Assessing Community Readiness for Preventing Youth Substance Use in Colombia: A Latent Profile Analysis

Abstract

The growing use of evidence-based preventive interventions for youth substance use in Latin American countries has prompted governments, researchers, and practitioners to ask if communities are ready for implementing these interventions, especially in light of the elevated costs and long-term commitment necessary for successful implementation. This study explores the construct validity of a measure of community readiness for prevention, using confirmatory factor and latent profile analyses of 7 measures theorized to be indicators of community readiness for implementing preventive interventions for youth substance use. Data were obtained from 211 community leaders in 16 communities in Colombia. Results indicate that community readiness can be represented as a unidimensional construct with multiple profiles of varying levels of readiness. Findings suggest community readiness can be measured adequately as a latent construct and that its indicators can be used diagnostically to assess areas where readiness could be improved for better implementation of evidence-based preventive interventions.

Introduction

Prevention science in Latin America is currently undergoing a dramatic expansion, with increasing development and implementation of evidence-based preventive interventions (Pérez-Gómez, Mejía-Trujillo, & Becoña-Iglesias, 2015). The approach to addressing youth substance use in countries such as Argentina, Brazil, Chile, Costa Rica, Mexico, and Peru is undergoing a progressive transformation, from criminalization towards the adoption of a public health framework (Pérez-Gómez et al., 2015). Even though different Latin American nations are currently at different stages of prevention science research and application, there seems to be a continental trend toward increasing the development, adaptation, testing, and implementation of evidence-based substance use prevention interventions (Pérez-Gómez et al., 2015).

As prevention (and its component health promotion) becomes the preferred approach to reducing the incidence and prevalence of substance use and related health and behavioral outcomes among youth (National Research Council and Institute of Medicine, 2009), the use of prevention systems to implement programs, strategies, and policies in the community also becomes increasingly important (Brown, 2015). Prevention systems consist of a set of prevention-specific components that are interdependent and function together, achieving more preferable results than they would separately (Hirsch, Levine, & Miller, 2007). They are implemented with the purpose of transforming structural elements that support or hinder preventive efforts focused on an issue of interest to a broader context, such as a community or an organization (Brown, Hawkins, Arthur, Briney, & Fagan, 2011). The main characteristics of a prevention system are: (a) it is guided by a clear theory of change; (b) it targets a broad set of outcomes that are relevant to the community or organization; (c) it depends on cross-domain and inter-sector collaboration to deliver preventive services efficiently; and (d) it helps in identifying mediators, or proximal outcomes, that can link the most appropriate intervention with targeted outcomes (Behrens & Foster-Fishman, 2007; Brown, 2015; Stevenson & Mitchell, 2003).

One example of a prevention system is Communities That Care (CTC; Hawkins & Catalano, 2002; Quinby et al., 2008). CTC is a manualized system designed to assist community coalitions in the development and implementation of a strategic prevention plan to protect the healthy development of local youth (Hawkins & Catalano, 2002; Quinby et al, 2008). CTC has been translated and adapted for implementation in a number of countries outside of the United States, such as Australia, Germany, Sweden, Chile, and Colombia. For implementation in Colombia, the system underwent a few necessary adaptations to ensure better fit with the local context, resulting in the development of Comunidades Que Se Cuidan (CQC; Pérez-Gómez, Mejía-Trujillo, Brown, & Eisenberg, 2016). Such adaptations ensured that the system was sensitive to the specific needs and interests of local authorities, researchers, and community leaders, and that it was adequate to assist them in achieving community-wide impacts.

Nonetheless, the implementation of prevention systems and programs in lower and middle-income countries (LMICs) is still very challenging. A persistent challenge to evidence-based prevention implementation in Latin America is the translation of academic research into operationalizable strategies that can be used by local governments to minimize rates of youth substance use (Pérez-Gómez et al., 2015). Additional barriers to evidence-based decision making and practice in these LMICs persist in the form of (a) reduced governmental and financial support, (b) social and cultural structures that hinder implementation, (c) incompatibilities between the design of chosen interventions and contextual needs of the receiving community, and (d) the profile and skill set of professionals in charge of implementing and delivering preventive interventions (Pérez-Gómez et al., 2015). Confronted with these challenges and the simultaneous growth in interest and use of evidence-based preventive interventions in Latin America, local authorities and researchers have been prompted to ask if their target communities are indeed ready to implement these interventions, especially in light of the elevated costs and long-term commitment necessary for successful implementation.

