Journal of Health Economics 19 2000 767–791
Hospital mergers and acquisitions: does market
consolidation harm patients?
Vivian Ho a,), Barton H. Hamilton b
aSchool of Public Health, UniÕersity of Alabama, Birmingham, Birmingham, AL, USA
bOlin School of Business, Washington UniÕersity in St. Louis, Campus Box 1133,
One Brookings DriÕe, St. Louis, MO 63130-4899, USA
Received 1 July 1998; received in revised form 1 March 2000; accepted 31 March 2000
Debate continues on whether consolidation in health care markets enhances efficiency or
instead facilitates market power, possibly damaging quality. We compare the quality of
hospital care before and after mergers and acquisitions in California between 1992 and
1995. We analyze inpatient mortality for heart attack and stroke patients, 90-day readmis-
sion for heart attack patients, and discharge within 48 h for normal newborn babies. Recent
mergers and acquisitions have not had a measurable impact on inpatient mortality, although
the associated standard errors are large. Readmission rates and early discharge increased in
some cases. The adverse consequences of increased market power on the quality of care
require further substantiation. q2000 Elsevier Science B.V. All rights reserved.
JEL classification: Ill; L41
Keywords: Mergers; Acquisitions; Quality
Competitive pressures have led to mergers among all types of providers in the
health care industry, including hospitals, HMOs, nursing homes and diagnostic
)Corresponding author. Tel.: q1-205-975-0532; fax: q1-205-936-3367.
E-mail address: email@example.com V. Ho .
0167-6296r00r$ - see front matter q2000 Elsevier Science B.V. All rights reserved.
PII: S0167 -6296 0 0 00052- 7
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791768
laboratories Lutz, 1996 . Recently, the number of mergers and acquisitions in the
hospital sector has been particularly high. In 1998, 144 consolidations involved
298 hospitals and approximately 44,000 beds Monroe, 1999 .
The recent flurry of hospital consolidation has generated interest in determining
its impact on prices, costs, and quality of care. While the previous literature has
analyzed the impact of mergers on prices and costs1, no general industrial
organization studies or health care studies explicitly examine the impact of
mergers and acquisitions on product quality. Yet, federal antitrust agencies, as
well as patient and consumer groups, have voiced concern that these mergers have
negative implications for the quality of health care. For instance, in three recently
proposed hospital mergers, the Justice Department and the Federal Trade Commis-
sion have argued that the merger of two hospitals would decrease competition and
therefore reduce quality of care in the local market United States of America,
Plaintiff v. Mercy Health Services and Finley Tri-States Health Group, 1994;
Rather, 1997; Vandewater, 1998 . This paper tests this hypothesis by analyzing the
impact of hospital mergers and acquisitions on inpatient outcomes.
In fact, the structure of pricing in the hospital sector suggests that the strategic
effect of mergers may reveal itself in the form of quality rather than price effects.
A large proportion of patients admitted for hospital treatment are Medicare
patients, for whom hospitals are reimbursed a fixed price, based on their Diagnosis
Related Group DRG regardless of the amount of care provided to patients. In
addition, health insurance companies may use these fixed prices as a guide for
negotiating the prices they will pay to hospitals to cover the care of privately
insured patients. Thus, if prices for treating many patients are fixed, hospitals may
attempt to lower quality in order to maximize profits. The ability to do so may be
facilitated in the case of hospital mergers or acquisitions, which reduce the need to
compete on the basis of product quality.
Even when national hospital systems acquire independent hospitals without
changing local market concentration, concerns arise regarding the consequences
for patient care. Hospital staff have argued that acquisition by a for-profit hospital
system leads to reductions in nursing staff, a shift towards employment of
lower-paid employees, and reductions in expenditures on hospital supplies that
harm the care of patients Lagnado et al., 1997; McGinley, 1998; Burkhart, 1997 .
Such allegations and anecdotes are no substitute for an objective study of
whether hospital mergers and acquisitions harm patient care. This study compares
patient outcomes in hospitals before and after mergers and acquisitions that have
occurred in California between 1992 and 1995. We focus our study on heart attack
and stroke patients, two leading causes of death and disability in the US, as well as
early discharge for normal newborn babies. Our results yield no evidence that
mergers and acquisitions have increased inpatient mortality. These results must be
See, for example, Connor and Feldman 1998 and Dranove 1998 .
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 769
interpreted with caution, because the standard errors for the mortality estimates are
large. Thus, we cannot rule out the possibility that AtrueBdetrimental effects of
consolidation on mortality could readily be three times the estimated coefficients
in our results. Moreover, some hospital consolidation leads to higher readmission
rates for heart attack patients and faster discharge for normal newborns.
The results have important policy implications for the expanding role of market
forces in the US health care industry. Policy makers disagree on the extent to
which government intervention is needed to regulate ongoing consolidation among
hospitals and other health care providers. As stated in a recent review, the key
question facing enforcers of antitrust policy in health care markets is whether these
changes enhance efficiency and quality or instead facilitate collusion and market
power Haas-Wilson and Gaynor, 1997 . This paper contributes to the debate by
analyzing the impact of these consolidations on quality.
Section 2 reviews the existing literature on health care mergers and acquisi-
tions. Section 3 describes the data and provides descriptive statistics. Sections 4
and 5 describe the econometric framework and the results. Section 6 concludes.
2. Existing literature on health care mergers and acquisitions
It is important to distinguish how hospital quality may be affected by hospital
mergers versus hospital acquisitions. Independent hospitals involved in mergers
tend to be in close geographic proximity, and therefore are more likely to raise
concerns regarding increased market power. In contrast, hospitals that are acquired
may be participating in consolidation across markets that have no customer
overlap. For instance, some hospitals in our California database were acquired by
an out-of-state hospital system. Such hospital acquisitions may not lead to
increased local market concentration, but may affect the quality of health care for
reasons we discuss below.
The industrial organization literature concludes that a wide array of outcomes
regarding market share, prices, and costs are possible in cases of Amerger for
oligopolyBBerry and Pakes, 1993; Baker, 1999 . Consolidation potentially
affects market variables for all hospitals in a local market, not just those engaged
in mergers and acquisitions. For instance, Salant et al. Salant et al., 1983 have
pointed out that expansion of output by rival firms can render a merger unprof-
itable. Mergers may also generate cost savings that encourage the merged firm to
lower prices. Thus, theoretical models suggest that merging firms may lead to
higher market concentration; but they may not be able to exercise market power.
The direction and magnitude of the impact of hospital mergers and acquisitions on
economic variables and health care outcomes must be determined empirically.
2This literature does not provide theoretical models of the impact of mergers on product quality.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791770
We first review the empirical literature on hospital mergers. In the hospital
sector, most empirical research has focused not on the potential detrimental effects
of consolidation, but on the potential operating efficiencies resulting from hospital
mergers Connor and Feldman, 1998; Dranove, 1998; Dranove and Shanley,
1995 . Connor and Feldman 1998 report modest cost savings associated with
mergers, with evidence that these costs savings vanish in more concentrated
markets. The result may reflect the finding that scale economies exist only for
small hospitals, so that efficiency gains are unlikely to be realized in the recent
mergers challenged by federal antitrust agencies Dranove, 1998 .
