|Home | About | Journals | Submit | Contact Us | Français|
Effective planning for community health partnerships requires understanding how initial readiness—that is, contextual factors and capacity-- influence implementation of activities and programs. This study compares the context and capacity of drug and violence prevention coalitions in Mexico to those in the United States. Measures of coalition context include community problems, community leadership style, and sense of community. Measures of coalition capacity include the existence of collaborative partnerships and coalition champions. The assessment was completed by 195 members of 9 coalitions in Mexico and 139 members of 7 coalitions in the United States. Psychometric analyses indicate the measures have moderate to strong internal consistency, along with good convergent and discriminant validity in both settings. Results indicate that members of Mexican coalitions perceive substantially more serious community problems, especially with respect to education, law enforcement, and access to alcohol and drugs. Compared to respondents in the U.S., Mexican respondents perceive sense of community to be weaker and that prevention efforts are not as valued by the population where the coalitions are located. The Mexican coalitions appear to be operating in a substantially more challenging environment for the prevention of violence and substance use. Their ability to manage these challenges will likely play a large role in determining whether they are successful in their prevention efforts. The context and capacity assessment is a valuable tool coalitions can use to identify and address initial barriers to success.
Multiple community partnership models for health promotion have been shown to influence health outcomes (Butterfoss, 2007; Hawkins et al., 2012; Spoth et al., 2013). However, implementation of such models must be sensitive to context to succeed (Chang et al., 2013; El Arifeen et al., 2013). More specifically, contextual factors and initial coalition capacity can have a powerful influence on coalition ability to support the implementation of prevention programs and policies (Brown, Feinberg, & Greenberg, 2010; Brown, Feinberg, Shapiro, & Greenberg, 2013). Yet we know relatively little about how community and coalition characteristics vary across contexts, and thus how partnerships may need to adapt their approach to be successful (Chang, et al., 2013).
This study seeks to enhance understanding of the role of community context and initial coalition capacity in shaping coalition success through two primary aims: 1) to describe the psychometric properties of an assessment of coalition context and capacity administered in both English and Spanish; and 2) to utilize this assessment instrument to examine how the context and initial capacity of prevention coalitions in Mexico compare to those in the United States. The measures provide the infrastructure necessary for future research examining how context and capacity influence the course of implementation and the attainment of health outcomes. Context and capacity assessments can inform adaptations to the implementation process that respond to environmental demands. Further, identifying international context and capacity differences is an important first step in understanding how coalition models developed in the U.S. need to be adapted to be successful in Mexico.
Figure 1 outlines the theoretical model of coalition functioning guiding this study, which has emerged from and in turn guided several previous studies (e.g., Brown, Feinberg, & Greenberg, 2012; Feinberg, Greenberg, & Osgood, 2004; Greenberg, Feinberg, Gomez, & Osgood, 2005). In the model, coalition capacities are one facet of coalition functioning, which impact health outcomes through their influence on coalition ability to implement programs and policies (Brown, et al., 2010; Spoth, Guyll, Lillehoj, Redmond, & Greenberg, 2007). Community contextual factors are a separate component of the model hypothesized to influence all components of the causal chain from coalition functioning to health outcomes (Butterfoss & Kegler, 2009). This study focuses on specific coalition capacities and community contextual factors, as described further in the following sections.
Within a socioecological framework (Bronfenbrenner & Morris, 1997), community contextual factors may hinder or support a coalition successfully achieving change, and can thus be conceptualized as part of community readiness for a coalition approach to prevention (Burgess, McDonald, & Roberts, 1955; Carmack, 1965; Chilenski, Greenberg, & Feinberg, 2007; Foster-Fishman, Salem, Allen, & Fahrbach, 1999; Plested, 2006). Community readiness has been shown to relate to early coalition functioning (Greenberg, Feinberg, Chilenski, Spoth, & Redmond, 2007), and to other important community demographics such as poverty and population density (Chilenski, et al., 2007). The current study examines four constructs that fit within the community readiness framework: perceived community problems, participatory leadership, sense of community, and support of prevention (Chilenski, et al., 2007; Edwards, Jumper-Thurman, Plested, Oetting, & Swanson, 2000). All characteristics are expected to influence a community coalition's functioning and ability to achieve outcomes.
