A sociodemographic analysis of health care utilization in the United States

Elizabeth A Dobis, Purdue University

Abstract

How people use health care is important knowledge for public policy and private businesses. Understanding where people go for care and how often they demand medical services enables governments and health care providers to produce healthy populations at minimal cost. The analysis of health care utilization is especially timely given the current debates and legislation with respect to national health care in the United States. Using 2008 data from the Medical Expenditure Panel Survey, this thesis looks at factors that influence health care utilization. It is widely believed that vulnerable populations use emergency rooms for regular care, thus driving up health care costs. One part of the thesis focuses on the type of provider a patient chooses, using a nested multinomial logit model to see what factors influence his or her choice of a usual source of medical care (specifically office, outpatient, or emergency room). Another part of the thesis discusses how many times per year people visit a medical professional. This frequency of health care analysis employs negative binomial and finite-mixture count data models to view utilization in terms of both sheer number and classes of intensity of use. In both the logit and count model analyses, a distinction is made between vulnerable and non-vulnerable populations. These two population groups are derived using a hierarchical cluster analysis. Results from our work on the mode of health care utilization indicate that individuals with health insurance or those who perceive their health as poor are more likely to have a usual source of care at an office or non-ER hospital. Not feeling comfortable speaking English, conversely, decreases the likelihood that an individual will have any type of usual health care provider rather than no usual source of care. Additionally, individuals that reside in the Northeast region of the United States are most likely to have a usual source of care, whereas Southerners are least likely to have a regular medical provider. The analysis pertaining to the frequency of health care use reveals the individual characteristics that have the largest effect on the frequency of health care utilization are pregnancy and having a usual source of medical care. The influence of economic variables is the same for both vulnerable and non-vulnerable population subgroups. Both a higher income and holding health insurance increases the number of medical visits due to ease of access, whereas the time constraints of being employed decrease the annual number of medical visits of an individual. When utilizing the finite mixture model for the empirical analysis we see that there is a larger difference in the means between "healthy" and "ill" populations than between the means of our vulnerable and non-vulnerable population subgroups. Ultimately, we find that a different set of individual characteristics influences the behavior of vulnerable and non-vulnerable populations. The characteristics that influence choosing an office or non-ER based medical provider versus not having a usual source of health care are different for vulnerable and non-vulnerable populations. However, the factors influencing the choice of an ER as a usual source of medical care rather than not having a usual source of care differs only slightly between vulnerable and non-vulnerable populations. The empirical results also show that vulnerable and non-vulnerable populations consume medical services at different frequencies. When using a basic count model, we find that vulnerable populations have more medical visits in a year than non-vulnerable populations. The finite mixture model sheds more light on this result, indicating that this difference is mostly due to the large number of visits of "ill" individuals from the vulnerable population cluster, rather than the "healthy" individuals.

Degree

M.S.

Advisors

Florax, Purdue University.

Subject Area

Economics|Health care management

COinS