Table 1

Care management interventions for type 2 diabetes

Evidence user	Perspective	Framing of the question	Clinical heterogeneity	Statistical heterogeneity	Quality of evidence (certainty in evidence)*
Policymaker/payer	Should I cover (pay for) this intervention?	To what extent is this bundle of interventions effective?	No concern, we are interested in the effect of the whole bundle.	No concern, we are interested in the effect of the whole bundle.	High
Physician and patient	Which part of the intervention should we provide to this patient?	Which component of the intervention is most effective?	Variability of components is concerning particularly if a component is not well described or not feasible in own practice.	No concern if metaregression explains heterogeneity (ie, demonstrate the effect of each component).	Moderate (if statistical heterogeneity was explained) Low (if statistical heterogeneity was not explained)
Diabetes counsellor/educator	How can I apply the intervention?	Under what circumstances does the intervention work best?	No concern	No concern	The effectiveness evidence is inadequate to judge certainty. Additional contextual evidence from qualitative research is identified leading to high certainty.†

Evidence: meta-analysis demonstrated that the interventions significantly reduced HbA1c (−0.21%, 95% CI −0.40% to −0.03%). Across studies, there was important clinical heterogeneity (varying components of the intervention) and statistical heterogeneity (varying effect size with I²>75%). Following a traditional approach, the evidence may be graded down twice for indirectness (numerous components make application challenging) and inconsistency (heterogeneity of effect).
*The focus in this example is only on the domains of indirectness and inconsistency. We assumed that there were no concerns in other quality of evidence domains such as precision, risk of bias or concerns about HbA1c being a surrogate outcome.
†Contextual evidence shows that a significant amount of qualitative studies are available to provide context and details on implementation. For example, studies demonstrated that before providing education in diabetes, we should address negative feelings that individuals have (anger about the initial diagnosis, and subsequently negotiate loss of control of a person's own lifestyle and environment).7 Qualitative studies also provided explanation to lack of adherence to a programme (individual, organisational and content factors).8 Addressing these factors is essential to programme success.
HbA1c, glycated haemoglobin.