The community problems of interest are risk factors for youth substance use, such as poverty, lack of supervised activities for youth, and easy access to drugs (Feinberg, Ridenour, & Greenberg, 2007; Hawkins, Van Horn, & Arthur, 2004). Assessing the severity of these problems is important because they are potential points of intervention, which can be prioritized by the coalition and modified through collective action (Feinberg, 2012). As contextual factors, community problems such as poverty can also influence the implementation process and may be relevant even when the problems are not directly targeted by the coalition (Brown, et al., 2010).
A participatory leadership style is one where community leaders take into consideration the views of others when making decisions, and work to build consensus among involved parties (Magzan, 2011; Van Wart, 2013). If this type of leadership style is common within a community, coalition members are more likely to be comfortable employing such an approach in their own coalition. The ability to effectively employ a participatory leadership style contributes to community readiness by facilitating shared decision-making, minimizing inter-group conflict, and strengthening coalition functioning (Chilenski, et al., 2007; Greenberg, et al., 2007; Lehman, Greener, & Simpson, 2002; Nelson, Raskind-Hood, Galvin, Essien, & Levine, 1999).
A sense of community involves emotional attachment to the place and the people, the feeling that community members matter to each other, and that the community can meet their needs (McMillan & Chavis, 1986). This construct influences community readiness through its impact on coalition functioning and its sustainability (Chilenski, et al., 2007; Feinberg, Chilenski, Greenberg, Spoth, & Redmond, 2007; Greenberg, Feinberg, Meyer-Chilenski, Spoth, & Redmond, 2007; Perkins et al., 2011). Communities with a strong sense of community may be easier to engage in coalition efforts because residents believe their needs can be met through the coalition's efforts (Armenakis, Harris, & Mossholder, 1993; Lehman, et al., 2002).
Another aspect of community readiness is the recognition by community members that adolescent substance use is a problem in their community (Beebe, Harrison, Sharma, & Hedger, 2001; Edwards, et al., 2000; Plested, Smitham, Jumper-Thurman, Oetting, & Edwards, 1999). Positive attitudes regarding prevention predict enhanced adoption and implementation of prevention programs (Aarons & Palinkas, 2007; Feinberg, Chilenski, et al., 2007; Hagedorn & Heideman, 2010; Spoth, Clair, Greenberg, Redmond, & Shin, 2007; Spoth, Guyll, Redmond, Greenberg, & Feinberg, 2011).
Figure 1 displays several coalition capacities that have been theoretically and empirically connected with coalitions’ reaching their proximal and distal goals (Brown et al., 2013; Spoth, Guyll, et al., 2007; Welsh et al., under review). In our framework, coalition capacities are defined as knowledge, resources, skills, and influence that coalition members bring to the coalition through their participation (Raine et al., 2014; Shapiro, Oesterle, & Hawkins, 2015). Collaborative relationships and community champions are two coalition capacities of interest in this study because of the role they play in enhancing coalition influence on key segments of the community, encouraging cooperation and support (Nowell, 2009; Zakocs, Tiwari, Vehige, & Dejong, 2008).
A history and positive experience with collaboration will likely support success, as multi-sector collaboration supporting a comprehensive health promotion approach can be challenging (Brown et al., 2013a; Hausman, Becker, & Brawer, 2005). Research suggests collaboration capacity can enhance coalition functioning, promote systems change, and lead to improvements in youth outcomes (Brown et al., 2013b; Feinberg, Chilenski, et al., 2007; Nowell, 2009).
