Effects of Health IT on Regional Health Care
The Spillover Effects of Health IT
Investments on Regional Health Care Costs
Full Paper
Hilal Atasoy
Pei-Yu Chen
Department of Accounting, Temple
Dept. of Information Systems, Arizona
University
State University
[email protected]
[email protected]
Kartik K Ganju
Department of Information Systems, Temple University
[email protected]
Abstract
In this paper, we examine the effect of the adoption of Electronic Medical Records (EMR) systems to
neighboring hospitals. We find that although EMR systems increases the operational costs for adopting
hospitals, it has significant spillover effects by reducing the health care costs of the other hospitals in the
same region (possibly due to patient mobility and better care that is provided by these systems). These
regional externalities are stronger especially in the long term. Our results provide support to the role of
EMR investments in reducing overall health care costs. Estimates, based on our results, suggest that EMR
investments can lead to net reduction in national health care cost by about $47 billion dollars over 4
years.
Keywords
Health IT, Spillover effects of IT, IT and Society
Introduction
The cost of health care continues to be the subject of several policy debates in the US. In 2010, health
expenditure in the US was $2.6 trillion, and this figure is estimated to grow to $4.6 trillion in 2020,
outpacing the expected growth rate in GDP in the corresponding period (CBO 2008; Centers for Medicare
& Medicaid Services 2010). The US health care system is criticized as fragmented and uncoordinated
leading to large number of preventable medical errors and wasteful resource allocation (McCullough et al.
2013). These problems have been estimated to cause an operational waste between $126 billion and $315
billion in the US health care industry (PwC Health Research Institute Report 2010). Considering the
alarming level and growth of medical costs, it has been a primary interest of policy makers to find possible
solutions to mitigate them. One example is the 2009 Health Information Technology for Economic and
Clinical Health Act (HITECH Act), part of the American Recovery and Reinvestment Act, which allocates
around $19 billion to increase the adoption of electronic medical records (EMR) by health care providers.
This policy provides financial incentives for digitizing records, as well as imposing penalties on
institutions that do not comply. The underlying belief for these public subsidies is that EMR adoption
would lower the health care costs and improve the quality of care: EMR is expected to improve
communication and information exchange, leading to better diagnostics, quality care, which in turn can
translate into lower costs with a reduction in medical errors, unnecessary re-admissions, over-testing, and
ER visits.
Despite these promises, there is still a lack of consensus on the impacts of EMR adoption.
Several studies have found that EMRs can lead to improvements in health care quality and patient
outcomes such as lower mortality rates, improved diagnostics, and fewer medical errors (Athey et al.
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
1
Effects of Health IT on Regional Health Care
2000; Borzekowski 2009; Buntin et al. 2011; Chaudhry et al. 2006; McCullough et al. 2013; Miller et al.
2009). In other words, EMR investments can benefit the patients through increased care quality.
However, the evidence on the relationship between EMR adoption and health care costs is limited and
characterized by mixed results with both negative (Athey et al. 2000; Hillestad et al. 2005) and positive
correlations (Agha 2013; Borzekowski 2009; Dranove et al. 2012). Overall, it is not clear whether and how
improvements achieved in health care quality with EMR adoption lead to lower costs of providing care.
These findings not only question the effectiveness of EMRs as a way of reducing health care costs, but also
suggest that hospitals may not be able to fully internalize the benefits from improvements in health care
quality from their EMR adoption, which can lead to insufficient incentive to further invest in health IT.
In this paper, we argue that some benefits of EMR adoption might not be internalized by EMR adopting
hospitals but are realized in the form of spillover effects to other hospitals, in addition to improved care
for patients. Most of the literature focuses on the impact of EMRs at the hospital level. While studying
within-hospital impact is important as it affects the adopting hospitals’ incentive to invest in EMR, a
decision or policy made based entirely on a hospital-level analyses can be misleading if the impact goes
beyond adopting hospitals.
To address this void, we analyze the spillover effects of EMR adoption of each hospital on the costs of
neighboring hospitals in the same Health Service Area (HSA). We argue that although the adoption of IT
systems is known to increase the operational cost of firms due to the training that needs to be undertaken,
there can be spillover effects to neighboring hospitals that are able to benefit due to the increased level of
care and better medical records that are provided at the adopting hospital. With this we seek to study:
1.