Community readiness has been succinctly defined as “the extent to which a community is adequately prepared to implement a prevention program” (Stith et al., 2006, p. 601). Prevention researchers have suggested that community readiness is a paramount prerequisite for successful preventive intervention implementation (Edwards, Jumper-Thurman, Plested, Oetting, & Swanson, 2000; Feinberg, Greenberg, & Osgood, 2004; Kostadinov, Daniel, Stanley, Gancia, & Cargo, 2015; Oetting et al., 1995; Stith et al., 2006;). For example, a study on collaborative partnerships found that attributes such as having skilled and effective leadership, sharing power and decision making, members sharing a common vision for the future, and the active participation of a variety of stakeholders were all important for community coalitions’ effectiveness in reaching their goals (Allen, 2005). Socio-ecological factors such as the acceptance of a prevention effort within a community, or the compatibility between community norms and values and a prevention framework, can impact intervention effectiveness and implementation quality (Oetting et al., 1995). Research suggests that preventive intervention implementation is unlikely to happen effectively in communities displaying low levels of readiness, and that the insistence on implementing in those communities will likely lead to failure (Edwards et al., 2000). Researchers also stress that community readiness is problem-specific, highlighting the need for assessing readiness levels separately for each issue targeted by the community (Edwards et al., 2000; Oetting et al., 1995). Given the importance of community readiness for successful preventive intervention implementation, it comes as no surprise that researchers would attempt to quantify readiness. The assessment of community readiness, however, is a relatively new field, and there currently is a disagreement on how to measure levels of readiness (Castañeda et al., 2012; Foster-Fishman, Cantillon, Pierce, & Van Egeren, 2007). The diversity of available measures to assess community readiness is also representative of the numerous theoretical frameworks conceived for understanding this phenomenon.

One framework which seems to encompass other accepted models identifies four core elements of community readiness that are key in the successful implementation of preventive interventions: (a) enough community capacity for implementation, (b) recognition of the magnitude and urgency of the problem by local leaders and residents, (c) identification and support of prevention efforts by a local champion, and (d) an appropriate climate for the implementation of preventive interventions, including a community’s openness to change and support for prevention efforts (Stith et al., 2006). Community capacity refers to a sense of shared responsibility for the health and well-being of the collective, combined with the competence to ensure said well-being and address any issues that may jeopardize it (Stith et al., 2006). Recognition of the problem consists of leaders and residents being aware of the issue at hand and how it affects the community, having a sense of urgency to address it, and recognizing that new interventions or efforts are necessary to do so. The local champion is a key leader who is recognized and respected by the community, who is committed to the program or effort being implemented, and who fosters support for prevention from their peers. The development of a favorable climate for implementation consists of fostering an environment that is conducive to collaborative work, where community partners are aware and knowledgeable about the preventive interventions to be implemented, and receive incentives for the continued use of those interventions (Stith et al., 2006). Moreover, some authors have suggested that community readiness for preventive intervention implementation is associated with a community’s capacity to develop and maintain a prevention coalition (Feinberg et al., 2004).

The need to assess readiness for communities to embark on a system-level transformation from traditional “prevention as usual” to “science-based prevention” is especially important in LMICs that may not have a tradition of science-led and government-supported research and service delivery. Moreover, as this is a relatively under-researched area, the idea of preparing communities (and community coalitions as a vehicle) for local decision making and maintenance of prevention initiatives may come across as a very foreign concept.

Purpose of This Study

In light of this, the purpose of the current study was to (a) examine the construct validity of a practical measure of community readiness from the perspective of those most knowledgeable about their community and local preventions efforts (i.e., key informants, such as leaders and staff from the mayor’s office, health department, local schools and businesses) and (b) identify if responses from key informants can be classified into profiles based on levels or dimensions of readiness (i.e., different constellations of readiness indicators, which would indicate that readiness is not a uni-dimensional construct). This study builds on the recent prevention system development in Colombia with the introduction and scale-up of CQC throughout the country (Pérez-Gómez et al., 2016). Establishing construct validity will contribute to determining how well the referred measure represents the community readiness construct, and thus how useful it would be for community leaders and implementers alike. Identification of readiness profiles can help inform the classification of communities with regard to their need for preparation prior to preventive intervention implementation, with the objective of enacting system change. Definition of such a classification system would be particularly important for prevention science in Latin America, constituting an important tool to support the expanding development and implementation of evidence-based preventive interventions.