Gaynor and Haas-Wilson 1999 summarize the literature on the relationship
between hospital consolidation and price. They note that Connor et al. 1997
found that merging hospitals exhibit smaller percentage price increases than
non-merging hospitals. This finding supports the hypothesis that the efficiency
gains of hospital mergers outweigh their anti-competitive effects. However, this
result is reversed in concentrated markets; where merging hospitals display higher
percentage price increases. Moreover, Krishnan 1999 finds that merged hospitals
increase prices for those services where the merger raises market share. The
combined conclusions of the above studies of costs and prices suggest that hospital
mergers raise price–cost margins in concentrated markets.
We next consider the empirical literature on hospital acquisitions. Dranove and
Shanley 1995 hypothesize that local multi-hospital systems gain reputation
benefits from standardizing product offerings and quality. The reputation benefits
these systems enjoy over non-system hospitals yield higher price–cost margins.
However, acquisitions do not necessarily imply an increase in local market power.
For instance, some of the hospitals in our sample were acquired by hospital
systems with headquarters outside of California. Nevertheless, those multi-state
systems that did acquire hospitals in California during our study period often
acquired multiple hospitals in the state, perhaps with the intent of achieving local
market power. Even where hospital acquisition does not increase local market
concentration, one might argue that transfer of control from a local hospital board
to a more distant one could alter a hospital’s behavior. Hospitals with strong ties to
the local community may trade off profit maximization in favor of other goals
such as quantity or quality maximization. In contrast, non-locally based systems
particularly for-profit systems acquiring non-profit hospitals may focus more on
profit maximization. Thus, while the literature on the effects of hospital acquisi-
tions is limited, there are reasons to believe that such consolidation may raise
price–cost margins as well.
Although much of the previous literature suggests that market consolidation
leads to higher price–cost margins, none of these studies explicitly examined the
consolidation’s impact on hospital quality. Yet Gaynor and Haas-Wilson note that
it is essential that studies of the economic effects of market power include an
analysis of their impact on quality. The same factors that imply higher price–cost
margins after merger or acquisition may also lead to lower quality after consolida-
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 771
tion. On the other hand, Gaynor and Haas-Wilson note that higher hospital prices
may simply reflect better quality. While this argument runs counter to the
hypothesized effects of increased market concentration, it cannot be ruled out
given the role that non-price competition is believed to play in health care
Dranove et al., 1992; Luft et al., 1986 .
Moreover, the structure of pricing in the hospital sector suggests that the
strategic impact of mergers may occur in the form of quality as well as price
effects. Because a substantial proportion of hospitals’ revenues are based on
Medicare reimbursement rates3, market consolidation may enable providers to
compromise quality, with less concern for its impact on patient demand or price.
Empirical evidence showing that costs decline only slightly or remain constant
after mergers contradicts this reasoning. Yet, federal antitrust agencies and con-
sumers strongly voice their concerns regarding the potential detrimental impact of
hospital consolidation on hospital quality. Thus, an examination of the impact of
hospital mergers and acquisitions on patient outcomes will provide a more
complete picture of market consolidation and social welfare, and contribute
objective information to an antitrust issue which is hotly debated in the public
3. Data and summary statistics
We chose to analyze the California market, where a great deal of hospital
consolidation has occurred, and where patient discharge data is readily available.
The analysis requires data on the characteristics of hospitals and the timing of
hospital mergers or acquisitions if they occurred, as well as information on the
care of hospital patients who were treated in the same time period.
Hospital Data. The American Hospital Association AHA Annual Survey of
Hospitals provides characteristics of California hospitals from 1991 to 1995. Each
year of the survey records hospital mergers that occurred since the previous
survey. These mergers represent two or more corporations coming together into a
single surviving entity, although both physical facilities may continue to treat
patients after the merger.
The AHA survey also lists each individual hospital’s membership in a multi-
hospital system, if applicable.4This information, as well as data in the annual
Hospital Acquisition Report published by Irving Levin Associates was used to
3For instance, approximately 60% of heart attack patients and 75% of stroke patients in this study
are covered by Medicare.
4The AHA Data Base defines a multi-hospital health care system as two or more hospitals owned,
leased, sponsored, or contract managed by a central organization.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791772
identify acquisitions of individual hospitals by health care systems, and acquisi-
tions of systems by other systems.5
Patient Data. Data on patients come from the California Office of Statewide
Health Planning and Development OSHPD discharge data, Version A for the
years 1991 to 1996. The OSHPD database contains standardized information from
hospital discharge abstracts for all patients admitted to California hospitals. All
patients admitted to an acute care hospital in California between January, 1991 and
March, 1996 with a primary diagnosis of acute myocardial infarction heart attack
or stroke, and normal newborn babies are included in the sample.6Approximately
4% of patients in the sample were missing information on length of stay and were
excluded from the sample. In addition, due to size constraints, we analyze a 50%
random sample of newborn babies. This yielded a sample of 256,193 heart attack
patients in 461 hospitals , 268,506 stroke patients in 476 hospitals , and 510,572
newborn babies in 335 hospitals over the course of the analysis.
3.1. Trends in California mergers and acquisitions
Hospital mergers and acquisitions occur in a variety of forms, which we
distinguish in our analysis. Between 1992 and 1995, there were 21 independent
California hospitals involved in mergers recorded by the AHA. Over this same
time period, 54 independent hospitals were acquired by a hospital system. Prior to
this time period, many hospitals already belonged to a hospital system; and
between 1992 and 1995, many of these hospitals were again involved in hospital
transactions. In fact, over this time period, 65 hospitals that already belonged to a
hospital system were then acquired by another system. The following analysis
distinguishes between acquisitions of independent hospitals versus hospitals be-
longing to a system that was then acquired by another system. Because the latter
group already belonged to a system, any potential impact of acquisition on local
market variables may have already occurred. Thus, the impact of acquisition on
quality of care for hospitals already belonging to a system is hypothesized to be
smaller than that for independent hospitals which are acquired.
3.2. Measuring hospital quality
The primary goal of this paper is to determine whether the quality of patient
care declines after a hospital is merged or acquired. The measures of quality in this
5Acquisitions of individual hospitals by health care systems, or acquisition of systems by other
systems for 1992 and 1993 were identified based on changes in health care system ID listed between
the 1991–1992 surveys and the 1992–1993 AHA surveys. For subsequent years, hospital acquisitions
were defined based on hospital transactions listed in The Hospital Acquisition Report published by
Irving Levin Associates, which began tracking transactions in 1994.
6Acute myocardial infarction patients are those with ICD-9 codes 410.0–410.1 and 410.3–410.9 as
a primary diagnosis. Stroke patients have ICD-9 codes 430, 431, 434.00, 434.10, 434.90, or 436 as a
primary diagnosis. Normal newborn babies are discharged with DRG code 391.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 773
study are inpatient mortality, readmission rates, and early discharge of newborns.
Inpatient mortality is an important and objective measure of hospital quality. Cet.
par., an increase in inpatient mortality associated with mergers and acquisitions
implies that consolidation reduces quality and harms patient care. One may
surmise that inpatient mortality is a relatively insensitive quality measure; patient
well being or satisfaction may rise or fall substantially, with no accompanying
change in mortality. However, past research has found that price declines associ-
ated with the introduction of DRG reimbursement another significant change in
economic incentives facing hospitals had a significant detrimental impact on
inpatient mortality Cutler, 1995 . Therefore, it seems reasonable to hypothesize
that mergers and acquisitions may affect inpatient mortality as well.
