In this study 94% of therapeutic interventions were evidence-based by Ellis score. It may seem encouraging that more than 90% of therapeutic decisions in a referral paediatric hospital of a developing country are evidence based. However, the level II of evidence from Ellis includes interventions based in cohort studies, case-control studies, case series, expert's opinion, and even those based in basic sciences. We attempted therefore to classify the primary interventions according to more specific criteria. Using the Oxford classification, 75% of therapeutic interventions were based in expert opinions or in basic sciences (Grade of Recommendation D).
Some limitations of our study include the possible author' bias when assigning the primary diagnosis and primary intervention. The assignment of diagnosis by the clinician may have been influenced by both the choice of treatment and the available evidence. Only one of us classified the primary intervention. In addition, we evaluated only a primary intervention for a single primary diagnosis. Actually, many patients had more than one diagnosis and obviously more than one therapeutic intervention.
We used as evidence-base for rating the interventions assessed in our study, guidelines and evidence-based resources published in the developed world. This raises the issue of whether they are fully applicable to our setting. The most prevalent conditions found were diarrhoeal dehydration and community acquired pneumonia. For both of them we used British produced guidelines [20
] because there were not Cochrane reviews on them and because the guidelines fulfilled recommended criteria for methodological quality of published guidelines [11
]. The main recommendation of the guidelines on diarrhoea favours rapid oral rehydration over intravenous rehydration for children with mild to moderate dehydration [20
]. This recommendation is based on several studies performed in both developed and developing countries and thus it can be applicable to both settings. The only concern on the applicability from setting to setting is that related to the osmolarity of the oral rehydration solution (ORS). The guidelines recommend a solution with 60 mmol/l of sodium, whereas a recent expert consensus found sufficient evidence to recommend the universal use of an ORS containing 75 mmol/l of sodium [49
Regarding community acquired pneumonia, the British guidelines recommend antibiotic treatment for all children with pneumonia, due to the difficulties in identifying the aetiology, and they also specify criteria for hospitalization [21
]. These recommendations are in agreement with the World Health Organization published guidelines [50
]. The main difference is that the WHO guidelines rest on fast breathing and chest retraction for the diagnosis of pneumonia, whereas the British guidelines emphasize the role of chest x-rays. Chest x-rays are widely available in referral hospitals in developing countries and thus they should be used in addition to the clinical findings.
We acknowledge that the evidence derived from studies performed in developed countries should be translated with caution to developing settings. However, when the native research is scarce or of low quality, we think that the transfer of knowledge from the developed countries is an acceptable approach, as far as the particular characteristics of patients in developing countries are considered on an individual basis.
Dehydration due to diarrhoea and pneumonia were the most frequent diagnoses. Oral rehydration for diarrhoeal dehydration and antibiotics for pneumonia are considered as interventions with sufficient evidence for implementing them widely [9
]. In our study, all children with mild to moderate dehydration were treated with slow intravenous infusion, and most children with uncomplicated community acquired pneumonia received intravenous antibiotics.
In addition to their enormous potential for saving lives, outpatient antibiotic therapy for pneumonia and outpatient oral rehydration can drastically reduce the rate of hospitalizations, the hospital stay, the hospital mortality rate, and the costs incurred. At our hospital, the mean stay time for hospitalized children is 4.7 days, and the mean crude mortality rate is 3.6%. We estimated the cost of managing hospitalized children with pneumonia and diarrhoeal dehydration as US$ 10.6/day and US$ 8.6/day, respectively. These costs are referred only to hospital bed and laboratory tests. A substantial amount of money could be saved treating these conditions on an outpatient basis.
We planned the dissemination of our results among the hospital policy makers. The tools that will be suggested for improving the standards include the development and systematic application of locally produced guidelines and/or the adaptation of published guidelines. A useful alternative that has been experienced for several years at our inpatient ward unit is to make available personal computers connected to Internet for attending physicians, residents and interns, and to encourage the use of online evidence-based resources. This last alternative may work better, particularly where there are motivated physicians who are able to lead the efforts for improving the health care standards. However, the ultimate decision to systematically introduce and monitor the suggested changes will rest on hospital managers. Such changes should also depend on taking into account the role of several other determinants of the clinical decision making by individual practitioners, such as continuous training, motivation, time, availability of drugs, equipments and supplies, supervision, and long-term health system strengthening strategies.