Doing includes using iterative approaches such as Plan-Do-Study-Act change cycles or incremental steps instead of implementing all components in all planned areas within a specified period (Ashok et al. 2018; Dy et al. 2015). The original CFIR (Damschroder, Aron, et al. 2009) elaborated, stating that taking an incremental approach leads to breaking the innovation down into manageable parts that can be implemented incrementally (Damanpour 1991). The ability to implement an innovation incrementally (sometimes referred to as divisibility (R. P. Grol et al. 2007)) can help decrease perceptions of innovation complexity and thus implementation difficulty. Thus, more complex innovations especially benefit from incremental implementation (Damanpour 1991; Grossman 1970; Normann 1971). In addition, incremental approaches allow deliverers to have enough time to do their work and to learn new skills associated with the new innovation (Helfrich, Weiner, et al. 2007). Successes in early increments of the implementation help increase confidence, give an opportunity to adjust the innovation or implementation (see Implementation Process: Adapting), and gain new “believers.” Usability testing (with deliverers and recipients) promotes successful adaptation of the innovation (Feldstein and Glasgow 2008). Dry runs (simulations or practice sessions) to allow team members to learn how to use the innovation before going live (Edmondson et al. 2001) is also beneficial.
Doing also includes trialing, which allows for time to build experience and expertise, reflect upon and test the innovation (Rycroft-Malone, Kitson, et al. 2002), gain confidence, and build an environment of psychological safety (Edmondson et al. 2001). The ability to trial is a key feature of the PDSA approach, allowing users increased coordination to manage interdependence (see Innovation: Innovation Trialability) (Rabin et al. 2008; Leeman et al. 2007). Implementers or deliverers need to be able to stop the innovation and reverse its effects if it causes problems or is ineffective (Feldstein and Glasgow 2008). Results from trials will provide needed information about how best to implement to other units to minimize workflow disruption.