Methods One urban county, with 44 unique municipalities, wanted to begin a long term prevention campaign for cities with highest level of need and most resources. We collected municipal level data (administrative and primary) from several sources (e.g. census, police, liquor control commission) and selected 7 variables pertinent to intervention planning (e.g. crashes, fatalities, treatment admissions, juvenile violations, and alcohol outlet violations). The variables for all 44 municipalities were standardized into z-scores and submitted to a k-means cluster (Hartigan, 1975) using SPSS. Cluster analysis (Wilkinson et al., 1996) was used to identify similarities and differences between the 44 cities.
Results A 9 cluster solution was decided upon as an effective planning tool because city groupings were distinct and meaningful to community stakeholders. Cities were chosen with attention to within and between cluster groupings, allowing for within cluster comparisons as the interventions were implemented. In Year 1, three target cities, with matched comparison cities were selected from 3 distinct clusters, for intervention and evaluation. Once selected, additional data on youth substance use, related norms, community alcohol consequences, consumption patterns and resources important to underage drinking and drunk driving was collected for each of these matched pairs through interviews, surveys and administrative records. Data from the cluster analysis and descriptive data specific to the community were used in an intervention planning, aimed at residents with tailored messages about norms related to substance use.
Conclusions and Implications Using the cluster analysis technique improves practitioner's ability to identify communities for interventions when the size and characteristics of community is complex. It also yields data for within cluster comparisons useful for evaluating community interventions. Collecting data from school, state and county sources was challenging but produced community profiles of underage drinking and drunk driving, heretofore undocumented at the city level. We found that the validity of secondary data was critical and dependent upon use of systematic codes, law enforcement practices and local laws and ordinances. Youth data had limitations, but communities valued the data to tailor strategies and tactics that fit their community.