Abstract: Cluster Analysis to Identify Groups for Health Promotion (2011 National Conference on Health Communication, Marketing, and Media)

Frederico M. Cohrs, MSc, Candidate¹, Fernando Sousa, MSc, Candidate², Luiz Ramos, MD, PhD³ and Ivan Pisa, PhD², 1Departament of Preventive Medicine, Universidade Federal de São Paulo, Sao Paulo, Brazil, 2Departament of Health Informatics, Universidade Federal de Sao Paulo, Sao Paulo, Brazil, 3Departament of Preventive Medicine, Universidade Federal de Sao Paulo, Sao Paulo, Brazil

Theoretical Background and research questions/hypothesis: A clear communication to a target market can lead to better results. Thus, it is important to identify that target market. Most of the time, policy makers use segmentation study to separate groups and then apply a model of promotion. Data from the first interview from the Project Epidoso, a cohort study to understand the process of aging in São Paulo, was used as a scenario to apply a cluster analysis to identify an optimum number of clusters of people with similar socioeconomic characteristics as well as diseases and vulnerabilities.

Methods: This was a case study using data from the Project Epidoso¹. In this study it was used a table with 207 variables, containing data from 1,666 people. To make the cluster analysis, 56 of 207 variables were chosen after discards for several reasons. It was used the algorithm TwoStep from IBM-SPSS to find out the optimum number of clusters with a closer look to socioeconomic characteristics as well as diseases and vulnerabilities from people who participated in the Project Epidoso. TwoStep was chosen due to no need to provide an initial number of clusters. This study was approved by UNIFESP’s Ethics Committee, process number CEP1890/08.

Results: Data were clustered into three groups, clearly segmented that can help better targeted audiences for public health promotion². Variables were sorted according to the importance in identifying groups. Within each cluster, variables were sorted again showing their inside importance. This method can show participation of each variable in a cluster by importance. Thus a general view of each cluster can be made to lead a targeted communication, choosing better channels for disseminate health information.

Conclusions: Cluster analysis showed interesting to identify groups of people for better design of health promotion to identified targeted groups. TwoStep is a good choice to identify an optimum number of clusters.

Implications for research and/or practice: Planning and Designing of a more target health promotion and communication.

25909 Cluster Analysis to Identify Groups for Health Promotion