TRiPLaB is a multisite, longitudinal, mixed methods case study. Currently, we are working with NHS Bradford and Airedale (an NHS commissioning and community provider organisation) to translate research-based findings into practice in the areas of maternal mental health and stroke care, and with Leeds Partnership Foundation Trust (a provider of mental health services) to enhance the implementation of recent and relevant NICE guidance.

Each case study will have three sequential phases (see Figure 1): the findings from the development phase (phase one) lead into the implementation phase (phase two), and the outcomes of this are assessed in the evaluation phase (phase three). The three-phase approach has been informed by the Medical Research Council's framework 11 for developing and evaluating complex interventions, acknowledges the need to use theory in planning and analysis, recognises the importance of local context, piloting, and evaluating intervention components, and the use of multiple outcome measures to evaluate intervention effectiveness. The three phases are summarised below.

Phase one is a development phase in which the innovation that is the focus of each case study will be selected and its key characteristics mapped. Theory-informed factors hypothesised as influential in health professionals' adoption of the selected innovation into routine practice are explored and mapped.

Phase two is an implementation phase in which tailored behaviour-change interventions are developed piloted and delivered using personnel from TRiPLaB and its partner organisations.

Phase Three is an evaluation phase in which changes in structure, process, and outcome are described and evaluated. We will be looking at change both within and, towards the end of the programme, between case studies.

TRiPLaB will use the resources of the Centre for Reviews and Dissemination (CRD) to increase the accessibility of research evidence to decision makers (particularly commissioners) in the NHS. Primarily, we will do this by using tailored briefings relating research evidence to specific decision problems and context in Bradford and Leeds. These 'evidence briefings' will be based on existing sources of synthesised and quality assessed evidence - for example, CRD's databases of systematic reviews (DARE) and economic evaluations (NHS EED). We will develop and implement methods for producing and disseminating evidence briefings and evaluate their perceived usefulness, costs, and use by decision makers.

Development of the intervention will be systematic, specifying intervention objectives, developing specific implementation strategies to satisfy these objectives, and piloting strategies to assess their likely impact and test how they will be received by the health professionals. This piloting and modelling prior to rolling out implementation strategies/behaviour-change interventions is a necessary prerequisite stage 11. The objectives and design of the intervention will be determined by the outcome of the development phase, in particular the results of the planned multilevel modelling. The selection and design of the actual intervention components will be informed by existing systematic, and other, reviews of the relevant literature.

Having decided on the innovation in phase one and possible implementation strategies in phase two, we will make the final choice on our implementation approach with reference to the idea of 'policy' cost effectiveness 14. Summary data on: the innovation from Phase One (net cost per patient and likely health gain per patient); the implementation strategies under consideration (net cost of planned implementation and likely change in adoption/adherence); and local scale factors (for example, the number of NHS organisational units involved and number of patients targeted) will be combined to arrive at a policy cost effectiveness figure for each option. The combination with the highest cost effectiveness will be the option pursued.

The failure to adequately describe interventions in the context of research and the commensurate reduction in others' ability to then use successful programs -- or conversely, avoid making the same mistakes as unsuccessful ones -- is common in healthcare research 15. For each of the case studies in the TRiPLaB program we will describe: the intervention and its component parts in sufficient detail that others could reproduce it; why the specific intervention was chosen; and a fidelity measure of how well the intervention was delivered. For example, if we undertake educational outreach or training as a component of an intervention, we will detail how many sessions each unit of analysis receive, and when and where the training took place.

Following the recommendation to conduct exploratory trials prior to embarking on more definitive randomised controlled trials 11, TRiPLaB will employ three different methods to evaluate the impact of the tailored implementation strategies delivered in each case study. The findings from these evaluation measures will inform (if worthwhile) later randomised controlled trials. The three methods to be used are: interrupted time series analysis of either tailor-made audit data or routinely collected data to estimate the impact of the intervention upon suitable process of care and outcome measures; comparison of pre- and post-intervention scores of survey-gathered measures of individual attitudes, team culture, and changing nature, composition, and size of social networks; and a qualitative process evaluation of why the intervention worked (or did not work).

Alongside these three primary evaluation methods we will also collect cost data on the resources used in the delivery of implementation approaches. The micro costs 16 associated with each strategy will be estimated alongside the extent of behavioural change achieved to arrive at summary estimates of implementation cost effectiveness 14 for each of the case studies in the programme.

Interrupted time series designs compare multiple 'before and after' (the introduction of a change strategy) measures to detect whether an intervention has had an effect over and above any underlying trend in the data 17. Time series analysis has been used as a technique for evaluating the effectiveness of health care interventions 18. In the case studies, routinely collected process (health professionals' adoption of the innovation) and health outcome measures (dependent on the innovation selected) will provide the multiple time points necessary to perform a time series analysis. Time, possible seasonal trend, and possible upward trend, commonly occurring following the introduction of a new innovation 10, will be modelled into the analysis. This will be the primary outcome measure for each case study.

The interrupted time series analysis will estimate the impact of the intervention upon process of care and health outcome measures; however, a comparison of pre- and post-intervention scores is also necessary to estimate whether the intervention successfully changed the factors (for example, individual attitudes, social networks and team culture) in the underlying theoretical framework that it was designed to target (based on the data synthesis through multilevel modelling in TRiPLaB's development phase in each site). This 'meditational analysis' 9 is critical when evaluating the theory used to develop change interventions, as it will inform our understanding of why an intervention either works or fails to work in the ways we intended.

Qualitative interviews with health professionals receiving the intervention will enable an exploration of their perceptions of what worked and what did not work in the intervention, providing insight into the 'black box' of intervention effectiveness 19. In combination with the measure of fidelity taken during the implementation phase, these qualitative interviews comprise a process evaluation of the intervention, addressing recommendations to monitor intervention delivery and receipt by participants 1120. The data will be analysed using a framework approach 21: familiarisation with the data, identification of a thematic framework, indexing, charting, and finally, mapping and interpretation with reference to the overall aim of TRiPLaB as well as the themes revealed by the data.