Method

Participants

Data for these analyses were compiled from three waves of data collected from Colombian key informants in 2013, 2016, and 2018. Participants were 211 key informants from 16 different communities. The first wave of assessment included 50 leaders from five communities in Quindío, Colombia (i.e., Calarcá, Circasia, La Tebaida, Montenegro, and Quimbaya); the second wave included 64 leaders from five communities in Cundinamarca, Colombia (i.e., Sopó, El Rosal, Pacho, Chocontá, and Villeta); and the third wave included 97 leaders from six communities in the departments of Cundinamarca and Boyacá (i.e., Guayatá, Tabio, La Mesa, Mosquera, Fusagasuga, and Puerto Salgar). Respondents were key informants representing a variety of community sectors (e.g., civic, educational, and law enforcement). All participants held positions of leadership in the community (e.g., mayors, police chiefs, and principals) or had specialized knowledge of prevention-related efforts in their respective communities. All data were collected prior to the implementation of Comunidades Que se Cuidan in the community.

Measures

Seven measures of community readiness were adapted from the CTC Coalition Board Interview (Feinberg, Bontempo, & Greenberg, 2008) and the CTC Community Key Informant Survey (Arthur, Glaser, & Hawkins, 2005), instruments that are used traditionally to collect prevention coalition and system readiness data and inform CTC implementation in the United States (Hawkins et al., 2008). All measures were translated into Spanish by a team of prevention researchers from Colombia and the United States, who also reviewed each measure’s cultural appropriateness for use in Colombia.

Norms Against Youth Substance Use was measured by eight items (coefficient alpha = .80) that represented perceptions of normative beliefs about adolescent substance use in the community. This construct was operationalized by two sets of three items: (1) In this community, how wrong do most adults think it is for youth to (a) drink alcohol, (b) smoke cigarettes, (c) use marijuana?, (d) use other illegal drugs such as cocaine, ecstasy or heroin; and (2) Adults in your community think that using (a) alcohol, (b) tobacco, (c) marijuana, (d) other illegal drugs (for example, cocaine, ecstasy or heroin) is a normal part of youth growing up. Responses to the first set of questions were coded 1 = Not wrong at all, 2 = A little wrong, 3 = Wrong, and 4 = Very wrong. Responses to the second set of statements were 1 = Strongly Disagree, 2 = Somewhat Disagree, 3 = Somewhat Agree, and 4 = Strongly Agree.

Openness to Change was measured by three items (coefficient alpha = .62). Participants were asked to rate, in general, how open their communities were to change. Responses to this question were coded 1 = Open to change, 2 = A little open to change, 3 = A little resistant to change, and 4 = Resistant to change. The other two items measured the degree of willingness to try new ideas in the community and flexibility to change. Responses to those questions were 1 = Strongly Disagree, 2 = Somewhat Disagree, 3 = Somewhat Agree, and 4 = Strongly Agree.

Community Support for Prevention was operationalized using four items (coefficient alpha = .63) that measured community leaders’ assessments of community beliefs in prevention effectiveness, knowledge of prevention efforts, and willingness to pay for prevention programs.

Cohesion was measured by seven items (coefficient alpha = .80) that addressed unity, helpfulness, common interests and values, ability for people to get along, and feelings of being at home in the community.

Conflict Resolution was measured by four items (coefficient alpha = .60) that measured disagreements, conflict, and lack of ability to solve community problems in the community.

Effective Leadership was measured by eight items (coefficient alpha = .86) that measured the ability of leaders in the community to represent everybody in the community, gain consensus for decisions, obtain needed resources, share the decision-making process, and manage “turf battles” in the community, as well as how knowledgeable and committed they were to local prevention efforts focusing on youth substance use.

