Care-seeking and appropriate treatment for childhood acute respiratory illness: an analysis of Demographic and Health Survey and Multiple Indicators Cluster Survey datasets for high-mortality countries

According to the 2010 Global Burden of Disease estimates, acute lower respiratory infection (ALRI) is the leading cause of death among children aged under five years in developing countries [1]. Although viruses are the more common etiologic agent of ALRI in children [2], bacterial pathogens are associated with higher risk of morbidity and mortality [3] and viral ALRIs are often followed by bacterial infections [4]. While two of the most common etiologies of bacterial ALRI, Haemophilus influenza type b (Hib) and Streptococcus pneumoniae, can be partially prevented through vaccination, vaccine coverage is not yet widespread in resource-limited settings [5]. Vaccines are not available for any of the other bacterial pathogens that cause ALRI. Until effective vaccines become broadly available and vaccination coverage is high, proper clinical management remains the most important method of preventing childhood mortality from ALRI.

Appropriate clinical management of ALRI consists of several elements: a caregiver recognizing the signs and symptoms of severe infection, the child being brought to a healthcare facility, the healthcare worker making a correct diagnosis, and the availability of appropriate therapy. Each of these components provides one or more pathways by which a child with ALRI might not receive appropriate treatment.

As a clinical syndrome, ALRI is characterized by fever, cough, difficult breathing, and rapid breathing although each of these may be variable in young children and difficult for caregivers to evaluate. In one analysis, only 20% of caregivers were able to identify difficulty breathing and fast breathing as characteristic symptoms [6].

Severity of ALRI upon presentation predicts response to treatment [7], and therefore prompt care-seeking by the parents of ill children is critical. However, delays in care-seeking often occur and can be related to cost, non-recognition of the severity of illness, differences in the perception of the cause of disease, or geographic barriers [8‐11]. Additionally, poor healthcare infrastructure and/or lack of appropriate care and medicine contribute to delays [12].

Correct diagnosis of ALRI is often difficult in areas that are endemic for diseases with similar clinical presentations, such as tuberculosis or malaria [13]. Additionally, severity must be accurately assessed and antibiotic treatment and oxygen therapy provided, if necessary. The World Health Organization (WHO) currently recommends intramuscular ampicillin and gentamycin for inpatients with very severe pneumonia, and oral amoxicillin or co-trimoxazole for less severe pneumonia and outpatients [14]. Other common antibiotics for severe ALRI include oral cephalosporins and macrolides [15]. When needed, antibiotic therapy has dramatic effects on child survival; case-fatality rates for treated ALRI are under 1% while untreated ALRI case-fatality rates may reach over 10% [16]. However, there are a number of barriers to receiving appropriate antibiotics at the clinic level, including stock-outs, understaffing, prescribing patterns of health workers, and diagnostic overlaps with other infections [11, 12, 17, 18].

Given these challenges, both parental care-seeking and treatment patterns within facilities vary widely across countries. The WHO has estimated that globally, only 30% of children with ALRI are appropriately treated with antibiotics [19], though this proportion is likely to be even lower in high-mortality countries.

To effectively reduce the burden of ALRI mortality, it is important to understand patterns of care-seeking, as well as to quantify the number of cases of untreated ALRI within high mortality countries. However, direct comparisons among countries with high mortality are made difficult by a lack of routinely collected estimates that use similar methodology. Two exceptions to this are the Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS). These household-level surveys measure health and demographic indicators approximately every 5 years in developing countries using nationally representative samples. They are particularly useful for aggregate estimation of burden because they use a common, standardized approach. DHS data provides information on fertility and health topics, while the MICS is focused on the state of women and children to inform policy and program decisions. DHS and MICS surveys have been intentionally developed to be comparable to each other.

To date, no study has aggregated DHS and MICS indicators to compare ALRI care-seeking and treatment outcomes globally. Though DHS and MICS data are the best source of comparable data, many high-mortality countries have not conducted a DHS or MICS. Fortunately it is possible to use the information from countries which have been surveyed to provide information on those which have not yet been surveyed. This study uses the most recent DHS and MICS data to address two objectives: 1) gain an improved understanding of patterns of care-seeking and antibiotic treatment among countries with DHS and MICS data, and 2) determine whether country-level factors can be used to estimate the burden of cases of untreated ALRI in all high-mortality countries.

