Search

Introduction: Data mining in healthcare is a nascent arena of research in healthcare. Heterogeneity of Diabetes Mellitus in terms of clinical presentation calls for newer methods of research to study potential risk factors. Aim: The paper aims to use clustering techniques to identify the relationship between the four variables, namely the pre-prandial and postprandial sugar level, age and sex. Methods: The data was taken from a diagnostic laboratory in Wagholi, Pune. We conducted K-mean algorithm, EM algorithm, model-based clustering and t-mixture model. Results: It is evidenced that the data was best fitted to the t-mixture model. Our 50% samples were people with diabetes, 17% had prediabetes. Trivial correlation existed between age and sugar level. Males and females were equally at risk of having diabetes. Data presented concludes that age and sex have no effect on the risk of having diabetes. Data mining can be used to deduce meaningful clusters to drive plan-based interventions in the population. Conclusion: Methods of data mining can be used to deduce meaningful clusters in a heterogeneous dataset thus providing policymakers and healthcare researchers with novel information that will potentially contribute in formulating evidence-based policies. [ABSTRACT FROM AUTHOR]

Community-based groups (CBGs) are a frequent feature of community mobilization. In HIV/AIDS prevention programs in India, CBGs and networks are seen as a vehicle to strengthen demand for services and manage programmatic activities. Community mobilization is an important component of a participatory approach to health and development interventions.1 and2 have developed a methodology for measuring community mobilization score using community ownership preparedness index (COPI). These authors further use COPI to assess the effectiveness of the community mobilization in a 10-year, large-scale HIV Prevention intervention program, viz., "Avahan, the India AIDS Initiative". The COPI was calculated as a weighted average of 22 sub index scores, where the weights are based on the domain knowledge (or perception). In this paper, we attempt to develop a new data-driven (adaptive) composite index to measure the community mobilization score of a CBG. The suggested data-driven composite score index has been calculated using multivariate statistical methods such as principal components and factor analysis. The nearness of the suggested index to COPI has been established, thus justifying the use of COPI as a community ownership and preparedness index. The effectiveness of the program over the period of evaluation can be clearly seen from the improvement in the adaptive scores during the period 2009 to 2012. [ABSTRACT FROM AUTHOR]