Shared Responsibility consisted of five items (coefficient alpha = .65) that measured community members’ attribution of responsibility and commitment to preventing youth substance use, and how willing they were to participate in community activities and decision making. Response options to items in Community Support for Prevention, Cohesion, Conflict Resolution, Effective Leadership, and Shared Responsibility were coded 1 = Strongly Disagree, 2 = Somewhat Disagree, 3 = Somewhat Agree, and 4 = Strongly Agree.

Statistical Analyses

We first conducted a Confirmatory Factor Analysis (CFA; Brown, 2006) of the seven readiness measures with correlations between the latent factors to test how well the scales represent the constructs of interest. Evidence of construct validity for community readiness would be realized if item-factor loadings for the indicator variables were large (i.e., > .50, Cohen, 1988) and statistically significant (ps < .05), and if the overall factor model fit the data well. Upon finding adequate fit of the measurement models to the data for the seven separate constructs, we went on to test a higher order molar readiness construct. Goodness of fit was established by assessing a preponderance of fit indices: Comparative Fit Index (CFI; Bentler, 1990) greater than .95, Tucker-Lewis Index (TLI; Tucker & Lewis, 1973) greater than .95, and Root Mean Square Error of Approximation (RMSEA; Browne & Cudeck, 1993; Steiger & Lind, 1980) less than .06. Adjustments to the factor structure were guided by theoretical considerations and inspection of modification indices. As scale items were categorical, CFA modeling relied on weighted least squares means and variances (WLSMV) estimation (Muthén & Muthén, 1998–2012), under the assumption of data missing at random (MAR; Little & Rubin, 1987).

As a second step, we conducted a latent profile analysis (LPA; Berlin, Williams, & Parra, 2014) of the seven measures to identify if participants could be classified into different levels or profiles of readiness. Participant scores were computed for each of the seven readiness measures as factor scores derived from the CFAs in Step 1, and modeled in the LPA to identify unobserved heterogeneity in patterns of community readiness across the seven measures. One-class to five-class models were examined. Entropy, Bayesian Information Criteria (BIC), Sample Size Adjusted BIC (SSABIC), the Sample Size Adjusted Lo-Mendell-Rubin likelihood ratio test (LMR-LRT; Lo, Mendell, & Rubin, 2001), and the Bootstrap likelihood ratio test (BLRT; Nylund, Asparouhov, & Muthén, 2007) criteria were used to select the best fitting model across the varying number of classes. Entropy values of at least .80, lower BIC and SSABIC values, and statistically significant (p < .05) LMR-LRT and BLRT results were all indicative of better model fit to the data (Nylund et al., 2007; Ram & Grimm, 2009). Plots of the factor mean estimates for each profile solution were also assessed for interpretability. The average missing rate across all study variables was 6%, and full information maximum likelihood (FIML; Enders & Bandalos, 2001) was used to address missing data in the LPAs. All analyses were conducted using Mplus v.8 (Muthén & Muthén, 1998–2012).

Results

Confirmatory Factor Analyses

CFA results initially suggested poor model fit to the data, χ2 (681, N = 144) = 1,374.35, p <. 001, RMSEA = 0.08, CFI = 0.89, TLI = 0.88. Table I shows standardized factor loadings, model fit indices, and coefficient alphas for each of the seven measures. Inspection of the correlation matrix indicated that the Norms Against Youth Substance Use measure appeared to be multidimensional; that is, it could be split into two correlated factors, each consisting of one of the two sets of items used to measure the original construct (i.e., In this community, how wrong do most adults think it is for youth to use [drug] versus Adults in your community think that using [drug] is a normal part of youth growing up). A subsequent CFA, with the two sets of items indicating two correlated factors, demonstrated substantially improved model fit, χ2 (674, N = 211) = 1,172.41, p < .001, RMSEA = 0.06, CFI = 0.94, TLI = 0.93. Coefficient alphas for two respective factors (labeled as “Wrong to Use” and “Normal to Use,” respectively) were .89 and .84. Inter-factor correlations are shown in Table II. Correlations were mostly large and all significant (rs ≥ 0.50, ps < .05) among Community Support for Prevention, Openness to Change, Cohesion, Conflict Resolution, and Shared Responsibility. Inter-factor correlations among “Normal to Use” and other factors were small (rs < 0.30), and those among “Wrong to Use” and other factors were small, except for Community Support for Prevention (r = 0.40) and Shared Responsibility (r = 0.39).

