Investigating effect versus impact in a participatory action research project: Building end-of-life communication competence in residential care facilities for older people.

Background: Active involvement of residents, family members and staff in proactive end-of-life conversations remains uncommon in residential care facilities for older people, although these are common settings for the end-of-life. This study derives from a Participatory Action Research project with an intervention aimed at developing Death Literacy among residential care staff, to promote implementation of proactive end-of-life conversations. Objectives: This article has a two-fold aim: (1) to investigate effects and forms of impact derived from this project intending to improve Death Literacy, and (2) to discuss challenges in this investigation and contribute to critical debate and further inform the use and development of the Death Literacy concept and related index (DLI). Methods: We combined quantitative analysis of pre-post intervention DLI survey data from staff and qualitative framework analysis of official documents along with transcribed interviews and workshop data from various staff groups. The framework was based on a matrix with Death Literacy features on one axis and forms of impact on the other. Results: The DLI showed no positive change over time. Effect evaluation was challenged by high baseline mean scores and ceiling effects on the DLI, and issues associated with the use of a participatory research approach, e.g. baseline differences between groups related to recruitment strategies. However, qualitative findings indicated all three forms of impact described in Kleijberg et al.'s model: impact on individual and group development, action-oriented impact, and strategy-oriented impact. These were seen to be related to all Death Literacy features, most notably experiential learning and social action. Conclusion: Features of Death Literacy were relevant in understanding impact, although the usefulness of the DLI to measure change over time was limited in this study. A matrix framework provided nuanced insight beyond either model alone. Congruence of different methodological assumptions should be carefully considered when designing investigations of effect and impact of complex interventions with limited researcher control.