A community champion is someone influential, vocal, and effective in promoting support for coalition goals in one or more community sectors, such as local government and schools (Zakocs, et al., 2008). Influential champions can work within a network or organizational bureaucracy to provide coalitions with visibility and aid in the commitment of resources (Steckler & Goodman, 1989). Champions can also help the coalition frame the issue in ways that will resonate with key stakeholders, thereby enhancing readiness to adopt targeted programs (Cookston, Sandler, Braver, & Genalo, 2007).
The U.S. and Mexican coalitions in this study focused on the prevention of youth substance use and violence. Both were multi-sector partnerships, engaging parents, teachers, youth, schools, media, law enforcement, faith community, health providers, social service agencies, and government agencies. Coalition coordinators helped organize day-to-day operations, typically as part-time paid staff. Substantial training and technical assistance were available in both countries, helping to support the use of best practices and troubleshooting implementation challenges. A key difference was that the U.S. coalitions followed the Communities That Care (CTC) system whereas coalitions in Mexico followed the Strategic Prevention Framework (SPF; Hawkins, Catalano, & Arthur, 2002; Orwin, Edwards, Buchanan, Flewelling, & Landy, 2012).
Both CTC and SPF use epidemiological data to identify prevention priorities. However, CTC employs a specific survey to assess risk and protective factors for problem behaviors, whereas the SPF is less proscriptive in its approach to community assessment (Hawkins, et al., 2004). The CTC system is again more proscriptive in selecting strategies to address priorities, emphasizing the use of evidenced-based programs that meet rigorous inclusion criteria (e.g. programs listed at www.blueprintsprogams.com). The Mexican coalitions received training from Community Anti-Drug Coalitions of America (CADCA) emphasizing the use of environmental change strategies, for which the evidence-base is more difficult to establish (Pettibone, Friend, Nargiso, & Florin, 2013).
The U.S. coalitions were located throughout the state of Pennsylvania and funded by the Pennsylvania Commission for Crime and Delinquency. The Mexican coalitions emerged through a partnership between Programa Compañeros Asociación Civil and the Alliance of Border Collaboratives, with financial support from the United States Department of Anti-Narcotic Affairs at the US Embassy in Mexico. The partnership – now called the Red de Coaliciones Comunitarias (Network of Community Coalitions)--started nine coalitions in four border municipalities: Ciudad Juárez (pop. 1,506,908), Nogales (pop. 220,292), Agua Prieta (pop. 79,138), and Tijuana (pop. 1,300,983). Ciudad Juarez and Tijuana were larger urban areas, whereas Nogales and Agua Prieta were smaller towns. Each of the Mexican coalitions focused on one specific neighborhood within these cities, with target population size ranging from 3,295 to 220,000 residents. The U.S. coalitions focused on school districts or counties, targeting rural and suburban areas, along with small and large towns. The population size of the areas where U.S. coalitions operated ranged from 4,199 to 438,965 people.
The goals of the current study are twofold. First, we examine the psychometric properties of measures for the previously described coalition context and capacity constructs in English and Spanish. These measures can be helpful in future research aimed at understanding how coalition context and capacity may influence coalition implementation of programs and policies, along with changes in community level health outcomes. The second goal of the study is to compare the context and capacity of U.S. and Mexico coalitions to see if there are systematic differences across settings. We do not hypothesize specific differences between the U.S. and Mexico, but the identification of systematic differences can help to inform how coalition approaches to health promotion may need to be different in the U.S. as compared to Mexico. We accomplish these goals by analyzing coalition context and capacity survey data collected from the previously described coalitions in Mexico and the U.S.
The coalition context and capacity assessment was originally developed in English, drawing from several existing measures (Arthur, Hawkins, Catalano, & Olson, 1998; Bumbarger, Kelsey, Mastrofski, & Witmer, 2010; Chilenski, et al., 2007; Feinberg, Chilenski, et al., 2007; Feinberg, et al., 2004; Greenberg, et al., 2007; Substance Abuse and Mental Health Services Administation, 2009) The survey was translated in Spanish, using back translation to identify discrepancies followed by discussion to reach consensus on the most appropriate translation (Cantor et al., 2005). Study instruments and data are available from the first author.
