Assessing Patient-Centered Communication Using Decision Tree Classification

Patient-centered communication is increasingly seen as a critical requirement for improving health care. The primary objectives of this study were to identify subgroups of U.S. adults who were likely to be satisfied (or dissatisfied) with the communication with their health care providers, and to explore demographic and other characteristics that were most related to communication satisfaction in the context of patient-centered care.

We used decision tree induction for secondary data analysis. Data for this study were obtained from the nationally representative 2005 Medical Expenditure Panel Survey. The outcome variables were determined from the responses of 4 survey items related to communication with health care providers (e.g., how often providers showed respect for what you had to say). We examined each item individually, as well as created a composite value of the items for an outcome measure. The outcome responses on a 4-point Likert scale were converted into binary variables. As predictors, several demographic and other variables were used in decision tree construction, including age, gender, race, ethnicity, income, education, insurance, marital status, spouse, health status, residential area. We evaluated the performance of the constructed decision tree models using cross-validation and independent test data sets. The resulting classification models identified distinct subgroups of the population with varying satisfaction rates, thus helping us better understand dissatisfied subgroups. And a comparative analysis of the models provided particular sets of demographic and other characteristics that contributed most to the partitioning of the population. A comprehensive and accurate assessment of patients' communication needs could be vital to the delivery of enhanced health care that is more concordant with their values and perspectives.