The Extra Costs of ED Visits Associated with Multiple Chronic Conditions

Purpose: This paper compares hospital emergency department (ED) use for those with and without multiple chronic conditions (MCC).  The twofold purpose is (1) to test the hypothesis that ED visits made by patients with MCC are more likely to be non-urgent than those made by patients without MCCs, controlling for relevant confounding variables, and (2) to estimate the extra costs of ED treatment associated with the presence of MCC using a national sample of ED visits from 2007-2012.

Background: Multiple chronic conditions (MCC), defined as having two or more physical or mental/cognitive chronic conditions, are highly prevalent in the United States.  According to Medical Expenditure Panel Survey data from 2006, one in four Americans of all ages has MCC, with a greater burden among older adults (Anderson 2010).  In dollar terms, MCC accounts for approximately 66% of total health care national spending.  ED visits also account for a large part of national health care spending and are often the target of interventions for health system improvements.  However, little is known about MCC as a driver of the rising trend of ED use and their long-term impacts on the patterns of utilization and costs of services used by ED visitors.  Both MCC and increasing ED use present health system challenges to practitioners, policy makers, and researchers.  As ED use increases over time, more EDs become overcrowded increasing the risk of medical errors, ambulance diversion of emergent patients, reduced ability to respond to disasters, and decreased reliability of the emergency care system (Trzeciak and Rivers 2003).

Methods: We will use the HCUP ED discharge data to determine the probability of non-urgent ED visits and the extra costs associated with the burden of chronic diseases in a multivariate framework.  The HCUP ED discharge dataset comprises discharge data for ED visits from 950 hospitals located in 30 States, representing approximately a 20-percent stratified sample of U.S. hospital-based ED visits.  Our treatment group comprises ED visits by patients with 2 or more chronic conditions and the definitions of MCC are similar to those used by Lochner, Goodman, Posner and Parekh (2013).  The treatment ED visits will be compared with a control group of ED visits selected by propensity score matching.  Our two main outcome variables are (1) the probability (0 – 1) of the having a non-urgent ED visits defined using the definition of the NYU Non-urgent ED Algorithm (Billings et al. 2000) and (2) the cost of care per ED visit.  The key control variables include patient characteristics, patient’s severity of illness or degree of urgency when presented at ED, hospital characteristics, and local market characteristics.  The key variable of interest is a dummy variable which takes on the value of 1 if the visit is made by a patient with MCC and 0 if by a patient without MCC.  We will also present predicted costs (together with their 95% CI) for each of the predictors holding other variables at their mean values for the treatment group and the control group.