In both the United States and Mexico, the coalition context and capacity assessment was administered soon after coalitions received funding. In the United States, the assessment was administered as a web-based survey. Coalition leaders provided researchers with the email addresses of coalition members, who were then invited to participate in the survey. Of the 313 individuals invited to complete the survey, there were 139 respondents across the 7 community coalitions, for a response rate of 44%. In Mexico, the survey was administered via paper and pencil at the initial coalition training. Everyone who attended the meeting completed the survey, for a response rate of 100%. However, membership lists for the Mexican coalitions were not developed until after the initial training, thus making the response rates across countries difficult to compare. In Mexico, there were 221 respondents from 9 community coalitions in Juarez, Nogales, and Tijuana. Among the completed surveys in the U.S. and Mexico, the percentage of missing data across items ranged from 3% to 21%.
Table 1 provides coalition and respondent characteristics from Mexico and the United States. Mexican coalition members were significantly younger, with less formal education, and were more likely to be male. Further, Mexican respondents had less experience in prevention and were less likely to have prior coalition involvement. With regard to occupational sector representation, the Mexican coalitions had more participants from the business and concerned citizen categories, whereas the U.S. coalitions had more human services representation. The Mexican coalitions operated with substantially less grant funding and tended to be in areas with higher population density.
Table 2 presents psychometric properties of the measures examined in this study, including the number of items, intraclass correlation coefficient (respondents are nested within coalitions), mean, standard deviation, Cronbach's alpha value, and inter-measure correlations separately for the U.S. and Mexico. Following is an example item from each scale along with a description of the response options.
E.g. “For each one, indicate how serious of a problem you feel it is in your area...Poverty.” Response options ranged from (1) Not at all serious to (7) Very serious (adapted from Arthur, et al., 1998).
E.g. “Community leaders are...Able to represent all sectors of the community, including disadvantaged groups.” Response options ranged from (1) Strongly disagree to (7) Strongly agree (adapted from Feinberg, et al., 2004).
E.g. “Most people in your area feel a strong tie to the community.” Response options ranged from (1) Strongly disagree to (7) Strong agree (adapted from Chilenski, et al., 2007).
E.g. “Does the community overall recognize that there is a substance abuse problem that needs to be addressed?” Response options varied across items but ranged from (1) The community is in denial to (4) The community is extremely aware of substance use problems, for the example item (adapted from Feinberg, et al., 2004).
E.g. “In what manner does your coalition collaborate with each of the following organizations...Primary and secondary schools.” Response options included: co-sponsor of events or activities, share funding or in-kind resources, exchange information, no collaboration, and does not exist in our community (adapted from Chilenski, Ang, Greenberg, Feinberg, & Spoth, 2014). Scores for each item ranged from 0 to 3, computed as a sum, with each of the first three response options counted as one point if marked.
E.g. “Is there an enthusiastic champion for substance abuse prevention—someone who is influential and effective in promoting substance abuse prevention in their own organization and in the community—in the...Local government.” Response options ranged from (1) No champion to (7) A very effective champion (adapted from Bumbarger, et al., 2010).
We used Mplus version 7.2 to conduct a Confirmatory Factor Analysis (CFA) of our context and capacity measurement model. For the regression analyses predicting each coalition context and capacity measure at the individual level, we used SAS version 9.3. Proc Mixed estimated multilevel regression models that accounted for the nesting of survey respondents within coalitions. To account for missing data, we used multiple imputation, generating 10 imputed datasets (Sinharay, Stern, & Russell, 2001). All measures of coalition context and capacity were standardized prior to analysis to ease interpretation (mean = 0; standard deviation = 1). Coalition nationality was the focal predictor, with gender, age, educational attainment, years working in prevention, previous coalition involvement, and population density entered as covariates. We conducted follow-up analyses on meaningful scale sub-components if coalition nationality was a significant predictor of the overall scale at p < .05. Only Community Problems, Collaborative Relationships, and Community Champions contained meaningful sub-components.