Use of inpatient mortality as an outcome measure is complicated by the fact
that this variable is censored by live discharge. For example, the probability of
dying within 30 days for heart attack patients may be equal for merged and
unmerged hospitals. However, inpatient mortality rates could appear lower for
merged hospitals even if true rates were similar, because they tend to discharge
their patients earlier than non-merging hospitals do. We will control for this
censoring of inpatient mortality using Cox proportional hazards models to examine
the hazard rate for inhospital death, with live discharge treated as a censored
It would be useful to validate the inpatient mortality results over a longer time
period, such as 90-day mortality. However, the OSHPD datasets do not provide
information on mortality after discharge. Moreover, mortality after discharge is
more difficult to attribute to hospital care, because it is also affected by post-dis-
charge outpatient care, which may or may not depend on which hospital the
patient was initially treated in. Nevertheless, given the relatively short time
horizon over which inpatient mortality is observable, we also examine 90-day
readmission rates for heart attack patients as a measure of health care quality.7In
the medical literature, higher readmission rates are often assumed to indicate lower
quality hospital care.
Finally, if shorter hospital stays decrease the amount of care patients receive,
then an association between mergers and acquisitions and declining length of stay
may also imply a reduction in quality of care. This argument is particularly
compelling for the case of early discharge for newborn babies. Concern that early
discharge endangers the health of mothers and newborns led to the passage of the
7We consider readmissions to any California hospital for which the primary diagnosis is AMI, old
AMI, congestive heart failure, or ischemic heart disease. We exclude admissions occurring within 30
days of discharge because they are more likely to represent the treatment of the AMI itself, rather than
later complications. Similar results were found using 6-month readmission rates and readmission for
any reason. Results are available from the authors upon request. Less clinical research has been
conducted on readmission following stroke. Therefore, we are reluctant to use readmission for stroke
patients as a measure of quality of care.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791774
Mothers’ Health Protection Act of 1996, which mandates coverage of a 48-h
hospital stay. Previous studies provide conflicting evidence on the impact of early
discharge on health outcomes for newborns Edmonson et al., 1997; Gazmararian
et al., 1997; Liu et al., 1997 . However, given the absence of information on
patient outcomes after discharge in our data, we use discharge within 48 h as a
process measure of quality in the period preceding this legislation.8
3.3. DescriptiÕeeÕidence on the impact of mergers and acquisitions on hospital
To provide a first look at the relationship between M&As and outcomes, we
examine descriptive statistics on patients who were treated in hospitals before and
after mergers or acquisitions took place. Table 1 describes outcomes for patients
admitted to California hospitals for two selected years in our data set, 1991 and
1995.9The patients are subdivided according to whether or not they were treated
in a hospital that merged with another hospital, an independent hospital acquired
by a hospital system, or a hospital belonging to a system that was acquired by
another hospital system. For instance, the column subtitled AMergedBpresents
information on patients who were admitted in 1991 to a hospital that subsequently
merged with another hospital between 1992 and 1994; the column then character-
izes patients who were admitted in 1995 to one of these same hospitals. Inpatient
mortality and length of stay are reported for heart attack and stroke patients. We
also report 90-day readmission rates for heart attack patients. Length of stay and
the rate of early discharge are reported for newborn babies.
The descriptive statistics illustrate the reasoning underlying the regression
specification we choose in the forthcoming analysis. First, it seems reasonable to
compare inpatient mortality in hospitals before versus after merger or acquisition
to assess the impact of consolidation on patient outcomes. For example, heart
attack patients treated in hospitals that merged after 1991 had an inpatient
8The OSHPD data does not provide information on time of birth, so that our early discharge
definition is an approximation of discharge within 48 h. The OSHPD data sets length of stay equal to 1
for babies born and discharged on the same day, as well as for babies discharged 1 day after birth.
Therefore, our early discharge measure includes all babies discharged between 0 and 24 h after birth, as
well as some of the babies discharged between 25 and 48 h of birth. A baby born at 0300 h one day
and discharged at 1500 h the next day has a length of stay equal to 1 in the OSHPD data, and therefore
is identified as an early discharge. However, a baby born at 2100 h one day and discharged at 0900 h 2
days hence is not an early discharge, even though her length of stay was actually 36 h as well.
Assuming that the amount of mismeasurement for early discharge is equal across all hospital types,
then we can accurately assess the impact of M&As on changes in early discharge patterns.
9The descriptive statistics only examine information on patients admitted to hospitals in 1991 and
1995, in order to simplify the before-and-after comparison of mergers and acquisitions. We do not use
information on patients from 1996 in the descriptive statistics, because we only have information on
patients admitted through March of this year. The regression analysis will utilize data on patients
admitted in all years.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 775
Patient outcomes, by mergerracquisition status of treating hospital
Consolidation in 1992, 1993, or 1994
All No Consolidation Merged Independent acquired System acquired
Patients admitted inr1991 1995 1991 1995 1991 1995 1991 1995 1991 1995
Babies born in
Heart attack patients
Mean length of stay 7.46 5.79 7.54 5.82 6.85 6.30 7.32 5.56 6.27 5.16
% died 0.096 0.078 0.096 0.077 0.097 0.093 0.10 0.081 0.086 0.079
90-Day readmission 0.094 0.085 0.095 0.085 0.087 0.083 0.084 0.084 0.089 0.083
Sample Size 44046 50805 38553 44165 1293 1305 2668 3523 1532 1812
Mean length of stay 11.00 7.98 11.19 8.17 9.19 6.31 10.60 6.55 7.70 6.42
% died 0.117 0.102 0.115 0.100 0.136 0.107 0.132 0.117 0.107 0.103
Sample size 47719 55566 42194 49272 1495 1453 2663 3026 1367 1815
Mean length of stay 1.84 1.55 1.84 1.55 1.81 1.55 1.81 1.49 1.80 1.49
% discharged within 1 day 0.49 0.65 0.49 0.65 0.52 0.65 0.51 0.67 0.52 0.69
Sample size 122,833 108,459 112,001 97,144 3061 3637 4233 4447 3538 3231
aReadmission rates are calculated based on rehospitalization with a primary diagnosis of acute myocardial infarction, old AMI, congenstive heart failure, or
ischemic heart disease for those patients discharged alive. We exclude admissions occurring within 30 days of discharge.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791776
mortality rate of 9.7% in 1991 prior to merger , versus a mortality rate of 9.3% in
1995. The decreases in mortality in 1995 versus 1991 for all hospitals that merged
or were acquired over the time period would lead one to conclude that consolida-
tion tends to reduce inhospital death.
However, between 1991 and 1995, inpatient mortality rates also declined for
hospitals that did not undergo consolidation from 9.6% to 7.7% . Therefore,
inhospital death rates for hospitals that did not undergo consolidation between
1991 and 1995 can be used as a control group when assessing the impact of
mergers and acquisitions on outcomes. Comparison of outcomes within hospitals
before and after consolidation provides an estimate of the direct effect of mergers
and acquisitions; while comparison relative to hospitals that did not consolidate
accounts for changes in outcomes that would have occurred if consolidation did
not take place. The analysis follows a Adifferences-in-differencesBframework
Heckman et al., 1999 for comparing hospital outcomes, although the unit of
observation in the regressions will be the patient.