Does EMR adoption create regional spillover effects by reducing health care costs of other
hospitals in the same region? What is the degree of the spillover effects?
2. What factors (such as the type of IT system and the number of years after the EMR system has
been adopted) affect the strength of spillover effects?
Our findings indicate that while EMR adoption increases operational costs for the adopting hospital itself
for the initial year, it significantly decreases the total costs of the other hospitals in the same HSA in the
subsequent years. These findings provide evidence for the cross-hospital externality and show that EMR
adoption, while costly for the adopting hospitals as they incur high expenditures, can lead to cost
reduction for other hospitals in the area through patient sharing, especially in the long-run. Based on back
of the envelope calculations, EMR investments can lead to net reduction in national health care cost by
about $47 billion dollars over 4 years
Originality Theoretical Development
Background Literature
Several studies analyzed the effects of EMR adoption on health care quality. Systematic reviews of the
literature indicate that majority of the studies found positive results of the adoption of EMRs (Buntin et
al. 2011). Others concluded that EMRs lead to increased adherence to guideline-based care, enhanced
surveillance and monitoring and decreased medication errors (Chaudhry et al. 2006). A wide body of
literature has found that EMRs can improve care performance and patient outcomes (Athey et al. 2000;
Borzekowski 2009), with a few studies suggesting negligible impacts (Agha 2013). Studies have also
argued that EMRs may help hospitals coordinate care especially for complex cases in which patients
require regular assistance between different departments (McCullough et al. 2013). Overall, these results
suggest that patients do benefit from EMR investments with higher quality care.
For hospitals adopting EMR, an important concern is whether these investments help reduce health care
costs—a major concern for both hospitals and the society. The adoption of EMR technologies is expensive
with the typical installation costs ranging between $3 million for a 250-bed hospital and $7.9 million for a
500-bed hospital and corresponding operating costs ranging between $700,000 to $1.35 million (CBO
2008). The evidence on the relationship between EMR adoption and health care costs is more scarce and
characterized by inconclusive conflicting results. Hillestad et al. (2005) use results from previous studies
and extrapolates the potential cost savings of EMRs net of adoption costs. They estimate that if 90 percent
of the US hospitals adopt EMR, potential savings could add up to $80 billion over fifteen years. Recent
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
2
Effects of Health IT on Regional Health Care
studies that evaluate more comprehensive set of EMR systems at the hospital level found that EMR
adoption increases operational costs (Agha 2013; Dranove et al. 2012). In other words, it is not clear
whether and how previously discussed improvements in health care quality from EMR adoption translate
into reduced costs. The evidence has been divided on whether these technologies compensate their initial
large implementation costs. These results suggest that hospitals may not be able to fully internalize the
benefits of improvements in health care quality due to EMR adoption, indicating potential incentive
issues regarding EMR adoption.
Previous studies have focused solely on hospital level effects of EMRs, however, due to cross-hospital
externality arising from shared patients, hospital-level effect may underestimate the benefits of EMR,
leading to a potentially insufficient investment from society’s point of view. There can be geographical
network externalities exceeding the hospital level impacts due to spillovers at the regional level. There is a
long stream of economics literature on externalities starting with Katz et al. 1985. Several IT investments
are also subject to positive externality as documented in the literature (e.g. Katz et al. 1985; Markus et al.
2006; Zhu et al. 2006). Moreover, IS research documents IT spillover effects through several mechanisms
such as labor mobility, customers and inter-industry transactions (Chang et al. 2012; Cheng et al. 2007;
Cheng et al. 2012).
Current spillover effects of EMR adoption on regional health care costs
There are two forces for the potential impacts of EMR adoption on operational costs for the adopting
hospital. First, EMR investments are usually expensive, and therefore will directly increase the costs,
especially in the current period. EMR adoption can also increase the future cost levels, as these systems
have significant maintenance and update costs (CBO 2008). On the other hand, these technologies can
introduce several improvements in care quality that can in turn lead to reduced costs. EMR applications
are found to increase health care quality by enabling better diagnostics and decision-making tools for
physicians, and better patient control and tracking.
Hospitals are often connected to each other through sharing patients, because patients may be
examined/treated in or admitted to different hospitals (at different time) in the same region (Huang et al.
