Table 2 Summary of the GRADE system for guideline development

We used the GRADE system to develop the recommendations since it provides an explicit and transparent assessment of the quality of evidence and the strength of recommendations for clinical questions.

PICO, an acronym that represents four attributes (Population, Intervention, Comparator and Outcome) is widely used in formulation of clinical questions. The PICO elements for our 5 clinical questions of interest for childhood pneumonia in the Kenyan treatment guidelines are given below.

Population: Kenyan children aged 2 - 59 months meeting WHO criteria for:

Intervention (proposed new treatment)

Comparator (standard treatment)

Outcomes of interest

i. non-severe pneumonia

I. Amoxicillin

Co-trimoxazole

i. Mortality

ii. severe pneumonia. (HIV-unexposed without severe malnutrition)

II. Amoxicillin

Benzyl penicillin/ampicillin

ii. Treatment failure

III. Benzyl penicillin/ampicillin plus gentamicin

Benzyl penicillin (or amoxicillin)

OR

iii. very severe pneumonia. (HIV-unexposed without severe malnutrition)

IV. Benzyl penicillin/ampicillin plus gentamicin

Chloramphenicol

Time to resolution of signs of pneumonia

V. Ceftriaxone

Benzyl penicillin/ampicillin plus gentamicin

iii. Cost

GRADE classifies quality of evidence into four categories (high, moderate, low, or very low) and recommendations (for or against treatments) into two grades (strong or weak). With study design as the starting point, RCTs ranking highest and observational studies lowest, it allows for downgrading of the quality of evidence in the presence of factor related to study limitations, consistency, directness and publication bias. It also allows for upgrading the quality of evidence in the presence of large treatment effects, a dose-response gradient and residual confounding likely to underestimate the true effect. The current revised interpretation of the GRADE levels of evidence is shown below.

Quality level

Interpretation

High

Highly confident that the true effect lies close to that of the estimate of the effect

Moderate

Moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different

Low

Limited confidence in the effect estimate: The true effect may be substantially different from the estimate of the effect

Very low

Very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect

Unlike other guideline development tools, GRADE uniquely separates judgments on the quality of evidence and the strength of recommendations - recognizing that making recommendations involves tradeoffs between benefits and harms, and contextual factors (e.g. costs, baseline risk of population, clinician values and preferences, etc). Thus, using the above assessments, we classified recommendations into categories summarized in figure 2.