Not studying a relevant, high-risk, or representative sample

Use population-based recruitment or over-recruit high risk subgroups

Do not include too many exclusion criteria

Not thoroughly understanding outcomes or how they come about, e.g., no knowledge of mediators, conflicting or ambiguous results, or inadequate control conditions to rule out alternative hypotheses

Assess broad set of outcomes including possible negative ones

Include measures of hypothesized mediators

Conduct subgroup analyses or include different assessment points

Select stringency of control condition to fit your question

Involve potential adoptees (real world) settings beginning with initial design phase

Approach a representative or broad group of settings early on when you can still revise

Assess if Tx is too complicated, too intensive, or not compatible with other duties

Involve potential (non-research) intervention agents in program design

Include maintenance phase in both protocol and in evaluation plan

Leave Tx behind after study and plan for institutionalization