2010; Lee et al. 2011; McCullough et al. 2013). Patient sharing usually happens through two channels:
direct sharing where a patient is referred from one hospital to another, and indirect sharing where a
patient is admitted to another hospital after a certain time without referral from the first hospital. Lee et
al. (2011) conduct a detailed social network analysis of hospitals in the Orange County, California, and
they find that 87% hospitals share patients. With the same data, Huang et al. (2010) find that 94% of the
inter-hospital patient sharing happens indirectly, meaning initiated by the patient without a referral
between hospitals. These studies find significant effects of patient sharing on hospital outcomes such as
infectious disease spread and control, patient education and prevention programs.
We expect this patient mobility to create spillover effects via two main mechanisms: First, better
diagnostics and care quality received in one hospital can improve patient health at the time of admission
to another hospital. Second, information sharing can decrease unnecessary tests, imaging and
procedures. For example, let’s consider a cancer patient who is diagnosed and treated at Hospital A for a
certain amount of time, and this hospital is equipped with advanced EMR systems that enable an accurate
diagnosis and high quality care for the patient, as well as electronic records for her history. When this
patient transfers to Hospital B, either directly by a referral or indirectly for other reasons, this transfer
might reflect in the costs of Hospital B. First, the patient can be in a better health condition at the time of
admittance to Hospital B because of the high-quality diagnosis and care she received in Hospital A, and
this can reduce the care costs for Hospital B. Second, if the some information and records are transferred
to Hospital B (by the patient or between hospitals), Hospital B will not incur costs of series of tests and
imaging that have already been done. The adoption of EMR systems does not directly lead to a seamless
electronic exchange of information between hospitals even if they adopt the same software system for
EMRs. For this to occur, hospitals need to implement a Health Information Exchange (HIE). However,
better record keeping by one hospital enables an easier transfer a data to neighboring hospitals via
patients even in the absence of hospitals adopting HIEs. Additionally, hospitals making Electronic Health
Records (which are subsets of EMRs) available to their patients would strengthen our results, as
Electronic Health Records are more transferable between hospitals and can be monitored by the patient
as well.
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
3
Effects of Health IT on Regional Health Care
In sum, these two mechanisms, improved patient health due to EMRs and information sharing (via
patient or between hospitals), create spillover effects to different hospitals in a region through shared
patients. We argue hospitals that are located in the same region can affect each other’s cost combining the
network externality theory and evidence from medicine and healthcare literatures. Particularly, we
hypothesize the spillover effects as follows:
H1: EMR adoption in one hospital leads to lower operation costs at other hospitals in the same region.
Lagged spillover effects of EMR adoption on regional health care costs
There is reason to believe that the benefits resulting from EMR adoption may take time to be realized. An
impressive body of literature on positive returns of IT on productivity and output indicates that returns of
IT are generally larger over the long term (e.g. Brynjolfsson et al. 2003; Devaraj et al. 2003; Peffers et al.
1996). There are several theoretical reasons for lagged returns of IT investments. First, organizations do
not simply adopt IT and enjoy benefits—they go through a challenging process of changing their existing
processes and tasks (Bresnahan et al. 2002). The benefits of IT are thus accompanied by large and timeconsuming investments in complementary processes and assets (Brynjolfsson et al. 2003; Brynjolfsson et
al. 2002), and these improved complementary changes lead to larger returns from IT over time. In the
context of EMR systems, Dranove et al. (2012) indeed show that the complementary resources and
changes are essential for hospitals to be able to achieve cost reductions from EMR adoption.
Additionally, we expect the spillover impacts of EMR adoption on the other hospitals in the region to be
stronger in the long-term as patients are more mobile across hospitals over time. Better diagnostics,
improved patient tracking, lower medical errors and disease control and prevention enabled by EMRs are
more likely to result into better health of patients in the future. Combined with increased patient sharing
and mobility over time, there can be a stronger decrease in redundant testing longitudinally. We expect
that, through these mechanisms, the spillover impact of IT on cost reduction will be stronger in the long
term. Thus, we propose:
H2: Long-term spillover effects of EMR adoption on health care costs exceed current impacts.