This approach is complicated though, by a selection problem; namely, hospitals
that merge or are acquired may differ systematically from those that do not
consolidate, even prior to consolidation. Note that in 1991, the mortality rate for
heart attack patients discharged alive from hospitals that did not consolidate
9.6% is higher than that observed for hospitals involved in system acquisitions
8.6% . Systematic differences also arise among stroke patients. In 1991, patients
in independent hospitals that later merged or were acquired had higher inpatient
mortality rates 13.6% and 13.2% versus hospitals which did not subsequently
merge or undergo acquisition 11.5% . If one does not account for systematic
differences in mortality in merging or acquired hospitals both before and after
consolidation, then one may incorrectly attribute these pre-existing differences to a
merger or acquisition effect. Section 4 will discuss the identification of the effect
of mergers and acquisitions on outcomes controlling for systematic differences in
quality that may exist in consolidating hospitals.
The descriptive statistics do not account for censoring of the patient outcome
data for stroke and heart attack patients. For example, between 1991 and 1995
inpatient mortality declined for all stroke patients. However, the overall decline in
inpatient mortality between 1991 and 1995 does not necessarily represent an
improvement in quality. Descriptive statistics on length of stay provided in Table 1
reveal that overall length of stay for stroke patients fell from 11 to 8 days between
1991 and 1995. Because patients were staying 3 days less on average in 1995,
there was less time to observe inpatient mortality. This data censoring will also be
controlled for in the regression model to follow.
An appendix available from the first author’s website10 provides further
descriptive statistics on patients treated in hospitals in 1991 and 1995. A number
10 http:rrlhcwww.soph.uab.edurhcoprho.htmrjhe appendix.pdf.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 777
of differences exist in patient case mix and insurance status by mergerracquisition
type and year that must be controlled for when assessing the impact of hospital
consolidation on patient outcomes. There is a noticeable increase in the number of
comorbidities reported for heart attack and stroke patients admitted in 1991 versus
1995. This increase may be due to DRG creep; as competition in the health care
industry becomes more aggressive hospitals have an incentive to report more
comorbidities per patient, which will increase their payments under the Medicare
DRG reimbursement system.
The percentage of normal newborns who are white is more than 10 percentage
points higher in independent hospitals acquired by systems than in hospitals that
did not consolidate. In addition, the number of births covered by private insurance
is higher among this group of acquired hospitals than in other facilities. These
differences in case mix by mergerracquisition status will be controlled for in the
following multivariate regression framework.
4. Empirical framework
Let ydenote the quality measure of interest for patient iadmitted to hospital
hin year t, such as whether the patient was readmitted within 90 days after the
initial discharge, or whether a newborn was discharged early. Suppose that yiht
can be written as
iht ht h iht ht iht
where MA is a vector of three dummy variables indicating whether a patient has
been admitted to a hospital after the hospital had: merged; been acquired; or been
part of a system acquired by another hospital system.11 The parameter
sents the effect of merger or acquisition on outcomes, relative to expected
outcomes if the admitting hospital had not consolidated.
The previous literature on hospital consolidation suggests that mergers and
acquisitions lead to higher price–cost margins, although the reasons underlying
increased price–cost margins may differ e.g. market power for local hospital
mergers, reputation benefits, distancing from local community interests, or market
power for hospital acquisitions . Correspondingly, we hypothesize that the effect
of hospital consolidation on outcomes will differ by consolidation type. However,
given the absence of previous research on the relation between consolidation and
quality, we do not hypothesize whether the impact of consolidation will be larger
for mergers versus acquisitions. But the impact of acquisition on quality of care
for hospitals already belonging to a system is hypothesized to be smaller than that
for independent hospitals which are acquired.
11 Less than five hospitals consolidated more than once during the sample period. For these hospitals,
we measured the effects of the second consolidation on quality.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791778
The data in Table 1 suggest that there are fixed differences across hospitals,
such as location, that may be correlated with merger status. For example, hospitals
that merged or were acquired had inpatient mortality rates that differed from
non-consolidating hospitals even prior to the consolidation. To account for these
time invariant differences between hospitals, Eq. 1 includes a vector of hospital-
specific intercepts denoted by
.12 Because the regressions include hospital
indicators, all characteristics that do not change over the sample period will be
captured by these controls. Therefore, we do not include hospital-level variables
for other factors such as non-profitrfor-profit status, which affect outcomes but
which are also unchanging over the sample period.
In general changes in hospital performance after acquisition may result from an
accompanying change in ownership status — from non-profit to for-profit. In the
sample over the period of our study, the overwhelming majority of acquisitions do
not involve a change in ownership status. Non-profits acquire non-profits, and
for-profits acquire for-profits. Thus, although changes in ownership status after
consolidation remains an important policy issue, it is not one we can address with
this sample of hospital patients.
The vector Xconsists of patient characteristics including age, gender, race,
and number of comorbidities to control for the fact that M&A effects may reflect
systematic differences in observed case mix across hospital types. Because of the
potential for DRG creep, we estimate the model with and without the comorbidity
variables. In addition, Xalso includes year dummies to control for time trends
that are evident in Table 1. An indicator variable for patients admitted as transfers
from other acute care hospitals is also included as an explanatory variable in the
regressions for heart attack patients. These patients are often transferred to
facilities with more advanced care such as angioplasty or open heart surgery, and
therefore may differ systematically in health status from other patients. Finally, to
investigate the possibility that changes in outcomes represent a volume effect, the
number of patients treated with the same primary diagnosisrDRG in the hospital
during the year in which the patient was admitted VOL is also included as a
Because hospitals involved in consolidation often had different proportions of
private insurance or Medicare patients, or altered their mix of Medicare and
private insurance patients after consolidation, we also estimate the model including
indicators for primary payment source. To examine whether outcomes after merger
12 For most mergers, both facilities continued to operate after consolidation. Therefore, separate fixed
effects were included in the regressions throughout the sample period for each hospital.
13 The volume variable will capture the association between number of patients treated and outcomes
for any one hospital. However, it will not capture the benefits of increased volume across hospitals
which may result from merger or acquisition. This latter effect will be reflected in the merger and
acquisition dummy variables.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 779
and acquisition differ by payment source, we then estimate Eq. 1 separately for
Medicare or Medi-Cal in the case of newborns and private insurance patients.
Previous analyses of the impact of hospital consolidation on prices and costs
suggest that consolidating hospitals may only be able to exercise market power in
concentrated markets Connor et al., 1997 . We will determine whether our
findings change when we allow the effects of hospital mergers and acquisitions to
vary by local market concentration. We define local hospital markets using the
Health Service Areas HSA formed by Makuc et al. 1991 for the entire
coterminous United States. This method was also used by Connor and Feldman
1998 in their analysis of hospital mergers.
Makuc et al. identified HSAs based on travel patterns between counties by
Medicare beneficiaries for routine hospital care. Cluster analysis was used to
combine counties linked by high border crossing into HSAs. Information on each
patient’s county of residence in our data was used to assign each patient to an
HSA. The patients in our sample resided in 27 different California HSAs. The
lowest number of hospitals in an HSA in a given year was one, and the largest
number of hospitals in an HSA in a given year was 157.
The Herfindahl Hirschman Index HHI , equal to the sum of squared market
shares of all hospitals in a local market, was constructed for each HSA and year.
The lowest HHI for an HSA among heart attack patients in a given year was
0.012. The highest market concentration for an HSA in a given year was 1.0. The
mean HHI for all heart attack patients in the sample was 0.090.
