Survey data analyses focused on the calculation of descriptive statistics, such as frequencies, central tendencies, and distributions for all questions. Standard difference of means and difference of proportions tests were used for making statistical comparisons between sustained and not sustained coalitions. In cases where multiple items were used to assess a concept, the factor structure of the data was analyzed and an additive index or factor variable was created. Confirmatory factor analysis was used to provide evidence that the multiple items load together and are related to the latent concept. When the confirmatory factor analysis supported the latent concept, principal components analysis was used to extract an empirical indicator, using the Bartlett method. These indicators were then utilized as predictor variables in subsequent analyses (e.g., the influence of coalition leadership strength on sustainability). This procedure was used for the validated scales included in the questionnaire from past research as question ordering effects and the application of the scale to a slightly different population can jeopardize the statistical integrity of the scale.
Logistic regression models were employed to assess specific hypotheses derived from the conceptual framework (see Exhibit 3). Use of multivariate models allowed the explanatory power and fit of theories from the literature to be assessed for the HCAP population. Significance was determined through two-tailed tests and statistical significance is noted at the p<.10, .05, and .01 level. All references to significant differences refer to statistical significance. Analyses were conducted in STATA 10. Appendix B contains the results from the survey.
Qualitative data analysis of the 25 phone interviews was conducted using QSR NVivo 9 (NVivo). NVivo facilitated the identification of common themes across community coalitions in addition to major differences between sustained and disbanded coalitions. All notes from the interviews were uploaded into NVivo and sorted into folders based on the coalition statuses of Not Sustained, Sustained and Expanded, and Sustained and Not Expanded (see Exhibit 3). Based upon the topics addressed in the interview protocol, nodes (containers for categories and coding that represent concepts, processes, people, abstract ideas, or places, etc.) were developed to capture data from the transcribed interviews and subsequently coded (linking text to nodes). From this analysis, a list of key themes was developed for each topic area that emerged as particular to sustained coalitions and not sustained coalitions. Sub-nodes were made according to these themes to capture additional data specific to these themes.
Site visit data analysis began with the site visit teams engaging in a process of respondent validation to confirm the key information and themes emerging from the visit and the framing of these key themes in the site visit report. First, NORC reviewed the notes from each site visit, identified significant quotes and comments, and drafted the reports. The site visit teams conducted debrief conversations with each grantee director either at the end of the site visit or shortly thereafter by telephone to confirm the interpretation of key findings and ensure that important information was not omitted. Reports highlighted important facilitators for the coalition’s sustainability and lessons learned. Once internal reviews were completed, the draft reports were shared with the lead organization at each site for verification.
Results from the surveys, key-informant interviews, and site visits were analyzed to identify overarching themes and differences among sustained coalitions and not sustained coalitions and are integrated with the key survey findings throughout the report.