The Affordable Care Act (ACA) has been promoted with two major goals: expanding health insurance coverage and reducing healthcare costs. Part of the cost reduction is expected to be achieved through the expanded coverage. The proposal is that uninsured patients make use of the emergency department of their local hospital as a primary healthcare option more often than an insured patient with a comparable condition. The insured patient will visit a primary care physician for certain ailments because that visit will be covered. By contrast, the uninsured patient will go the emergency department knowing that they legally cannot be denied treatment because of their insurance status.
Since emergency department visits are most costly than a comparable visit to a primary care physician, this pattern of behavior could contribute to higher healthcare costs. Further, if an uninsured individual delays seeking treatment because of cost concerns, when they finally do seek treatment they may be in a more critical condition, requiring additional care than if they had been seen earlier, and possibly developing more complications and needing a longer period of recovery. All of these factors can further drive up costs beyond the difference inherent in usage of the emergency department.
However, insurance status is not the only factor driving emergency department utilization. It has been observed in some hospitals that nonurgent visits to the emergency department are primarily the result of uncertainty on the part of the patient as to the true severity of their condition. While trained medical personnel could distinguish the urgent and nonurgent cases after the fact, the initial presentations of both classes of patients were very similar. Access to different healthcare options also affects decision making, as do other socioeconomic and educational factors. In that context, changing insurance status may not actually decrease emergency department utilization.
EpiCenter has been collecting data on emergency department usage for nearly a decade in several states. The latest version of the American Community Survey included population-level data on coverage for several classes of health insurance by geography. Combining these two data sources provided a unique opportunity to ask the question of how insurance coverage relates to emergency department usage at the population level, and how it compares as a predictor of usage relative to other traditional factors in healthcare discrepancies such as poverty rate and racial population distributions.
Our first observation is that emergency department utilization by residents of a zip code declines as the distance between a zip code and the nearest emergency department increases. This effect is not just for patients using the emergency department nearest to their home zip code, but for all emergency department visits. It suggests that access to healthcare alternatives is an important consideration in healthcare utilization decisions.
It has been widely observed that the number of people with an income below the federal poverty standard in a geographic area is a significant predictor of healthcare-related discrepancies. We divided zip codes into four groups using standard cutoffs for the population rate of individuals with income below the federal poverty standard. These groups are defined as less than 5% of the population with income below the federal poverty standard, between 5% (inclusive) and 10% (exclusive), between 10% (inclusive) and 20% (exclusive), and greater than or equal to 20% of the population.
This plot shows the distribution of emergency department usage for each of the four groups of zip codes. The median of the usage rate for each group increases with the increasing percentage of people with incomes below the poverty standard. While the range for the first category in particular is comparable to the range for the final category, the central mass of the distributions do not fully overlap. The fact that the notched regions do not overlap provides a heuristic indication that there is a statistically significant difference between the groups.
This plot demonstrates that the percentage of the population in a zip code with public insurance coverage is also an indicator of emergency department usage; the more people in a zip code with public insurance, the more visits to the emergency department were observed. This holds true for just the rate of Medicaid insurance in a zip code as well. Both categories are more strongly associated with more emergency department usage than the rate of not having any insurance coverage. If this observation bears out, it could have implications for ACA implementation, particularly in those states which are increasing Medicaid coverage to meet ACA goals.
As with any study of aggregate data, there is the possibility of conflating group observations with individual effects. Before concluding anything about the role that insurance coverage plays in healthcare seeking, it would be necessary to replicate these correlations at an individual level. For example, the population of people with Medicaid coverage is not necessarily representative of the general population, since certain medical diagnoses can qualify a person for Medicaid coverage. These diagnoses may predispose an individual to require more emergency department visits. Thus it would be important to compare usage rates between individuals who are matched as well as possible in terms of their health status.
With these caveats in mind, the consistent association between the distribution of insurance options within a zip code and the emergency department usage of residents from that zip code seems strong enough to warrant further investigation. In particular, the effects observed persist after adjusting for age distributions. Given the relatively recent availability of census-level insurance information, it should be expected that further investigations of this kind will refine our understanding of the relationship between insurance coverage and healthcare decision making.