Table I.

Summary Statistics for the Seven Community Readiness Scales

	Standardized factor loadings			Model fit				Coefficient Alpha
Measure	Min	Max	Avg	CFI	TLI	RMSEA	χ2
Norms Against Youth
Substance Use	0.802	0.958	0.877	0.926	0.897	0.389	659.63***	0.80
Normal to Use	0.919	0.982	0.945	0.993	0.978	0.392	66.288	0.89
Wrong to Use	0.875	0.944	0.917	0.982	0.945	0.350	53.339	0.84
Community Support for Prevention	0.513	0.720	0.612	0.963	0.890	0.124	8.51*	0.63
Openness to Change	0.503	0.753	0.675	0.987	0.981	0.06	24.73*	0.62
Cohesion	-	-	-	-	-	-	-	0.80
Conflict Resolution	0.423	0.834	0.577	0.829	0.487	0.284	34.80***	0.60
Effective Leadership	0.646	0.850	0.734	0.953	0.934	0.135	94.39***	0.86
Shared Responsibility	0.197	0.914	0.596	0.996	0.991	0.058	8.56	0.65

Note.

^*p < .05,

^**p < .01,

^***p < .001.

Table II.

Inter-factor Correlations for the Seven Community Readiness Scales

	1.	2.	3.	4.	5.	6.	7.	8.
1. Wrong to Use
2. Normal to Use	0.214*
3. Community Support for Prevention	0.397*	0.025
4. Openness to Change	0.225*	−0.109	0.646*
5. Cohesion	0.245*	0.052	0.545*	0.897*
6. Conflict Resolution	−0.019	0.149	0.555*	0.549*	0.529*
7. Effective Leadership	0.271*	0.070	0.559*	0.366*	0.431*	0.369*
8. Shared Responsibility	0.386*	−0.007	0.713*	0.609*	0.564*	0.457*	0.418*

Note.

^*Correlation significant at p < .05 (two-tailed).

We then tested a higher order molar readiness factor, composed of eight factors—six original measures and the two new factors derived from the Norms Against Youth Substance Use measure. The higher order factor model had good fit, χ2 (694, N = 211) = 1,208.06, p < .001, RMSEA = 0.06, CFI = 0.94, TLI = 0.93. Inspection of standardized factor loadings suggested that the Normal to Use factor did not load significantly in the higher order readiness factor (λ = 0.08, p = 0.38). Nevertheless, we retained this measure for the subsequent LPA analyses to examine its, and the other measures’, ability to identify latent topologies of community readiness.

Latent Profile Analyses

Based on the results of the CFA analyses in Step 1, we proceeded to test latent profiles. LPA models were examined, allowing means for each readiness measure to be freely estimated for each class; however, measure variances were constrained to equality across classes. Fit statistics for the 1- through 5-class models are shown in Table III. Results indicated that the entropy statistic favored the 3-class solution, the Adjusted LMR-LRT favored the 4-class solution, and the BIC and SSABIC favored the 5-class solution. The BLRT did not distinguish among the solutions. Figures 1, 2, 3, and 4 present the standardized means for the 2- through 5-class models. Visual inspection of the profiles for the 3-, 4-, and 5-class models suggested that each progressively higher class model was merely splitting a class from the prior model into two new classes. Standardized means (z-scores) for the 2- and 3-class models demonstrated clear separation among latent profiles for most indicators with unidimensional ordinality ranging from “low” to “high” levels of readiness. The exception to this pattern was found for the “Normal to Use” measure, which did not differ significantly across classes in any of the examined models. All other readiness measures exhibited significant (p < .05) differences in scale means across classes in the 2- and 3-class models. The 4-class model also demonstrated ordinality between “low,” “medium,” and “high” levels of readiness; however, this model did not demonstrate meaningful separation among the two “low” readiness classes, except for Shared Responsibility (see Figure 3).

Table III.

