Adoption of a Data-Driven Bayesian Belief Network-Investigating Provider-Related Domains that Influence Patient Experience

Abstract

Patient experience is a key performance indicator of healthcare quality. It is driven by different provider-related domains in patient journeys during hospital visits. Several studies have provided various insights on patient experience, related domains, and its associations with patient safety and clinical outcomes. However, there has been limited research investigating the interdependencies between multiple provider-related domains and patient experience. To address this research gap, this study has developed a probabilistic graphical model, Bayesian Belief Network (BBN). The aim is to explore the role and relative importance of provider-related domains influencing patient experience. The model is based on leveraging the British National Health Service aggregated patient experience survey data from 2017 to 2019. In this study, eight provider-related domains were explored along with overall patient experience score (i) information, communication, and education, (ii) respect for patient-centered values, preferences, and expressed needs, (iii) emotional support, (iv) confidence and trust, (v) coordination and integration of care, (vi) food choice, (vii) hydration, and (viii) respect and dignity. Three structural learning algorithms were attempted Bayesian Search (BS), Greedy Thick Thinning (GTT), and Tree Augmented Naïve Bayes (TAN), along with three discretization schemes. The prediction accuracy of each model was evaluated using 10-fold cross-validation. Overall, the three algorithms presented high prediction accuracy. However, the TAN algorithm with the three states discretization scheme outperformed GTT and BS. Moreover, further analyses were performed on the model, such as sensitivity and scenario-based analyses. The results showed that the most influential domains that lead to a high patient experience score are: (i) confidence and trust, (ii) respect for patient-centered values, preferences, and expressed needs, and (iii) emotional support. In conclusion, the constructed model is an enhanced method of assessing and evaluating patient experience derived from survey data. This study guidesresearchers and healthcare practitionersin understanding the role of various provider-related domains in supporting patient experience enhancement. The probabilistic graphical characteristics of the model allows users to visualize and quantify the causal relationship between the domains. Further, the additional analyses, such as sensitivity and scenario-based analyses, provide deeper insights into the effect of different hypothetical interventions and how the system responds. It also explains what leads to a specific outcome. Based on the importance of provider-related domains, this study results can help healthcare managers utilize and allocate their resources more effectively to improve overall patient experience during hospital visits.

Date of Award	May 2022
Original language	American English