To assess the importance of market concentration for health outcomes, we add
the HHI as an explanatory variable to the specification described above. If mergers
and acquisitions only affect the quality of care by increasing local market
concentration, then the coefficient on the HHI and its corresponding coefficient
will capture part of this effect. However, the market power that any consolidating
hospitals will be able to exercise will depend on the size, cost structure, and
casemix of the facilities involved in merger or acquisition. Therefore, we also add
interaction terms of the merger and acquisition dummy variables and the HHI to
the above regression. These variables capture systematic variations in the ability of
consolidating hospitals to exercise market power at any given level of market
concentration and thus change quality after merger or acquisition. For ease of
interpretation, we examine the results of these analyses only after we have
discussed the results of the simpler specification without market concentration
4.1. Adapting the framework to surÕiÕal data
Eq. 1 is easily estimated by standard limited-dependent variable models when
the outcome is a binary measure such as 90-day readmission, or newborn early
discharge. However, investigating the impact of M&As on inpatient mortality
requires estimation of a survival model for time until death in hospital, with
censoring for live patient discharges. For example, suppose that mergers lead to
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791780
shorter lengths of stay, but have no effect on inhospital mortality conditional upon
length of stay. In this case, a simple logit regression of mortality on merger status
would erroneously lead one to conclude that hospital mergers reduce mortality,
since inhospital deaths are less likely to be observed for patients with shorter
lengths of stay. Consequently, the empirical framework must allow for the
potential non-independence between length of stay and inpatient mortality.
The impact of mergers and acquisitions is therefore estimated using the
following likelihood function:
Ž. Ž. Ž.
where the hazard function
mis the probability that the patient dies in hospital
after mdays in hospital, conditional upon remaining in the hospital for at least m
days. Because the likelihood function yields the probability of death in hospital on
day mgiven survival up to day m, the estimated inpatient mortality hazard rate is
estimated accounting for the fact that length of stay varies across patients.
The hazard function
mis parameterized to depend upon the MA , X
and VOL vectors. We follow a common approach and adopt a proportional
hazards specification, so that
mMA ,X,VOL sexp
iht ht iht ht ht iht ht 0hiht
mrepresents the hospital-specific baseline transition intensity func-
tion. Measured characteristics thus shift the transition intensity above or below its
baseline. While a wide variety of parametric assumptions may be made for the
baseline hazard, misspecification of the functional form may lead to biased
parameter estimates. Consequently, we estimate the model using Cox’s Partial
Maximum Likelihood estimator, which does not require a specification for the
baseline hazard to obtain estimates of
. While we could include hospital-specific
indicators in Eq. 3 , we adopt a more general specification of hospital specific
baseline transition intensities
, which allows not only the intercept, but also
the shape of the baseline hazard to differ by hospital to account for fixed
differences across hospitals Kalbfleisch and Prentice, 1980 .
5. Empirical results
This section describes the estimated effects of mergers and acquisitions on each
measure of hospital quality. Unlike the descriptive statistics presented in Section 3,
data from all the years 1991–1996 are used in this analysis.
5.1. The impact of mergers and acquisitions on inpatient mortality
We first examine the impact of mergers and acquisitions on inpatient mortality
for heart attack and stroke patients. The Cox proportional hazard estimates of Eq.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 781
3 for various specifications of our model are presented in Table 2. The regression
is estimated using data on all patients regardless of discharge status, with survival
treated as a censored variable for patients discharged alive. Positive coefficients
indicate that an increase in the variable implies an increase in the conditional
probability of dying on a given day.
The first four rows of Table 2 reveal that none of the merger or acquisition
variables is precisely estimated. Thus, consolidation has no tangible impact on
inpatient mortality for either heart attack or stroke patients. Although there is no
evidence here that mergers and acquisitions affect patient outcomes, a number of
other explanatory variables behave as hypothesized. Larger patient volume is
associated with a lower conditional probability of inpatient death for heart attack
patients, as is a lower number of patient comorbidities. Previous studies Luft et
al., 1979 attribute the volume effect to learning economies in performing coro-
nary artery bypass grafts and angioplasties. In contrast, hospital volume does not
affect inpatient mortality for stroke patients. This may reflect the fact that stroke is
more often treated through medical management rather than invasive surgery. In
addition, a larger number of comorbidities appears to be associated with a reduced
inpatient mortality hazard rate among stroke patients.14
In order to investigate potential differences in consolidation effects by insur-
ance status, we re-estimated the Cox regressions separately for Medicare and
private insurance patients. Again, we found that mergers and acquisitions do not
affect the conditional probability of inpatient mortality for either insurance type.15
The descriptive statistics discussed in Section 3 suggested a propensity for
increased coding of comorbidities over time, which may have been greater for
hospitals which merged or were acquired. Therefore, we re-estimated the specifi-
cations in Table 2, excluding the number of comorbidities as an explanatory
variable.16 Again, the results are consistent with the hypothesis that mergers and
acquisitions do not affect inpatient mortality.
The apparent lack of a tangible effect of mergers and acquisitions on inpatient
mortality may be due to an insufficient sample size of patients in consolidating
hospitals. Among heart attack patients, 3.1% were in hospitals that merged, 6.6%
were in independent hospitals that were acquired, and 4.5% were in hospital
systems acquired by other systems. These small samples can lead to relatively
large standard errors. For example, for heart attack patients the merger of two
independent hospitals has a standard error of 0.065. To find a significant effect of
The raw data reveals that stroke patients who die have more comorbidities 4.76 than those
discharged alive 4.38 . However, stroke patients who die also have substantially longer lengths of stay
than patients discharged alive 16.8 versus 10.2 days . Thus, the Cox regression estimates reflect the
fact that sicker stroke patients are less likely to die on any giÕen day given that they are still in hospital
— because they are likely to stay much longer than healthier patients, then die in hospital.
15 The results are available from the authors upon request.
16 The results are available from the authors upon request.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791782
Cox proportional hazards estimates for inpatient mortality)
Heart attack patients Stroke patients
Ž. Ž .
Merger 0.022 0.339 y0.010 y0.171
Independent acquired y0.031 y0.712 0.014 0.373
System acquired y0.010 y0.216 y0.020 y0.508
Patient volume y0.0006 y2.625 y0.0002 y1.067
Ž. Ž .
Comorbidities 0.149 47.011 y0.027 y10.297
Transfers y0.364 y14.415 –
Medi-Cal y0.007 y0.217 y0.164 y6.143
Private insurance y0.294 y8.595 y0.159 y5.707
Self-pay 0.141 2.511 0.230 6.151
Indigent y0.493 y4.531 y0.590 y7.460
Other payment y0.153 y1.759 y0.084 y1.146
t-statistics are in parentheses. Ns256,193 heart attack patients, Ns268,506 stroke patients.
Each model also contains year dummies, indicators for ages 0–29, 30–39, 40–44, 45–49, . . . , 85–89,
90q, gender, black, Hispanic, Asian and other races. The excluded category for insurance dummy
variables is Medicare. Each model also includes hospital specific baseline hazards.
these mergers on inpatient mortality, the parameter estimate would be twice the
standard error, or 0.130. Thus, we cannot rule out the possibility that the AtrueB
effect may indeed be positive, but smaller than 0.130. Even for heart attack
patients in independent hospitals acquired by systems, which is the most precisely
estimated M&A effect in Table 3, we cannot rule out true effects up to three times
Linear probability estimates for readmission within 90 days for heart attack patients)
All patients Medicare Private insurance
Ž. Ž. Ž.