Fit Statistics for Latent Profile Analysis

Fit statistics	1 Class	2 Classes	3 Classes	4 Classes	5 Classes
BIC	3443.500	3029.788	2861.619	2744.356	2731.355
SSABIC	3446.431	2925.224	2703.188	2532.058	2465.191
Entropy	-	0.897	0.951	0.921	0.906
Adj. LMR-LRT (p-value)	-	0.002	0.004	0.014	0.439
BLRT (p-value)	-	<0.001	<0.001	<0.001	<0.001
Group size (n, %)
Class 1	1.000 (211)	0.559 (118)	0.569 (120)	0.223 (47)	0.152 (32)
Class 2		0.441 (93)	0.313 (66)	0.318 (67)	0.299 (63)
Class 3			0.118 (25)	0.118 (25)	0.104 (22)
Class 4				0.341 (72)	0.209 (44)
Class 5					0.237 (50)

Note. BIC = Bayesian Information Criteria. SSABIC = Sample Size Adjusted BIC. Adj. LMR-LRT = Sample Size Adjusted Lo-Mendell-Rubin likelihood ratio test. BLRT = Bootstrap likelihood ratio test. n = class sample size.

Figure 1.

Standardized factor mean estimates for a 2-class model.

Figure 2.

Standardized factor mean estimates for a 3-class model.

Figure 3.

Standardized factor mean estimates for a 4-class model.

Figure 4.

Standardized factor mean estimates for a 5-class model.

Discussion

As adolescent substance use, delinquency, violence, and other associated health and behavior problems continue to grow as significant social problems in Latin America, the need to address these problems with effective prevention programs and strategies has also been growing. However, systems to implement preventive interventions in families, schools, and communities have not kept pace with this need. One exception is the Comunidades Que se Cuidan initiative in Colombia, which implemented the community coalition-driven system in 13 communities over a 6-year period. This study used data from the implementation of CQC in Colombia to examine indicators of community readiness for prevention system change. Although survey data with community leaders have been used in Latin America to descriptively assess a community’s ability to successfully embark on system-level change, psychometric assessment of these quantitative indicators of community readiness had not been conducted prior to the current study.

Results of this study demonstrate the construct validity of community readiness, as a latent construct indicated by several measures that assess characteristics of communities which have been theorized to be essential elements of effective system change (Arthur et al., 2005). With one exception (Norms Against Youth Substance Use), all measures demonstrated acceptable model fit and significant loadings on a molar measure of readiness. Interestingly, the set of items that measured normative beliefs about adolescent substance use in the community (i.e., “Normal to Use” ) did not load well on the global readiness factor and did not discriminate among the latent profiles of community readiness. Items on the “Wrong to Use” scale, however, did load on the global readiness factor and did discriminate among the latent profiles (although the discrimination among profiles was somewhat less for this measure than for the other measures). This finding highlights the distinction between perceptions of “what is” and “what should be,” commonly known as the “is-ought” problem (also known as Hume’s law), where the two perspectives are often confounded yet may describe very different phenomena (Black, 1964). Whereas community informants seem to agree that using alcohol and drugs is a normal part of growing up, there is varying consensus as to how wrong these behaviors are.

Although community leaders’ perceptions of normative drug use among adolescents in the community provided little information regarding a communities’ readiness for system change, the other examined measures demonstrated a good deal of information on readiness. Community Support, Openness to Change, Cohesion, Shared Responsibility, Conflict Resolution, and Leadership provided clear separation among latent profiles in the 2- and 3-profile solutions, and Shared Responsibility contributed with unique class separation in the 4-profile solution. These findings imply that it is possible to “diagnose” community readiness prior to implementing service delivery systems, and to identify areas of readiness that may be in need of strengthening. The community diagnosis model (Feinberg, 2012), for example, describes how quantitative measures of risk and protective factors, derived from school-based surveys of youth, can be used to identify targets for specific preventive interventions. We extend this line of thinking to the organizational level, where consideration of how to implement services is as important as which services to implement, and where these component parts of community readiness can be considered as system-level risk and protective factors for effective implementation and sustainability of community-based prevention efforts. Quantitative assessment of these elements of community readiness in the early stages of prevention planning could provide useful information on how best to prepare communities for system change. Communities with low levels of support for prevention activities, little openness to change, or high rates of internal conflict, for example, could consider specific “readiness interventions” that address these elements. Preparing systems for implementing evidence-based preventive interventions is in its infancy. To date, addressing community readiness issues has only been done through the delivery of training modules that are part of broader prevention programs or systems. The CTC system, for example, provides local leaders with trainings focused on collaboration, increasing knowledge about prevention, agreeing on a shared vision for prevention efforts, and organizing the work. The adoption of a quantitative assessment of core readiness constructs such as those presented here will facilitate the development of interventions to impact community readiness, ensuring that they can be tailored to each community’s needs.