Table 2 reports the psychometric properties of the coalition context and capacity scales. Cronbach's alpha values generally range from .68 to .96. However, Community Support for Prevention had substantially lower alpha values, at .45 for Mexico and .52 for the United States. Alphas for each construct were similar in the U.S. and Mexico. Correlations between the measures of coalition context and capacity are generally similar in the Mexican and U.S. samples, with one exception. The correlation between Community Problems and Community Champions is positive in Mexico (r = .18, p < .05) and negative in the U.S. (r = −.05, n.s.).
Figure 2 presents the standardized parameter estimates and fit indices for the CFA of the context and capacity measurement model. The initial CFA had χ2 of 3463.65 (df = 1469, p < .05), a Comparative Fit Index (CFI) of .80, and an RMSEA of .06. Several item residual variances were significantly correlated within three factors: Community Problems, Community Champions, and Collaborative Relationships. Freely estimating the residual variances within these constructs that had a magnitude greater than r = .19 improved model fit substantially. The revised CFA had χ2 of 2453.37 (df = 1437, p < .05), a Comparative Fit Index (CFI) of .90, and an RMSEA of .04. Standardized factor loading magnitudes varied across constructs. For Community Champions, Collaborative Relationships, Participatory Leadership, and Sense of Community, factor loadings ranged from .59 to .85. Community Problems factor loadings ranged from .31 to .78 and Community Support for Prevention factor loadings ranged from .22 to .72.
Table 3 presents findings from the regression models predicting coalition context and capacity. Coalition nationality was a significant predictor of Community Problems, Sense of Community, and Community Support for Prevention. Specifically, Mexican coalition nationality predicted a 1.00 standard deviation unit increase in Community Problems. Mexican coalitions were .54 standard deviation units lower on sense of community. With regard to Community Support for Prevention, Mexican coalitions were .57 standard deviation units lower. The covariates of gender, age, educational attainment, years working in prevention, and previous coalition involvement were generally not significant predictors of coalition context and capacity. The only exception was for gender; being female predicted a .31 standard deviation increase in Community Problems.
Coalition nationality predicted significant differences in one scale containing meaningful sub-components - Community Problems. Figure 3 presents follow-up regression analyses using the same set of covariates to predict the specific issues measured by each item in the Community Problems scale. The magnitude of the estimate for the coalition nationality covariate ranged from −.07 for mental illness to 1.28 for lack of enforcement for laws. The regression estimates presented are Cohen's d effect sizes, where ±.8 is considered a “large” effect, ±.5 is a “medium” effect, and ±.25 is a “small” effect (Cohen, 1988). The perceived severity of problems with youth violence, lack of enforcement for laws, lack of trust in law enforcement, school quality, easy access to illegal drugs, and crime were all predicted to be more than .8 standard deviation units higher among respondents from Mexican coalitions.
This study makes a unique contribution to the literature by providing data on the psychometric properties of coalition context and capacity measures in both Spanish and English. The psychometric properties of the scales appear to be strong. The internal consistency of each scale is similar across U.S. and Mexican coalitions, supporting the validity of comparisons between countries. Scales measuring community problems, participatory leadership, collaborative relationships, and community champions all had strong internal consistency (α = .87 - .96). Sense of community had moderate internal consistency (α = .68 - .72). CFA fit indices suggest community problems, collaborative relationships, and community champions are not unidimensional constructs however, as several residual variances within these constructs had to be estimated freely to produce good fit for the overall measurement model. Each item represents a different problem or sector of the community, thus it is not surprising some items within a construct are more correlated than others.
The only scale with poor internal consistency and low factor loadings in the CFA was community support for prevention (α = .45 - .52), which may be improved with additional items. Community support for prevention may also be additive in nature, with each item measuring unique issues that together yield important information. If the low alpha values indicated the measure was not valid, the large correlations observed with participatory leadership would be unlikely (r = .52 in the U.S. and r = .48 in Mexico). Thus, we believe the scale is valid and useful as constructed, although future research with additional items may further improve the scale.