Spillover effects by EMR application characteristics
EMR application type can affect the relationship between health IT and costs and therefore can have
implications on government policies regarding the level of EMR investment that should be targeted. IT
applications in general vary in purpose and scale and can have differential effects on outcomes, such as
productivity. Health care literature makes a distinction between basic and advanced EMR applications
(Dranove et al. 2012). Basic EMR systems allow hospitals to replace paper forms and records with
electronic copies. Technologies that enable different departments in the hospital to connect and exchange
information with each other as well as allow physicians to code patient observation chart, and computer
assisted diagnostics are known as advanced EMRs (HIMSS 2011). We adopt this conventional distinction
and differentiate between regional spillover effects of basic and advanced EMR adoption on health care
costs and examine these technologies separately to assess the differential impact they would have on
spillover
operational
costs.
Data and Model Specification
We obtain hospital level EMR adoption information from the Healthcare Information and Management
Systems Society (HIMSS) database. The database obtains information of hardware and software adoption
for healthcare providers across the nation. The data is collected via a survey administered to the different
hospitals. HIMSS database is used in several studies, as it is the most representative and detailed
database for health IT adoption in the United States (some recent examples are Dranove et al. 2012,
McCullough et al. 2013).
To measure the spillover effects, we consider hospitals that are in the same Health Service Areas (HSA) as
a network. HSAs are the nationally legislated local health care markets for hospital care that are created
for health planning purposes. An HSA is a collection of ZIP codes whose residents receive most of their
hospitalizations from the hospitals in that area. A health service area can be a county, a large metropolitan
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
4
Effects of Health IT on Regional Health Care
1
city, a state, or some other geographical unit that meets the HSA criteria . We choose HSA as a network
since it is the most granular and well-defined health care market. The spillover effects can certainly go
beyond the HSA, we can expect the overall spillover effect to be stronger when measurement is not limited
to a HSA. Since HSA is the smallest meaningful market for hospital care, our estimates present a lower
bound estimate of spillover effects compared to a case where patients are mobile beyond HSA. We also
conducted analysis where the spillovers effects are calculated at the Hospital Referral Region and county
levels, and results are qualitatively similar.
We merge hospital level EMR adoption data with Medicare cost data to construct a hospital level panel
2
between 1998-2010. Since our focus is on the HSA spillover effects, we drop HSAs in which there is only
one hospital, as spillover cannot be calculated. This leads to 2595 hospitals in 753 HSAs over our panel,
and an average of around 3.5 hospitals per HSA. The Medicare database contains yearly performance
information from hospitals that have reported costs to Medicare for their reimbursement rates. We use
the database to collect information of a number of hospital characteristics such as operational costs,
number of bed admittances, and number of employees, number of beds, Medicare and Medicaid
discharges. To calculate the HSA level demographic control variables such as population, age, education;
we use American Community Survey data at the Zip code level. We first match the Zip codes to the HSAs,
and then calculate the average statistics for each HSA, weighted by Zip codes’ population. Census County
Business Patterns data is used to calculate the high-tech industry intensity in the region. We measure
these complementary resources in the county as the ratio of high-tech establishments to total number of
3
establishments . We also control for number of major universities (top 100 nationwide) in the HSA as
they facilitate technology infrastructure in the area and attract skilled labor.
Measuring EMR adoption
For the independent variable, we measure the change in the level of EMR adoption for each hospital for
each particular year. There are five EMR applications reported in the HIMSS data: Clinical Data
Repository, Clinical Decision Support System, Order Entry, Computerized Physician Order Entry,
Physician Documentation. Table 1 summarizes the different EMR systems that we examine.
Table 1: Different EMR Applications
EMR application
Description
1998
2010
Clinical Data Repository (CDR)
Database that is used to maintain an update
record of the patient
39%
79%
Helps medical practitioners with diagnosis
and treatment plans
42%
77%
Order Entry (OE)
Allows hospitals to replace paper forms
with electronic documents
39%
81%
Computerized Physician Order Entry
(CPOE)
Systems contain information about the
specific needs of the patient
5%
38%
Clinical
(CDSS)
Decision
Support
System
1 HSAs are defined in the National Planning and Health Resource Act of 1974. The criteria for determining a HSA can
be found in the beginning of Part B of the act: https://www.adph.org/ALPHTN/assets/history_law.pdf
2 We include hospitals that have cost data for more than 3 years (from 1998-2010). For hospitals that have data
missing for particular years, the data is interpolated. However, if the cost or bed admittance or number of employees
data is negative, that hospital is dropped from the sample. This leaves us with 5099 hospitals for which we have data
for 13 years. This operational data for the hospitals is then merged with the IT data of the hospitals. For the five
technologies described above, we code the technology variable as 1 if the status is defined as “Live and Operational” in
the hospital and 0 otherwise. For missing years, we interpolate/extrapolate the based on the lagged values. For
hospitals where we do not have any IT information, we set the IT variable equal to 0.
