Using Surrogate Measures To Predict Patient Satisfaction In The Emergency Department

With healthcare organizations struggling to remain competitive and financially stable in a market where minimizing costs is a priority, hospital administrators feel the sense of urgency when it comes to keeping patients satisfied with services in order to expand volume and market share. The Emergency Department is considered the front door of a healthcare organization, and keeping its visitors satisfied in order to guarantee a future visit or a referral to a friend or family member is a must. While patient input on the services received in a healthcare facility is essential to improving quality of care, the costs associated with measuring, collecting and analyzing their feedback are remarkable. This research focuses on developing a linear regression model to predict patient satisfaction in the ED using surrogate measures related to patient\u27s socio-demographic characteristics and visit characteristics. With a model of this kind, healthcare administrators can potentially eliminate survey costs while still being able to determine where the hospital stands in the eyes of the patient. Three modeling approaches were used to develop a multiple regression equation. Modeling approach 1 used monthly patient satisfaction scores as the dependent variable collected by a third-party survey organization. The goal of this model was to predict monthly patient satisfaction scores. Modeling approach 2 used patient satisfaction scores collected by the discharge registrar prior to the patient leaving the ED. The goal of this model was to predict patient satisfaction scores on a patient-by-patient basis. Modeling approach 3 used patient satisfaction scores collected by a third-party survey organization. The goal of this modeling approach was to predict patient satisfaction scores on a patient-by-patient basis. Each modeling approach developed in this study used its own survey tool. Though this study had limitations when it came to developing the models and validating the findings, results are very promising. Analysis shows that predicting average patient satisfaction scores on a monthly basis gives the most accurate results, with socio-demographic characteristics and visit characteristics explaining 96% of variation in monthly average patient satisfaction scores. Other model indicators, such as normality of residuals, predicted error, mean square error, and predicted R-square show that the model fits the data very well and has strong predictive ability. Models that attempted to predict patient satisfaction on a patient-by-patient basis appeared to be ineffective, with very large predicted errors and prediction intervals and low predictive ability