As of August 2012, 17 countries within Africa and Asia had DHS datasets with information about care-seeking patterns for ALRI among children under 5 years of age. Eighteen datasets containing care-seeking information were available from the MICS datasets. For the predictive model, eleven countries had DHS datasets with information about antibiotic treatment of ALRI (two of which did not have care-seeking data), while 15 additional countries had antibiotic treatment information in MICS reports.

Antibiotic treatment is an inexpensive, effective way to increase child survival for children with ALRI in resource-limited settings. In this study, we estimated that over 66 million children in high-mortality countries who could benefit from antibiotic therapy are not treated, and that there is wide variability across countries in care-seeking and antibiotic treatment patterns for children with presumed ALRI.

Cultural and political factors may influence care-seeking behaviors. Some of the lowest proportions of parents seeking any care for their child’s illness were in Central African Republic, Burundi, and Sierra Leone. These were also the countries with some of the highest percentages of parents seeking public sector care. It is possible that the low level of care-seeking in these countries is related to a relative lack of private sector healthcare. Each of these countries has experienced extensive political instability which may have influenced the overall availability of care and the ability of parents to seek care. On the opposite side of the spectrum, Bangladesh, Tanzania, Syria and Pakistan each had high levels of access to care, but low percentages accessing public sector care. Although the private sector is well established, it is generally less regulated, and sales of inappropriate drugs are a recurrent problem [24].

In the countries with available DHS and MICS data, antibiotic treatment levels among children with presumed ALRI were quite low. There was wide variation in antibiotic coverage across countries, with the lowest percentage occurring in rural areas.

To our knowledge, no other studies have attempted to predict the burden of untreated pneumonia in high-mortality countries for which no publically available data exist. This prediction model provides meaningful estimates of the number of untreated children for mathematical models seeking to estimate how much disease could be resolved if treatment guidelines were followed, and how many prognoses can be improved. Although the DHS and MICS provide the most extensive methodologically comparable collection of datasets for global estimation, there are several limitations that may bias estimates of care-seeking and proportion of cases treated with antibiotics. The DHS data available spanned several years, so the model findings are most applicable to the years between 2005 and 2010. Due to the nature of self-report in the DHS, it is also likely that the DHS definition of ALRI is neither sensitive nor specific for radiographically confirmed pneumonia, and thus we would hope to observe fewer than 100% of cases treated with antibiotics [25]. Recent studies have posited that the antibiotic treatment estimates derived from DHS and MICS data cannot be accurate because of this low sensitivity and specificity of pneumonia recognition [26]. Furthermore, the DHS and MICS rely on maternal recall of antibiotic treatment, the challenges of which are well documented in the literature [26]. However, in this prediction model, we applied estimates of the proportion of children receiving antibiotics derived from DHS/MICS to CHERG estimates of pneumonia incidence, which is a more specific characterization of pneumonia burden than in the DHS. With large enough sample sizes, we expect that bias in the DHS/MICS antibiotic treatment proportions would only be introduced from differential treatment provision based on the severity of illness of those ALRI cases included in the DHS. The extent and direction of this bias is unclear, and further study is necessary.

The prediction model estimated that 66 million cases of ALRI go untreated every year. While the best model was fit from the data which were publically available, the predicted estimates from the linear model for countries with available DHS or MICS data were often much closer to the mean than the observed data. Because of this, the model is limited in its capacity to estimate the burden of untreated ALRI for individual countries. However, the model reflects improved estimates of the total burden for these high mortality countries by encompassing predictive indicators for countries without data. With the sampling variability of DHS estimations, particularly when combined with CHERG predictions, the estimates for each country have a wide predictive range; therefore the absolute number of untreated cases for each country should be interpreted with caution.

ALRI is a major contributor to child mortality in developing countries. Our analysis shows that there are large numbers of children with ALRI who are not treated with antibiotics, representing an enormous missed opportunity to prevent mortality. While the WHO estimated that globally 30% of children are untreated for pneumonia, our data suggest that, in high mortality countries, over 50% of children may be untreated. It is likely that this high proportion of untreated children is adding to the overall mortality rate of these countries. Additionally, there is likely unnecessary usage of antibiotics which, given the multibillion dollar global market in antibiotics, represents wasted funds that could be redirected to support evidence-based health care practices. Integrated initiatives such as community case management of pneumonia have recently been demonstrated to be effective for appropriately providing treatment to children with pneumonia [27‐30]. These community-based interventions promote both care-seeking and appropriate antibiotic treatment, and ultimately reduce mortality due to untreated ALRI.