Estimating Emergency Room Wait Times In Changepoint Weibull Hazard Model

This paper uses duration analysis to estimate emergency room wait times. To account for multi-phrasal hazard, analysis time is fitted in a piecewise model that allows hazard to change power and direction whenever it reaches a periodic length. Such model provides a marginal improvement from other competing models where hazard is assumed to be non-monotone or one-time increasing. Results, from applying the changepoint model in the 2010 National Hospital Ambulatory Medical Survey data, point to two major conclusions. First, the instantaneous rate to be seen follow non-monotonic decreasing path. Consequentially, a patient’s wait time is lengthened more aggressively for each instance he remains in the emergency room. Second, wait times are influenced by visit, patient, and hospital characteristics, in a descending order. Factors that correlate positively with wait times are female, Medicare/Medicaid insurance, low income, non-ambulance arrival, afternoon shift, and moderate urgency. In a hypothetical scenario where patients are segregated into two groups of opposite traits, the ones who possess the disadvantage characteristics are expected to have double to triple wait lengths compared to their counterpart. I would like to thank Dr. Zimmer for guidance and timely advising, Dr. Lebedinsky for his constructive critiques and revision, and Dr. Beckworth for his inputs in the very early stage of this project. I also thank my fellow students Chris, my friends Bobby, Emily, and Dr. Zhuhadar for their corrections and insights. All mistakes are mine.