In general, correlations between scales were modest but in the expected direction, suggesting scales measure unique factors and possess construct validity. Specifically, community champions, community support for prevention, sense of community, and participatory leadership were positively correlated with one another among both U.S. and Mexican coalitions. Each of these coalition strengths may help to support the others (Stone & Hughes, 2002). For example, community champions’ efforts may increase community support for prevention, which in turn may cultivate additional community champions (Dariotis, Bumbarger, Duncan, & Greenberg, 2008).
Collaborative relationships and community problems were generally not significantly correlated with other constructs. It may be that they influence coalition functioning largely independently of other aspects of context and capacity. However, community problems were negatively related to sense of community among U.S. coalitions. Low sense of community may exacerbate other problems in the community and vice versa (Chung & Lim, 2014). In Mexican coalitions, community problems were positively associated with the presence of community champions for prevention, which may help coalitions address problems in their community (Zakocs, et al., 2008).
This study is the first to our knowledge to measure coalition context and capacity in Mexico and the first comparison across international settings. Coalition context in Mexico is substantially different from that of the United States. Results of this study show that coalition members in Mexico perceive more severe community problems, especially a lack of enforcement for established laws, high rates of youth violence, and a lack of trust towards the police and government. Although familiarity with these border cities may make the findings obvious, it is interesting to note that problems such as poverty, which are also higher in these Mexican communities than in the U.S., are not perceived as significantly more severe by coalition members. If Mexican coalitions want to address problems with a lack of effective and trustworthy law enforcement, the strategies utilized will need to be vastly different from those taken by drug prevention coalitions in the U.S., which often focus on implementing school and community-based programs (Hawkins, et al., 2002; Spoth, et al., 2013).
The Mexican coalitions also face lower levels of support for community prevention activities and fewer community champions, making it challenging to build the momentum necessary to take action (Aoun, Shahid, Le, & Packer, 2013; Holmila, Mustonen, Sterberg, & Raitasalo, 2009). Further, the Mexican coalition members perceive lower levels of sense of community, which may make the establishment of trusting community relations particularly challenging but nevertheless essential. The ability of the Mexican coalitions to manage these challenges will likely play a large role in determining whether they are successful in their substance use prevention efforts. Mexican coalitions may need to focus additional energy on building community support for their implementation efforts to be successful.
A key strength of this study is its examination of coalition context and capacity in substantially different international settings using the same measures. One challenge in interpreting the results is that the geographic scope of the U.S. sites included counties and school districts, whereas the Mexican sites were always neighborhoods, typically in more urban areas. It is difficult to know how these differences impacted results, although the nationality effects are not due to population density, which was included as a covariate in analyses. An important limitation of the study is that the data are cross-sectional, limiting our ability to establish the validity of the scales in predicting outcomes such as the implementation of prevention programs and policies. One limitation of the measures is that they rely on coalition member perceptions of community context and capacity. Observational ratings or survey sampling of community members outside the coalition may provide a more accurate approach to examining community context. The measures presented here regarding community context are not intended to replace epidemiological assessments of community risk and protective factors (e.g., Feinberg, Ridenour, et al., 2007).
The assessment tools presented here can help to inform future coalition action aimed at strengthening coalition capacity and shaping community context to be more supportive of prevention. Future research should identify best practices coalitions can use to build capacity and address prioritized contextual challenges. This assessment is a first step in clarifying the context and capacity constructs that influence the achievement of coalition goals. With additional understanding of these important challenges, coalitions will be better able to identify and address weaknesses, thereby enhancing long-term goal achievement.
This research is supported by a grant from the U.S. Embassy in Mexico and the Pennsylvania Commission for Crime and Delinquency (PCCD). Additionally, this article is supported in part by the National Cancer Institute through a Community Networks Program Center grant U54 CA153505. Findings and recommendations herein are not official statements of the U.S. Embassy, PCCD, or NCI. We are grateful for the cooperation the many coalition members in Mexico and the United States.