrolled a total of 4840 MSD cases at the 3 study sites combined. The number of cases who met exclusion criteria for this analysis was 806 of 1678 (48.0%) in The Gambia, 468 of 1608 (29.1%) in Mali, and 1066 of 1554 (68.6%) in Kenya (Supplementary Table 2). The leading causes of exclusion across the sites were severe pneumonia and severe febrile disease.

Antibiotic Prescription Patterns
Among the MSD cases enrolled, 734 of 4840 (15.2%) had a diarrheal indication for antibiotic prescription and, of these, 732 of 734 (99.7%) had dysentery. Among those with dysentery, 641 of 732 (87.6%) were prescribed antibiotics. The most commonly prescribed antibiotics among dysenteric cases were ciprofloxacin ( The distribution of characteristics of MSD cases with no antibiotic indication who did and did not receive antibiotic prescription stratified by site is shown in Table 2. This univariate analysis showed that children with cough were significantly more likely to receive a prescription for inappropriate antibiotics in The Gambia (28.9% vs 19.9%; P = .039). A similar pattern was seen in Kenya for cough (35.7% vs 23.9%; P = .054) and difficulty breathing (5.7% vs 2.4%; P = .048). Children with signs of dehydration were less likely to receive an inappropriate antibiotic prescription in The Gambia (very thirsty: 78.0% vs 86.0% [P = .043]; dry mouth: 65.9% vs 76.6% [P = .02]) but more likely in Mali (dry mouth: 66.1% vs 42.1%; P = .029) and Kenya (very thirsty: 85.7% vs 62.2% [P < .0001]; decreased skin turgor: 50.0% vs 34.9% [P = .025]). In The Gambia, younger children appeared more likely to receive an inappropriate antibiotic prescription. In Kenya, children with fever (14.3% vs 6.2%; P = .024) were more likely to receive an inappropriate antibiotic prescription, whereas in The Gambia, children with belly pain were less likely (38.8% vs 52.0%; P = .004).

Factors Associated With Antibiotic Prescription by Multivariable Analysis
The trends observed in the univariate, bivariate, and multivariable analyses were consistent and generally produced concordant statistically significant associations with some exceptions. For example, in the multivariable analyses, there was no association between antibiotic prescribing and age or fever, whereas children who appeared irritable in Kenya were marginally less likely to receive a prescription.

DISCUSSION
Our findings indicate that nonadherence to IMCI guidelines for treatment of diarrhea is prevalent at the study sites in sub-Saharan Africa, manifesting as prescription of antibiotics that are ineffective or unnecessary, or not prescribing antibiotics when they are indicated. Overall, 88% of children with dysentery were prescribed antibiotics, but only 59% of them received the recommended first-line therapy with ciprofloxacin. In addition, 77% of MSD cases who had no apparent indication for therapy were prescribed antibiotics. Unindicated antibiotic prescription was associated with specific clinical features, which varied somewhat from site to site, suggesting that cross-cutting antimicrobial stewardship is needed that is informed by local clinical practices.
Overall, most children with watery diarrhea in The Gambia (58%) and Mali (98%) and 25% in Kenya received an antibiotic that the guidelines deemed unnecessary, corroborating results from other studies indicating the widespread nature of this problem in primary care across low-and middle-income countries (LMICs) [22], with some variation in the frequency and predisposing factors by country [23]. In many instances when antibiotics were prescribed without an apparent indication, a drug was selected that has limited utility for diarrheal pathogens. For example, metronidazole was prescribed for nonbloody diarrhea in approximately 28% of episodes in The Gambia and Kenya and 77% in Mali. Metronidazole is not indicated for pediatric diarrhea unless the child has dysentery that fails to improve with antibiotics directed against shigellosis, and it is not without adverse reactions [24]. Inappropriate prescription of cotrimoxazole was also common at all sites among children with watery diarrhea. The indication for empiric treatment with cotrimoxazole is limited to dysentery, and because resistance is now widespread, it should only be given when susceptibility is known or expected based on local data. Future studies should explore these drivers of decisions on antibiotic prescription to support the development of effective strategies that encourage appropriate, effective, affordable, evidence-based, guideline-driven antibiotics use for management of diarrheal diseases. Inadequate laboratory capacity, a common problem in Africa, has been suggested as a barrier to effective decision making by clinicians in LMICs regarding the need for antibiotic therapy [25][26][27]; however, this may not fully explain our findings. Although bacterial culture results could be accessed by clinicians caring for VIDA participants, the results were not available for several days after the patients had been treated and so were rarely used for decision making. Recognizing the limitations of culture to inform therapy for dysentery, IMCI advises that antibiotic treatment be guided by local susceptibility patterns, but numerous studies suggest this is not being done. For example, in 2014, Odhiambo et al [28] reported that clinicians at the leading referral hospitals in proximity to the VIDA sites in Kenya did not utilize the existing laboratory capacity on enteric pathogen susceptibility to guide patient management [29][30][31][32][33]. Clinic-based surveillance for diarrheal pathogens was established in this area of Kenya in 1997. Despite a series of publications from these sites demonstrating high rates of AMR to the antibiotics commonly used to treat dysentery [25][26][27], only approximately 44% of Kenyan children with dysentery in VIDA were treated with the recommended first-line ciprofloxacin therapy. In Mali, no dysentery cases were prescribed ciprofloxacin, whereas metronidazole was prescribed as first-line therapy for 83% of dysentery cases despite its lack of efficacy for shigellosis. Although emergence of resistance to cotrimoxazole prompted the current WHO recommendation that it only be used when local susceptibility is documented, it was prescribed for 68% of dysentery episodes in Mali. In fact, more than 80% of Shigella isolates were resistant to cotrimoxazole during GEMS [20] and VIDA [34]. Prescription of these ineffective drugs for unnecessary indications presents risk without benefit and may, in part, explain the frequent poor treatment outcomes and antibiotic resistance in these countries and other LMICs [27].
Antibiotics were prescribed significantly more often in our study when selected clinical findings were present that do not carry an indication. For example, in Mali and Kenya, antibiotic prescription was 2-4 times more likely when a healthcare provider observed signs of dehydration (dry mouth, excessive thirst, and/or decreased skin turgor). In The Gambia and Kenya, both rural sites, children with cough in the absence of respiratory distress were more likely to receive antibiotics than those without these findings. These observations corroborate those of Rhee et al [22] in Kenya and Ahmad et al [32] in India that cough without IMCI criteria for pneumonia was frequently associated with unnecessary antibiotic prescribing among children with diarrhea.
The main limitation in our current study is that it was not originally designed to evaluate the decision process in antibiotic prescribing, which could have resulted in misclassification of some cases as not warranting antibiotics. Important factors that this study did not explore include the prescribers' knowledge, attitude, and rationale for their antibiotic prescription decisions or barriers to implementing guidelines such as drug cost or availability.