Merger 0.017 3.122 0.014 1.813 0.025 2.731
Ž. Ž. Ž .
Independent acquired 0.009 2.280 0.013 2.512 y0.005 y0.674
Ž. Ž. Ž.
System acquired 0.007 1.763 0.006 1.151 0.009 1.305
Ž. Ž. Ž.
Patient volume y0.00001 y0.733 0.000003 0.117 y0.00003 y1.244
Ž. Ž. Ž.
Comorbidities 0.004 12.997 0.004 10.216 0.003 5.389
Ž. Ž. Ž.
Length of stay y0.0007 y6.890 y0.001 y6.855 y0.0005 y2.016
Ž. Ž. Ž.
Transfers y0.002 y0.945 0.002 0.753 y0.003 y1.019
Medi-Cal 0.017 5.859
Private insurance y0.018 y7.339
Self-pay y0.023 y6.027
Indigent y0.012 y2.256
Other payment y0.018 y3.377
t-statistics are in parentheses. Ns234,375 patients; 135,635 Medicare, 68,814 Private Insurance.
Each model also contains year dummies, indicators for ages 0–29, 30–39, 40–44, 45–49, . . . , 85–59,
90q, gender, black, Hispanic, Asian and other races. Readmissions are subsequent admissions to any
California acute care hospital for heart failure, acute myocardial infarction, or ischemic heart disease.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 783
the magnitude of the estimated coefficient of y0.031. Thus, the impact of mergers
and acquisitions on inpatient mortality may be too small to detect with the given
5.2. The impact of mergers and acquisitions on readmissions
Table 3 contains estimates of the determinants of readmission within 90 days
for heart attack patients who were discharged alive. These estimates were derived
from a linear probability model, where the dependant variable equals 1 if the
patient is readmitted within 90 days and 0 otherwise.17 Column 1 indicates that
readmission rises after two types of hospital consolidation. Treatment in a merged
hospital increases the probability of readmission 1.7 percentage points, while
treatment in an independent hospital acquired by a system raises readmission rates
by 0.9 percentage points. Acquisition of a hospital belonging to a system acquired
by another system also increases the probability of readmission, although the
effect is smaller 0.7 percentage points and less precisely estimated ts1.76 .
Table 1 shows that hospitals which later merged and independent hospitals which
were later acquired had readmission rates of 8.7% and 8.4%, respectively, in 1991.
Thus, the estimates in Table 3 suggest that consolidation raises the probability of
readmission for heart attack patients by at least 10% in each of these cases. Thus,
when readmission is used as a quality measure, mergers and most acquisitions do
have a detrimental impact on quality.
The estimated impact of mergers and acquisitions on readmission rates is
derived assuming that trends in hospitals prior to consolidation or trends in
hospitals which did not consolidate serve as an appropriate control group. But
given that the chosen quality measures are changing rapidly for all hospitals over
this period, one may be concerned that omitted variables that are correlated with
the timing of mergers and acquisitions in the data may be driving the results. In
order to check for potential omitted variables, we re-ran the specification for
readmission rates including interaction effects for each year and the 27 HSAs
included in the dataset. Both the magnitude and the precision of the estimates of
the effects of merger or acquisition on readmission rates remained virtually the
same.18 We also re-ran a specification including interactions between year and
for-profit versus non-profit status. In this case, the effects of hospital mergers
17 Because of the large sample size and the inclusion of hospital-specific indicators, the readmission
and early discharge regressions were estimated using a linear probability model. Using a subset of the
observations, no difference was found between linear probability estimates and those derived from a
18 The same interaction effects were added to the inpatient mortality regressions to check for omitted
variables. Again the standard errors for the M&A coefficients were large, so that the effects were
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791784
and acquisition by independent hospitals remains unchanged. However, the impact
of acquisition of a hospital system by another system on readmission rates drops to
Ž. Ž .
0.3% versus 0.7% , and the t-statistic falls to 0.696 versus 1.763 . Therefore,
omitted variables may explain higher readmission rates for hospital systems
acquired by other systems.
When consolidation effects are estimated separately by insurance status, the
detrimental impact of mergers on readmission is tangible for both Medicare and
private insurance patients. However, only readmission rates for independent
hospitals acquired by a system appear higher for Medicare patients.19 This finding
is consistent with the hypothesis that quality may more likely be compromised for
Medicare patients whose care is covered by fixed price reimbursement.
5.3. The impact of mergers and acquisitions on early discharge for newborns
Table 4 contains regression estimates of Eq. 1 for the probability of discharge
within 48 h for normal newborns. Column 1 indicates that being born in a hospital
system acquired by another hospital system increases the probability of early
discharge by 3.2 percentage points. The rate of early discharge for newborns was
52% in these hospitals prior to acquisition in 1991. Therefore, the estimated
impact associated with consolidation in this case represents a 6% increase in the
probability of early discharge. We had hypothesized that the impact of consolida-
tion on the quality of care would be smaller for hospitals already belonging to a
system. However, the estimates for newborns do not support this hypothesis.
To check for potential omitted variables, we re-estimated the regression first
including interactions between year and HSA, then including interactions between
year and for-profit status. In both cases, the impact of hospital systems being
acquired on early discharge rises slightly to 3.7% and 3.8%, respectively , and the
coefficient remains precisely estimated ts7.776 and 7.466, respectively . The
coefficients on the other merger and acquisition variables remain insignificant.
Therefore, the measurable changes in early discharge are robust.
Recall that the majority of babies delivered are covered by either private
insurance or Medi-Cal. The estimates in Column 1 indicate that private insurance
increases the probability of discharge within 48 h relative to Medi-Cal coverage.
Columns 2 and 3 contain separate regression estimates for Medi-Cal and private
insurance babies. In this case only babies covered by Medi-Cal who were born in
hospitals which consolidated experienced an early discharge effect. Hospital
mergers increased the probability of early discharge by 2.2 percentage points;
while acquisition of a hospital system by another system raised the probability of
19 Further tests showed that the impact of mergers on readmissions was not significantly different for
Medicare versus private insurance patients. Moreover, readmission rates for privately insured patients
in independent hospitals acquired by a system were significantly lower than for Medicare patients.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 785
Linear probability estimates for discharge within 48 hours for newborns)
All patients Medi-Cal Private insurance
Ž. Ž. Ž.
Merger y0.001 y0.178 0.022 2.232 y0.005 y0.758
Ž. Ž. Ž.
Independent acquired y0.003 y0.813 y0.010 y1.572 y0.004 y0.905
Ž. Ž. Ž.
System acquired 0.032 10.030 0.071 15.090 0.0001 0.020
Ž. Ž. Ž.
Ceasarean section y0.736 y558.846 y0.677 y318.162 y0.782 y454.297
Ž. Ž. Ž.
Comorbidities y0.075 y66.301 y0.074 y42.597 y0.073 y46.905
Private insurance 0.041 29.461
Self-pay 0.098 38.655
Indigent 0.053 5.207
Other payment 0.046 10.411
t-statistics are in parentheses. Ns510,572 babies; 228,779 Medi-Cal, 249,045 Private Insurance.
Each model also contains year dummies, and indicators for black, Hispanic, Asian and other races.
early discharge by 7.1 percentage points for Medi-Cal patients. Re-estimating the
equations in Table 4 with the number of comorbidities excluded as a regressor did
not change any of these conclusions.
