Patient text reviews and preference estimation

The goal of this paper is to illustrate how customer text reviews can be used to identify (a) the factors underlying consumers’ preference for a product offering and (b) the magnitude of each of these factors on the consumers’ overall assessment of the product offering experience. The authors do this using approximately 317k Google patient reviews for U.S. acute care hospitals. They first analyze the texts using Natural Language Processing and find eleven valenced topics well-describe the types of healthcare experiences. Then, after describing the structure of these reviews, they use regression analysis to estimate the magnitude of each type of experience on the patient’s overall evaluation of the experience after adjusting for any halo effect associated with the dominantly discussed topic, which has the potential of influencing the impact of the other discussed experiences. The authors conclude by providing numerous managerially significant insights coming from these analyses.

Duke Scholars

Author Richard Staelin Fuqua School of Business