Ignoring 'downstream infection' in the evaluation of harm reduction interventions for injection drug users

Harm reduction interventions to reduce blood-borne disease incidence among injection drug users (IDUs). A common strategy to estimate the long-term im pact of such interventions is to examine short-term incidence changes withi n a specific group of individuals exposed to the intervention. Such evaluat ions may overstate or understate long-term program effectiveness, depending upon the relationship between short-term and long-term incidence and preva lence. This short paper uses steady-state comparisons and a standard random -mixing model to scrutinize this evaluation approach. It shows that evaluat ions based upon short-term incidence changes can be significantly biased. T he size and direction of the resulting bias depends upon a simple rule. For modest interventions, such analyses yield over-optimistic estimates of pro gram effectiveness when steady-state disease prevalence exceeds 50% absent intervention. When steady-state prevalence is below 50%, such analyses disp lay the opposite bias.