Electronic copy available at: http://ssrn.com/abstract=1226522
Is Fragmented Financing Bad For Your Health?
Steven D. Pizer 1,2
John A. Gardner 1
January 7, 2011
Americans finance health care through a variety of private insurance plans and public
programs. This organizational fragmentation could threaten continuity of care and
adversely affect outcomes. Using a large sample of veterans who were eligible for
mixtures of VA- and Medicare-financed care, we estimate a system of equations to
account for simultaneity in the determination of financing configuration and the
probability of hospitalization for an ambulatory care sensitive condition. We find that a
one standard deviation change in financing fragmentation increases the risk of an adverse
outcome by one fifth.
Health Insurance; Outcomes; Coordination of Care; Ambulatory Care Sensitive
1 Health Care Financing & Economics, US Department of Veterans Affairs
2 Health Policy and Management, Boston University School of Public Health
Steven D. Pizer
Associate Director, HCFE
150 South Huntington Ave.
Mail Stop 152H
Boston, MA 02130
This article is forthcoming in the summer 2011 issue of Inquiry. Please cite accordingly.
Electronic copy available at: http://ssrn.com/abstract=1226522
Health care financing in America is fragmented. Individual patients may pay for
health services directly out of their own pockets, indirectly, through one (or sometimes
more than one) of a multitude of private health insurance plans, or through one or more
public programs. Even within a public program like Medicare individuals can choose
from a variety of health plans with different combinations of coverage and premiums,
adjusting their plan choices when their tastes or circumstances change. This variety of
health care financing choices has value to consumers (Danzon, 1992), but it also has a
cost because it makes continuity of care more difficult.
Continuity of care has been defined as a long-term relationship between a patient
and a physician, regardless of the presence of any specific disease (Haggerty et al., 2003),
and it is believed to be a critical determinant of quality in primary care (Starfield, 1998).
Continuity of care is associated with lower probability of hospitalization (Gill, 1997;
Mainous and Gill, 1998), fewer visits to the emergency room (Gill et al, 2000), and better
immunization (Mark and Paramore, 1996). Short or discontinuous relationships with
providers could lead to lower quality of care because the provider’s access to information
about the patient’s health will be more limited than it would be in long, continuous ones.
When access to information is difficult, physicians may be more likely to choose
treatments based on norms or past experience with other patients instead of treatments
optimally tailored to an individual’s particular condition and history. Frank and
Zeckhauser (2007) use the colorful term ―ready-to-wear‖ to characterize a treatment
strategy that relies on norms, in contrast to ―custom-made,‖ which describes a treatment
strategy that makes use of detailed individual-specific information.
The goal of this paper is to assess empirically whether disruption in clinical
relationships, resulting directly from fragmented financing, has a meaningful effect on
health outcomes. We have chosen a very broad outcome: hospitalization for ambulatory
care sensitive conditions (ACSC hospitalization), which is a widely used and accepted
measure that can be constructed easily from administrative data. To measure
fragmentation we focus on a population of low-income or disabled veterans who choose
mixtures of services from private networks financed primarily by Medicare and the
public network operated by the Veterans Health Administration (VA). We are able to
observe all the utilization for this population, and, because the VA network and private
networks have very little overlap, we can infer discontinuity of care from fragmentation
of financing with few errors.
Low-income or disabled veterans use combinations of Medicare and VA care for
a variety of reasons. Low-income veterans can receive comprehensive health care from
the VA at minimal financial cost, but large numbers regularly receive care from non-VA
sources because VA services may not be available or may require patients to wait or to
travel (Shen et al. (2003); Burgess and Fiore (1994); Prentice and Pizer (2008)).
Fleming et al. (1992) document the degree of dual use of VA and Medicare health
services among veterans who are enrolled in Medicare. They describe an equilibrium
that exists between VA and Medicare/Medicaid programs, mediated by eligibility
requirements and financial controls on both sides. When Medicare or Medicaid
eligibility rules tighten or out-of-pocket costs increase, more veterans turn to VA for
services. Fragmented financing between VA and Medicare can therefore reduce out-of-
pocket costs and improve convenience for individual veterans, but continuity of primary
care may be undermined.
Considerations like these apply to non-veterans as well. At a minimum,
Americans may change health coverage when they graduate from college, change jobs,
get married or divorced, move from place to place, or retire (Currie and Madrian, 1999).
If these changes in coverage are accompanied by disruptions in relationships with health
care providers, fragmented financing could substantially undermine the quality of care.
Cunningham and Kohn (2000) used data from the Community Tracking Study to show
that about 17% of privately insured Americans changed health plans in 1996-7. Those
who changed plans were twice as likely as those who kept their plan to change their usual
source of care (22.5% compared to 11.2%). When respondents switched from one HMO
to another, the likelihood of changing their usual source of care increased further to
28.6%. One reason to change usual source of care is because the new plan does not
include the previous provider in its network. Chernew et al. (2004) examined the degree
of network overlap among HMOs in the same metropolitan statistical areas and found
that a member switching from one HMO to another in the same market had about a 50%
chance of having to change primary care physicians.
When measuring the effect of fragmented financing on outcomes we face an
endogeneity problem. Individuals in poor health will be more likely to have fragmented
financing as well as bad outcomes. There are both institutional and individual reasons for
this. Institutionally, those who are 65 or older or receive Social Security Disability
Income can qualify for Medicare. Individually, as health status deteriorates, demand for
quantity and variety of services increases, making the use of multiple programs more
attractive. We can use risk-adjustment techniques to mitigate this problem, but
unobserved variation in health status will remain, causing naïve estimates of the effect of
fragmentation on outcomes to be confounded with the effects of unobserved health. To
address this problem, we need an instrumental variable that predicts fragmentation and is
independent of unobserved individual health status.
We use the distance between the patient’s home and the nearest VA facility as an
instrument to explain some of the variation in fragmented financing in this population
and identify the causal effect on outcomes. We find that a one standard deviation
increase in fragmentation of financing increases the probability of ACSC hospitalization
by 21%. This effect is comparable in size to the effects of major chronic diseases like
arrhythmia, anemia, or peripheral vascular disease.
The remainder of the paper is organized in four sections. We present the
statistical specifications for our models and discuss estimation and testing issues in the
next section. In the third section we describe our sources of data and how the sample
was derived. The fourth and fifth sections present results and conclusions, respectively.
2. Statistical Specification
Our goal is to measure the effects on health outcomes of potential coordination
failures arising from fragmentation of health services financing. Our principal empirical
challenge is that observed and unobserved health status simultaneously determine
financing arrangements (through eligibility policies) and outcomes, so fragmented
financing is endogenous with respect to the outcome. For example, a negative shock to
unobserved health might simultaneously cause fragmentation and poor outcomes because