Table 2 .

Key tasks for optimising an intervention: example of computer support for assessment of familial risk of cancer in primary care

Key task	Example
Identify key processes and outcome of intervention	The intervention sought to optimise the use of cancer genetics services available to primary care by reducing inappropriate referrals. A central problem was that referral guidelines are complicated to use and poorly implemented in general practice17
Identify mechanisms by which intervention will lead to improved outcome	The intervention would provide tailored advice on referral for individual patients by using computer decision support to implement referral guidelines. Systematic reviews have shown that this mechanism helps the implementation of clinical practice guidelines. Training would help practitioners to use the software, and paper questionnaires would help patients provide information to increase the accuracy of assessment
Identify barriers to application of intervention (which may manifest as rate limiting steps)	Previous studies of computer support suggested that lack of use during consultations was an important rate limiting step. Lack of training and poor software design contributed to this. A qualitative study of the prototype software (RAGs) using simulated patients identified important contextual barriers (eg, time needed for data entry, and loss of doctor-patient interaction) and helped development of the software to minimise adverse effects (eg, providing an interface that encouraged shared use by doctor and patient and avoided frightening “high risk” messages appearing suddenly on screen)18
Quantify potential for benefit and estimate the likely effect size	An experimental randomised block design study with simulated patients showed the software could improve risk assessment and double the number of appropriate referral decisions (the intermediate outcome)19
Refine the target group to take account of its likelihood of responding to the intervention	The potential target group was all primary care practitioners. The group was refined to a single general practitioner in each practice to maximise the effect of the training and increase the frequency with which they used the software
Consider the best achievable combination of intervention components and intensities	Findings from previous related research and the qualitative and experimental studies of the prototype software all helped with this. A concern was to optimise use of the software during consultations so, in addition to standard intervention and control arms, an exploratory trial included an “adaptive arm,” which permitted the software and protocols to be varied during the trial in response to comments from practitioners and reasons identified for low use20