3
The BLS high-tech sector categorizations are used: http://www.bls.gov/opub/mlr/2005/07/art6full.pdf
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
5
Effects of Health IT on Regional Health Care
Allow physicians to maintain electronic
records about patients’ conditions. System
can also inform doctors about conditions
they may have overlooked.
Physician Documentation (PD)
9%
35%
We further distinguish basic EMR applications and advanced EMR applications. The US EMR Adoption
Model defines 8 stages for the adoption of IT in the health care sector. These are based on the relative ease
of adoption of the different EMR systems. For example, HIMSS (2011) groups Clinical Data Repository,
Clinical Decision Support System and Order Entry in to the rubric of basic EMR and grouped
Computerized Physician Order Entry and Physician Documentation into the rubric of Advanced EMR.
We use the ratio of EMR applications present at the hospital to the total number of EMR applications (5
applications) as a measure of EMR adoption. This variable changes between 0 and 1, 1 indicating all the
EMR applications are adopted. Table 2 presents the summary statistics for the EMR adoption measures
as well as mean and standard deviations for main dependent, independent and control variables.
Table 2: Descriptive Statistics
Variable
Mean
Std.
Dev
Log Ratio of hospital costs to bed admittance
8.13
0.86
Log Ratio of other hospitals' costs to bed admittance
8.26
0.53
EMR adoption rate (between 0 and 1)
0.38
0.32
Basic EMR adoption rate (between 0 and 1)
0.53
0.42
Advanced EMR adoption rate (between 0 and 1)
0.14
0.30
Gini Coefficient of EMR
0.23
0.23
Gini Coefficient of Basic EMR
0.30
0.29
Gini Coefficient of Advanced EMR
0.14
0.22
Discharge Medicare (in 1000s)
2.35
2.63
Discharge Medicaid (in 1000s)
0.93
1.52
Number of beds
145
1118
Number of employees
819
882
Log (Outpatients)
11.21
1.59
Log of population
12.59
1.62
Log of median income
10.84
0.23
Population density (in 1000s)
1.80
6.14
Percentage population age 65+
14.69
4.50
Percentage population age 24-65
52.70
3.73
Percentage population black
9.14
13.64
Percentage population high school graduate
85.44
7.76
Percentage population college graduate
22.75
11.67
No of top 100 universities
0.38
0.86
Ratio of high-tech establishments
0.02
0.02
HSA level characteristics
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
6
Effects of Health IT on Regional Health Care
3.2 Model Specification
Our main goal is to estimate the degree of spillover effects by which EMR adoption of a hospital affects the
costs of other hospitals’ in the same region. We measure the regional spillover effects at the HSA level, by
calculating the total costs in the HSA except the focal hospital, which is our main dependent variable. To
account for different sizes of hospitals, we standardize health operational costs by dividing the total
operational costs by number of bed admittance days. Health care costs deflated to adjusted for price
inflation. We use yearly EMR adoption as the key independent variable. We investigate how EMR
adoption made by a hospital i, affects its own costs as well as costs of other hospitals from the same HSA
h.
(1) Hospital level effect: ()
,
= + + + + + + +
(2) Spillover level effect:()
,
= + + +
+ + + +
where the dependent variable in equation (1) , ()
, is the deflated operating cost per bed
admittance days of hospital i at time t. The dependent variable in equation (2), ()
, is the
deflated operating cost per bed admittance days of hospitals in health service area h at time t excluding
the focal hospital i. 4 The independent variables are same between the two equations. The main
independent variable of interest, is the level of EMR adoption at hospital i at time t, and
is the EMR adoption level at hospital i at time t-1. In different specifications, we add further
lagged values of EMR adoption levels. Additionally, we control for several hospital and regional
characteristics that might influence the impact of EMR adoption on hospital’s own cost and on the
regional spillovers on other hospitals’ costs. includes hospital level characteristics such as number of
Medicare and Medicaid discharges, number of beds and number of employees. represents HSA level
control demographics such as population, population density, race, age, and education. We further
control for the ratio of high-tech establishments and number of major universities (top 100 universities)
in the area as a measure of resources that are complementary for health IT. Additionally, hospital fixed
effects ( ) control for time invariant heterogeneity across different hospital characteristics. Year fixed
effects ( ) enable us to control for nation-wide shocks to the economy and health care system that are
experienced by all health care providers. is the random error, which captures unobserved random
factor that may have an effect on health care costs. In all specifications standard errors are clustered by
hospital and year.
