Using Social Indicators, Systems Science and Predictive Modeling to Improve Prevention Planning

Application of the risk and protective factor framework to prevention planning relies on information regarding the levels of risk and protection in the areas or populations to be served. Social indicators provide a significant source of data that can be used for this purpose. Social indicator studies are particularly valuable because they bypass the high cost and time commitments, as well as many of the methodological weaknesses and impracticalities, associated with primary data collection. However, these indicators are often examined individually despite known associations and an understanding that these factors are multidimensional in relation to human behavior.

State Epidemiological Outcomes Workgroups (SEOW) were funded to encourage recipients to use empirical data to document needs, justify planning decisions, guide resource allocation, and monitor performance.  Toward that end, the state of North Carolina used a social indicator approach to organize indicators of risk and protective factors; substance use and mental health; and related consequences to allow the state’s 100 counties to prioritize their needs.  However, the social indicators study (SIS) did not examine the relationship between indicators, nor determine which indicators were most associated with each other.    

To better understand the dynamic relationship between risk indicators and related consequences, a system science model development approach was applied. A conceptual model based on current practitioners’ beliefs and experiences was developed. This model included a causal diagram with nodes representing critical variables and arrows representing causal relationships. At the same time, a data-driven modeling approach was built to reflect what independent data is showing unbiased by individual opinion. At the next stage these two models: conceptual ad data-driven were compared to each other and through brainstorming sessions a systems model that has elements of individual experiences and hard data was developed. The model was programmed and calibrated based on existing data and policy experience, and then used to consider and evaluate policy scenarios.

The presentation shows how North Carolina combined a systems science approach and predictive modeling to demonstrate the utility of social indicator studies for empirically identifying risk factors to substance abuse, mental health, and related consequences.  Additionally, the presentation will demonstrate and discuss the utility of heat maps to visually display clustering among counties as a function of risk. The comparative analysis will be conducted to examine the predictive utility of state- and county-level social indicators to identify the variables most associated with outcomes of interest.