5.4. The impact of mergers and acquisitions in concentrated markets
In the preceding analysis, we found several instances where we could not detect
an effect of hospital mergers and acquisitions on hospital quality particularly for
inpatient mortality . Yet, detrimental effects may only be evident in the subgroup
of hospital consolidations that occurred in concentrated markets. We therefore
re-examine the impact of hospital mergers and acquisitions on patient outcomes
including measures of local market concentration.
We report results of the regressions for the determinants of AMI readmissions
and early discharge of newborns modified to account for market concentration in
Table 5. The table contains coefficient estimates for the merger and acquisition
dummy variables, the HHI, and the interaction of the hospital consolidation
dummies with the HHI. The remaining explanatory variables are the same as the
ones used previously.20 Test statistics for the joint significance of each indicator
20 Because the standard errors for the consolidation effects in the inpatient mortality regressions were
relatively large, it was unlikely that the inclusion of market concentration effects would modify these
results. Therefore, these results are not reported in Table 5. However, we did add market concentration
effects to the inpatient regressions, and we still found that hospital mergers and acquisitions had no
measurable effect on the mortality hazard rates. Parameter estimates for the entire regressions are
available from the authors upon request.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791786
Estimates including market concentration measures)
90-Day readmission Early discharge of
heart attack patients newborns
Ž. Ž .
Merger 0.020 2.840 y0.022 y3.176
Ž. Ž .
Independent acquired 0.008 1.723 y0.004 y0.735
System acquired 0.006 1.262 0.028 7.291
Merger=HHI y0.046 y0.612 0.319 3.218
Independent acquired=HHI 0.004 0.104 0.014 0.365
System acquired=HHI 0.006 0.181 0.086 2.000
HHI y0.059 y0.741 0.324 7.721
Joint significance tests F-statistic P-value F-statistic P-value
Merger 3.62 0.013 22.76 P-0.001
Independent acquired 1.91 0.126 20.11 P-0.001
System acquired 1.22 0.302 55.19 P-0.001
t-statistics are in parentheses. Sample sizes and additional variables are the same as those reported
in Tables 2–4. Joint significance tests provide test statistics for the null hypothesis that the coefficient
on each individual merger or acquisition dummy variable, the HHI, and the interaction between these
two variables are jointly equal to 0.
variable for merger or acquisition, the HHI, and their interaction terms are
reported at the bottom of the table.
We first examine the impact of market concentration on conclusions regarding
90-day readmission for heart attack patients. The analysis in Table 3 revealed that
hospital mergers, acquisitions of independent hospitals, and acquisition of system
hospitals all increased the probability of 90-day readmission. However, in Table 5,
only the coefficient on the hospital merger dummy variable remains precisely
estimated. Inclusion of market concentration measures reduces the precision of the
estimates relating consolidation to poor outcomes based on readmission data.
We also examine the link between market concentration and the impact of
hospital consolidation on early discharge of newborns. In Table 4, we found that
acquisition of hospital systems by other systems increased the probability of early
discharge. In Table 5 we found that increased market concentration as measured
by the HHI increases the probability of early discharge. Moreover, the coefficients
on the HHI, merger, system acquisition, and their interaction terms are jointly
significantly different from 0.
For any given newborn, the probability of early discharge associated with
hospital consolidation is computed by summing the coefficient on the appropriate
hospital consolidation dummy and the coefficient on its interaction term multiplied
by the HHI for the admitting hospital. The marginal impact of increased market
concentration on the probability of early discharge after hospital consolidation
appears to be substantial. The entire lowest quartile of deliveries in the sample
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 787
occurred in markets with an HHI of approximately 0.015.21 The estimates in Table
5 indicate that a merger of independent hospitals in a competitive market such as
this reduces the probability of early discharge by 1.7 percentage points. However,
an independent hospital in a market with an HHI of 0.1 the approximate sample
mean increases the probability of early discharge by 1 percentage point. Likewise,
acquisition of a system hospital in a market with an HHI of 0.1 raises the
probability of early discharge 3.7 percentage points. At the 90th percentile of the
HHI distribution 0.18 , the probability of early discharge rises by 3.5 percentage
points for newborns in merged hospitals; and by 4.3 percentage points for
newborns in acquired system hospitals.22 Note that the estimated effects for these
two acquisition types are larger than comparable estimates in Table 4, which did
not control for market concentration. Thus, for the case of early discharge of
newborns, hospitals in markets with higher concentration do appear to exercise
market power by lowering quality after consolidation.
5.5. Summary of estimation results
In summary, the regression results indicate that mergers and acquisitions have
no measurable impact on inpatient mortality, although the standard errors associ-
ated with these estimates are large. However, mergers and acquisition of an
independent hospital increase readmission rates for heart attack patients. In
addition, some hospital acquisitions lead to early discharge of normal newborns.
These effects for newborns are particularly notable for consolidating hospitals that
operate in highly concentrated markets. We had originally hypothesized that the
impact of consolidation might be larger for Medicare patients versus other
patients, and for independent hospitals versus hospitals already belonging to a
system. Although we found evidence consistent with both hypotheses in some
cases, neither hypothesis held for the majority of quality measures.
This paper provides a first look at the impact of hospital mergers and
acquisitions on patient outcomes and the quality of inpatient care for patients
admitted to California hospitals between 1991 and 1996, a location and period that
has seen a substantial amount of M&A activity. We find no evidence that mergers
and acquisitions measurably affect inpatient mortality. However, some types of
21 Most of these patients are in Los Angeles county.
22 These figures may in fact be a lower bound of the effects of market consolidation on readmissions,
because cet. par., the HHI also rises after merger or acquisition. However, this marginal effect appears
to be small. For the sample of newborns in cases where consolidation occurred, markets with hospital
mergers experienced the greatest increase in HHI between 1991 and 1995; but the increase in the HHI
was only 0.02.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791788
hospital consolidation are associated with increased readmission rates for heart
attack patients, and increased likelihood of early discharge for normal newborns.
The effects of mergers and acquisitions on quality identified in this study differ
by type of consolidation. For example, acquisition of independent hospitals raises
readmission rates for heart attack patients, but does not lead to earlier discharge of
newborn babies. In contrast, acquisition of hospital systems by another system
leads to earlier discharge of newborn babies, but does not raise readmission rates
for heart attack patients. Thus, increases in consolidation may not always compro-
mise the quality of patient care. The extent to which different types of hospital
consolidations change the quality of care is likely to depend on both demand and
cost factors, which can vary by merger and acquisition type and patient popula-
tion. Further analysis with data on revenues by patient mix, managed care
penetration, and hospital costs will help to explain the findings in this paper.
Some important caveats to our analysis remain. The descriptive statistics
indicated that patient outcomes differed between consolidating and non-consolidat-
ing hospitals and across different types of mergers and acquisitions even in 1991,
prior to when consolidation occurred. We controlled for these initial differences in
outcomes in our analysis by using hospital-specific strata in the Cox regressions
and fixed effects in the linear probability models. However, the results in this
paper reflect the change in patient outcomes after consolidation, conditional on a
hospital’s decision to merge or be acquired. Determining the factors that lead a
hospital to merge or be acquired remains an interesting issue, which is worthy of
We were not able to detect a detrimental impact of hospital mergers and
acquisitions on inpatient mortality. These results are valid even if average lengths
of stay in hospital have fallen over time, so that patients discharged alive are on
average more unstable when they leave the hospital. The econometric framework
estimates the hazard of dying in hospital on a given day, given survival in hospital
up until that day, so that shorter lengths of stay for live discharges are accounted
for when estimating the impact of hospital consolidation on inpatient mortality.