Just as “bad systems trump good programs” (McCarthy & Kerman, 2010, pg. 167), bad systems (or no systems) may lead to a mismatch between existing interventions and community needs; lack of long-term follow-up (i.e., maintenance and sustainability) to community-based initiatives; poor fidelity to intervention protocols; restricted scaling up of interventions to new areas; dwindling funding for prevention efforts; and weak, if any, effects on targeted health outcomes (Bickman, 2008; Seidman et al., 2010). Similarly, community readiness can act as a “deal-breaker” in the successful implementation of preventive interventions, since low levels of readiness have been associated with higher rates of implementation failure (Edwards et al., 2000). Factors such as a community’s acceptance of a prevention endeavor, or compatibility between their norms and values and a proposed prevention framework, can impact intervention effectiveness and implementation quality (Oetting et al, 1995). Promoting community readiness for system development and implementation, while not a panacea for all reasons why systems may go “bad,” is certainly a primary step toward making them “good.”

Limitations

In this study, we used data from key informants to measure community readiness. The analyses were conducted at the individual level, despite the fact that the unit of interest was the community. Ideally, we would have used a multilevel design, with informants nested within communities. However, this would have required a very large sample size, and unfortunately community leaders are a hard-to-reach population. For this reason, we opted for analyzing data at the individual level, knowing that the results may be subject to ecological fallacy – that is, wrongfully making assumptions about the communities based on individual-level results (Diez-Roux, 1998; Feinberg, 2012). Power analyses conducted using data from the current study indicated that there was sufficient statistical power to detect individual-level correlations among readiness factors as small as r = .21. While it is empirically possible for factor structures and latent profiles to be different at the community level than at the individual level, we believe that, in practice, individual-level data would typically be used to identify a given community’s readiness profile. Therefore, the results presented in this study have great utility, despite this design limitation. We leave it to future research to extend the external validity of our findings using larger multilevel designs.

The generalizability of findings from this study is limited; we cannot claim that the participating communities are representative of the country or region beyond the departments of Colombia assessed in this study. Furthermore, due to the cross-sectional nature of the data, we are not able to ascertain if communities with higher levels of Openness to Change or Cohesion, for example, were more successful in the implementation of prevention systems and programs than communities at lower levels of those constructs. As prevention efforts and research progress in Latin America, we expect that local authorities and foundations will invest in further data collection to build nationwide profiles of readiness for preventive intervention implementation, as well as evaluate the success of different profiles in attaining local system transformation.

Conclusion and Future Directions

The importance of community readiness for prevention system success prompts researchers and key leaders to assess readiness levels and address any readiness shortcomings prior to proceeding with the implementation of other preventive strategies. This study contributes by establishing construct validity for a quantitative measure of community readiness applied to an LMIC.

Future research should test this measure with other samples, both within Colombia and in other Latin American countries, to further investigate its validity for these populations and contexts. Research on local leaders’ Norms Against Substance Use should further explore the distinction between participants’ perception of youth substance use as being normal, and how much they perceive it as being wrong. This will contribute to a better understanding of the role that norms play in the global construct of community readiness. Moreover, conducting longitudinal research in Colombian and Latin American communities will contribute to understanding how well the identified readiness profiles predict the quality and success of prevention implementation.

This assessment tool, when used at early stages of prevention planning, will assist local communities in the identification of readiness levels, in preparing for system change, and in the implementation of evidence-based preventive interventions. For researchers, it will help inform the development and evaluation of interventions that can be tailored to best impact readiness according to each community’s specific needs. At the government level, identifying different readiness profiles will allow funding to prioritize getting all target communities to acceptable levels of preparedness, thereby contributing to better implementation and intervention outcomes. The utilization of such a measure within a context of expanding prevention efforts will prove crucial in sustaining the growth of Latin American prevention science.