Results
Regional Spillover Effects of EMR Adoption
Table 3 presents hospital level effects and spillover level effects of EMR adoption on costs levels. Panel A
of Table 3 presents the hospital level effects. We find that EMR adoption is associated with higher costs at
the hospital level in the current year, and do not have impact on costs in the subsequent years, consistent
with other studies that use HIMSS data (Dranove et al. 2012). Interestingly, we find that the coefficient of
EMR adoption is negative and significant in the spillover equation starting from one-year lagged effects.
That is, increase in EMR adoption in one hospital reduces the healthcare costs of other hospitals residing
in the same HSA, indicating strong spillover effects over time. Hence, combing the results in Panel A and
Panel B, there is evidence supporting H1 that, although the cost increases at the hospital that is making
the EMR investment, this can lead to reduction in the costs of providing care for co-located hospitals. The
results indicate that one standard deviation increase in EMR adoption in a hospital, leads to 1 percent
increase in its own costs per admittance in the current year and not longitudinally. Similar standard
This is done with the aim of examining the effect of the adoption of EMR systems on operational cost of
other hospitals in the HSA (that is the spillover effect). The impact of diffusion remains an open question.
4
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
7
Effects of Health IT on Regional Health Care
deviation increase in change in EMR adoption in a hospital is associated with 1.3 percent decrease in the
total costs per admittance of the other hospitals in the region after one year. The lagged effects remain
significant up to four years.
Table 3: Effect of EMR adoption on hospital’s own cost and on other hospitals’ costs
A. Hospital Level Effects: DV: Hospital's Own Cost
VARIABLES
(1)
(2)
(3)
(4)
EMR
0.044***
0.041***
0.036***
0.030***
(0.010)
(0.009)
(0.009)
(0.009)
0.019
0.008
0.008
0.007
(0.012)
(0.009)
(0.008)
(0.008)
0.010
0.007
0.006
(0.012)
(0.011)
(0.011)
-0.007
-0.008
(0.012)
(0.010)
EMR (t-1)
EMR (t-2)
EMR (t-3)
EMR (t-4)
-0.003
(0.015)
Hospital Fixed Effects
Yes
Yes
Yes
Yes
Time Fixed Effects
Yes
Yes
Yes
Yes
Observations
31,071
28,499
25,927
23,355
R-squared
0.453
0.455
0.441
0.417
B. Spillover Level Effects
DV: Other Hospital's Total Cost in the HSA
VARIABLES
(1)
(2)
(3)
(4)
EMR
-0.010
-0.009
-0.009
-0.009
(0.008)
(0.008)
(0.008)
(0.008)
-0.041***
-0.022***
-0.021***
-0.020***
(0.009)
(0.007)
(0.007)
(0.007)
-0.031***
-0.012*
-0.012*
(0.008)
(0.006)
(0.006)
-0.033***
-0.018***
(0.008)
(0.006)
EMR (t-1)
EMR (t-2)
EMR (t-3)
EMR (t-4)
-0.026***
(0.009)
Hospital Fixed Effects
Yes
Yes
Yes
Yes
Time Fixed Effects
Yes
Yes
Yes
Yes
Observations
31,071
28,499
25,927
23,355
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
8
Effects of Health IT on Regional Health Care
Adj. R-squared
0.377
0.381
0.368
0.342
Hospital level control variables included: Number of beds, Medicare discharges, Medicaid discharges,
number of employees, and initial level of EMR adoption.
HSA level control variables included: Population, population density, age, education, race, median
income, high-tech industry ratio, number of major universities.
Standard errors in parentheses. Standard errors are two-way clustered by hospital and time.
*** p<0.01, ** p<0.05, * p<0.1
4.2 Effects of EMR Adoption by Technology Characteristics
We differentiate between basic EMR and advanced EMR applications. Table 4 presents estimated
spillover level effects where EMR adoption is decomposed into basic EMR adoption and advanced EMR
adoption. Results indicate that spillover effects are mainly driven by basic EMRs up to three years.
