Applying data analytics to national survey data to explore patient satisfaction with health care providers among adults with multiple chronic conditions

As the number of adults with multiple chronic conditions is projected to increase considerably in coming years, effective management and prevention of these conditions are vital to their health and wellbeing.  A growing body of literature has shown that patient satisfaction is an important indicator of quality of care and health outcomes.  This study aimed to evaluate patient satisfaction with care providers among the U.S. adults with multiple chronic conditions and assess the impact of their chronic conditions and socio-demographic characteristics on patient satisfaction.

We extracted data from the nationally representative 2011 Medical Expenditure Panel Survey.  Target chronic conditions were selected based on criteria from CDC (Centers for Disease Control and Prevention) using the responses of corresponding survey items (e.g., Have you ever been told by a doctor or other health professional that you have diabetes?).  Patient satisfaction was estimated using the responses of survey items related to patient experience with their providers.  We first conducted association analysis to determine the prevalence and co-occurrence patterns of the conditions. We then partitioned the population by different patterns using the outcomes of the association analysis (e.g., most frequent co-occurrence patterns of 2 conditions) to explore how patient satisfaction was influenced by different patterns.  A number of demographic and other characteristics were considered, including age, gender, race, income, education, and insurance.  We applied a decision tree method to iteratively construct classification models.  We selected final models based on their descriptive ability as well as meaningfulness by domain experts.  The comparative analysis of the models revealed distinct co-occurrence patterns of the chronic conditions and their relationship to population subgroups with varying satisfaction rates by demographic and other factors.  Data mining methods were useful to discover, analyze and visualize the risk factors related to patient satisfaction in the context of chronic disease management and prevention.