Spillover effects of advanced EMRs become significant after four years, indicating longer time required to
gain cost benefits from more advanced health IT applications.
Table 4: Effects by Basic and Advanced EMR Adoption on the spillovers
DV: Other Hospital's Total Cost in the HSA
VARIABLES
(1)
(2)
(3)
(4)
Basic EMR
-0.004
-0.004
-0.005
-0.004
(0.007)
(0.006)
(0.006)
(0.006)
-0.031***
-0.019***
-0.018***
-0.018***
(0.007)
(0.006)
(0.005)
(0.006)
-0.020***
-0.006
-0.006
(0.006)
(0.005)
(0.005)
-0.023***
-0.016***
(0.006)
(0.005)
Basic EMR (t-1)
Basic EMR (t-2)
Basic EMR (t-3)
Basic EMR (t-4)
-0.008
(0.007)
Advanced EMR
Advanced EMR (t-1)
-0.007
-0.007
-0.005
-0.006
(0.007)
(0.007)
(0.007)
(0.007)
-0.007
-0.000
0.000
0.001
(0.009)
(0.008)
(0.008)
(0.008)
-0.011
-0.006
-0.006
(0.008)
(0.006)
(0.006)
-0.009
0.002
(0.008)
(0.008)
Advanced EMR (t-2)
Advanced EMR (t-3)
Advanced EMR (t-4)
-0.025***
(0.010)
Hospital Fixed Effects
Yes
Yes
Yes
Yes
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
9
Effects of Health IT on Regional Health Care
Time Fixed Effects
Yes
Yes
Yes
Yes
Observations
31,071
28,499
25,927
23,355
Adj. R-squared
0.378
0.381
0.368
0.342
Similar controls as in Table 3.
*** p<0.01, ** p<0.05, * p<0.1
Discussion and Conclusion
In this study, we analyze the current and long-term regional spillover effects of EMR adoption on health
care costs. We find evidence for positive externalities as EMR adoption leads to a decrease in the costs of
the neighboring hospitals. We find evidence that it takes longer time to realize regional spillovers via
advanced EMRs.
Public Policy Implications
Health care costs remain to be one of the most important policy outcomes and challenges in the US.
Health IT has taken a significant role in US health care policy and it has been subject to a big debate.
HICTECH Act devotes around $19 billion to provide incentives to health care providers for EMR
adoption, and there have been an ongoing debate among policy makers whether the benefits of EMR
investments compensate the costs.
We provide evidence on regional level externalities of EMR adoption that goes beyond the hospital level.
An increase in EMR adoption is associated with higher level of costs for the adopting hospital initially,
however it leads to lower total costs for the other hospitals in the same HSA. Since the geographical level
effects of EMR adoption can differ from hospital level effects, policy makers can provide incentives for
hospitals designed to achieve an optimal outcome for the region. The benefits of EMR subsidies can be
realized more over time as we find stronger lagged effects over time. These findings indicate that policy
makers should account for both the spillover effects and the time effects and consider that expected
benefits could be realized in other co-located hospitals and be observed in the long-term. Similarly, if we
use the $19 billion of financial subsidies to health care providers by the HITECH Act as an investment
measure, then it can potentially decrease health care costs by around $66 billion through regional
spillovers, leading to a net benefit of $47 billion over four years.
References:
Agha, L. 2013. "The Effects of Health Information Technology on Costs and Quality of Medical Care," in Working
Paper.
Athey, S., and Stern, S. 2000. "The Impact of Information Technology on Emergency Health Care Outcomes,"
RAND Journal of Economics (33), pp. 399-432.
Borzekowski, R. 2009. "Measuring the cost impact of hospital information systems: 1987–1994," Journal of Health
Economics (28:5), pp. 938-949.
Bresnahan, T. F., Brynjolfsson, E., and Hitt, L. M. 2002. "Information Technology, Workplace Organization, and
the Demand for Skilled Labor: Firm-Level Evidence," The Quarterly Journal of Economics (117:1) Mar
01, pp. 339-376.
Brynjolfsson, E., and Hitt, L. M. 2003. "Computing Productivity: Firm-Level Evidence," Review of Economics and
Statistics (85:4) Nov, pp. 793-808.
Brynjolfsson, E., Hitt, L. M., and Yang, S. 2002. "Intangible assets: Computers and organizational capital,"
Brookings papers on economic activity (2002:1), pp. 137-198.
Buntin, M. B., Burke, M. F., Hoaglin, M. C., and Blumenthal, D. 2011. "The Benefits Of Health Information
Technology: A Review Of The Recent Literature Shows Predominantly Positive Results," Health Affairs
(30:3), pp. 464-471.
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
10
Effects of Health IT on Regional Health Care
CBO 2008. "Evidence on the Costs and Benefits of Health Information Technology," Congressional Budget Office)
Jun.
Centers for Medicare & Medicaid Services 2010. "National Health Expenditure Projections 2010-2020."
Chang, Y. B., and Gurbaxani, V. 2012. "The impact of IT-related spillovers on long-run productivity: An empirical
analysis," Information Systems Research (23:3-part-2), pp. 868-886.
Chaudhry, B., Wang, J., Wu, S., Maglione, M., Mojica, W., Roth, E., Morton, S. C., and Shekelle, P. G. 2006.
"Systematic review: impact of health information technology on quality, efficiency, and costs of medical
care," Annals of internal medicine (144:10), pp. 742-752.
Cheng, Z., and Nault, B. R. 2007. "Industry level supplier-driven IT spillovers," Management Science (53:8), pp.
1199-1216.
Cheng, Z., and Nault, B. R. 2012. "Relative industry concentration and customer-driven IT spillovers," Information
Systems Research (23:2), pp. 340-355.
Devaraj, S., and Kohli, R. 2003. "Performance impacts of information technology: is actual usage the missing
link?," Management Science (49:3), pp. 273-289.
Dranove, D., Forman, C., Goldfarb, A., and Greenstein, S. 2012. "The Trillion Dollar Conundrum:
Complementarities and Health Information Technology.," American Economic Journal: Economic Policy
(6:4), pp. 239-270.
Hillestad, R., Bigelow, J., Bower, A., Girosi, F., Meili, R., Scoville, R., and Taylor, R. 2005. "Can electronic
medical record systems transform health care? Potential health benefits, savings, and costs," Health Affairs
(24:5), pp. 1103-1117.
HIMSS 2011. "U.S. EMR Adoption Model,") Aug 11, pp. 1-1.
Huang, S. S., Avery, T. R., Song, Y., Elkins, K. R., Nguyen, C. C., Nutter, S. K., Nafday, A. S., Condon, C. J.,
Chang, M. T., and Chrest, D. 2010. "Quantifying interhospital patient sharing as a mechanism for
infectious disease spread," Infection Control and Hospital Epidemiology (31:11), pp. 1160-1169.
Katz, M. L., and Shapiro, C. 1985. "Network Externalities, Competition, and Compatibility," The American
Economic Review (75:3) Feb 01, pp. 424-440.
Lee, B. Y., McGlone, S. M., Song, Y., Avery, T. R., Eubank, S., Chang, C.-C., Bailey, R. R., Wagener, D. K.,
Burke, D. S., and Platt, R. 2011. "Social network analysis of patient sharing among hospitals in Orange
County, California," American Journal of Public Health (101:4), pp. 707-713.
Markus, M. L., Steinfield, C. W., and Wigand, R. T. 2006. "Industry-wide information systems standardization as
collective action: the case of the US residential mortgage industry," MIS Quarterly (30:Special Issue), pp.
439-465.
McCullough, J. S., Parente, S., and Town, R. 2013. "Health Information Technology and Patient Outcomes: The
Role of Organizational and Information Complementarities," NBER Working Paper Series) Feb 21, pp. 147.
Miller, A. R., and Tucker, C. 2009. "Privacy Protection and Technology Diffusion: The Case of Electronic Medical
Records," Management Science (55:7), pp. 1077-1093.
Peffers, K., and Dos Santos, B. L. 1996. "Performance effects of innovative IT applications over time," Engineering
Management, IEEE Transactions on (43:4), pp. 381-392.
Zhu, K., Kraemer, K. L., Gurbaxani, V., and Xu, S. X. 2006. "Migration to open-standard interorganizational
systems: network effects, switching costs, and path dependency," MIS Quarterly (30), pp. 515-539.
Twenty-first Americas Conference on Information Systems, Puerto Rico, 2015
11