These results should be interpreted with caution, given that relatively small
proportions approximately 5% of heart attack and stroke patients were treated in
each type of merged or acquired hospital. The resulting standard errors are large
enough that we cannot rule out the possibility that AtrueBdetrimental effects of
consolidation on mortality could readily be three times the estimated coefficients
in our results. Analyses with data from multiple states are necessary to investigate
this issue further.
Although the regressions included controls for patient health status, unobserv-
able differences in patient health status may have biased the results. We tested for
differences in age, number of comorbidities, and race white versus non-white
after consolidation for each of the three patient groups in this study. Only some
patient characteristics were significantly different before versus after consolida-
tion, and the direction of the difference varied by merger or acquisition type and
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 789
patient group. Therefore, it is unlikely that unobservable differences in patient
casemix systematically biased the findings up or down.
Another weakness of this analysis is that it cannot identify hospitals that may
strategically discharge patients just before the end of life. Identification of such
behavior requires information on changes in patients’ health status throughout
their stay in hospital. Such detailed information is usually only available through
detailed chart abstraction for a smaller sample of patients and hospitals than was
used in this study.
A preliminary means of investigating this issue is to note that strategic behavior
by hospitals to discharge patients near death would suggest that sicker patients
would have a lower hazard of death in hospital. However, our estimates indicate
that a higher number of comorbidities increases the hazard of death in hospital.
We also interacted the merger and acquisition dummy variables with the number
of comorbidities to determine whether hazards by health status differed after
consolidation; but these interaction terms were all insignificant.23 While this
analysis is coarse, the results suggest that finding no evidence that hospital
consolidation harms inpatient mortality is not due to strategic decisions by
hospitals to discharge patients just before death.24 Future analyses that are able to
provide combined data on inpatient mortality, readmission rates, and post-dis-
charge mortality will be more informative on this issue.
We have chosen to analyze only the quality implications of hospital mergers
and acquisitions. Analysis of both price and quality are necessary to determine
how market consolidation affects social welfare. However, the California OSHPD
discharge data base is a poor source for examining price–cost margins, because
only information on patient charges as opposed to actual prices paid is listed.
Nevertheless, only a handful of papers exist that analyze the price effects of
hospital consolidation; and none of these studies examines patient outcomes.
Therefore, this study yields useful initial findings on the implications of consolida-
tion for product quality. Gaynor and Haas-Wilson have suggested that higher
hospital prices after consolidation that have been noted in prior studies may reflect
better quality. In contrast, we find no evidence that hospital consolidation lowers
mortality rates, readmission rates, or the probability of early discharge for
newborns. Thus, the rents which consolidating hospitals have been observed to
achieve through higher prices may not be dissipated by increases in quality. Future
studies that can simultaneously evaluate both price and quality changes associated
with hospital mergers and acquisitions are needed to confirm these findings.
23 The results are available from the authors upon request.
24 Likewise, this analysis addresses concerns that the hazard of being discharged alive from hospital
is negatively related to a patient’s mortality hazard. Hospitals are less likely to discharge patients at
high risk of death. However, the insignificant coefficients on the interaction terms between the number
of comorbidities and the consolidation dummies suggest that such behavior does not change after
merger or acquisition.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791790
Previous studies suggest that consolidation may affect price, cost, or quality for
all hospitals in a local market, not just those involved in merger or acquisition.
Estimating a more general model to account for this possibility is beyond the
scope of this paper. We chose to focus our analysis on the quality effects of those
hospitals directly involved in consolidation, because these are the facilities that are
currently of greatest concern to policy makers. Nevertheless, the range of quality
measures analyzed in this paper is fairly narrow. We examined two process
measures readmission rates and early discharge and one outcome measure
inpatient mortality of hospital quality. Further analyses of a broader array of
quality measures and a larger set of diagnoses are necessary to substantiate our
Debate continues on whether consolidation in health care markets enhances
efficiency and quality or instead facilitates collusion and market power. The
results in this paper are consistent with the hypothesis that recent mergers and
acquisitions have not had a unilateral detrimental impact on the quality of patient
care. Concerns regarding the adverse consequences of increased market power on
quality should not be ignored; but further investigation with detailed clinical data
is necessary to substantiate these concerns.
We are grateful to Richard Boylan, Bill Encinosa, David Eisenstadt, Dana
Goldman, Martin Gaynor, Robert Helms, and seminar participants at the 1998
Econometric Society Meetings, the American Enterprise Institute 1998 Health
Policy Discussion Series, the University of Minnesota Institute for Health Services
Research, and RAND for their comments on this paper. David Cutler, Joseph
Newhouse, and two anonymous referees also provided many helpful suggestions.
Vivian Ho acknowledges the support of the John M. Olin Foundation Faculty
Baker, J.B., 1999. Developments in antitrust economics. Journal of Economic Perspectives 13,
Berry, S., Pakes, A., 1993. Some applications and limitations of recent advances in empirical industrial
organization: merger analysis. American Economic Review 83, 247–252.
Burkhart, F., 1997. Hospital change is opposed. The New York Times, July 3, Metropolitan Desk.
Connor, R.A., Feldman, R.D., 1998. The effects of horizontal hospital mergers on nonmerging
hospitals. In: Morrisey, M.A. Ed. , Managed Care and Changing Health Care Markets. American
Enterprise Institute, Washington, DC, pp. 164–191.
Connor, R.A., Feldman, R.D., Dowd, B.E., Radcliff, T.A., 1997. Which types of hospital mergers save
consumers money? Health Affairs 16, 62–74.
V. Ho, B.H. HamiltonrJournal of Health Economics 19 2000 767–791 791
Cutler, D., 1995. The incidence of adverse medical outcomes under prospective payment. Economet-
rica 63, 29–50.
Dranove, D., 1998. Economies of scale in non-revenue producing cost centers: implications for hospital
mergers. Journal of Health Economics 17, 69–83.
Dranove, D., Shanley, M., 1995. Cost reductions or reputation enhancement as motives for mergers:
the logic of multihospital systems. Strategic Management Journal 16, 55–74.
Dranove, D., Shanley, M., Simon, C., 1992. Is hospital competition wasteful? RAND Journal of
Economics 23, 247–262.
Edmonson, M.B., Stoddard, J.J., Owens, L.M., 1997. Hospital admission with feeding-related problems
after early postpartum discharge of normal newborns. The Journal of the American Medical
Association 278, 299–303.
Gaynor, M., Haas-Wilson, D., 1999. Change, consolidation, and competition in health care markets.
Journal of Economic Perspectives 13, 141–164.
Gazmararian, J.A., Koplan, J.P., Cogswell, M.E., Bailey, C.M., Davis, N.A., Cutler, C.M., 1997.
Maternity experiences in a managed care organization. Health Affairs 16, 198–208.
Haas-Wilson, D., Gaynor, M., 1997. Introduction to the special issue on competition and antitrust
policy in health care markets. Health Economics Letters 1, 3–9.
Heckman, J., Lalonde, R., Smith, J., 1999. The economics and econometrics of active labor market
programs. In: Ashenfelter, O., Card, D. Eds. , Handbook of Labor Economics vol. 3 Elsevier,
Kalbfleisch, J.D., Prentice, R.L., 1980. The Statistical Analysis of Failure Time Data. Wiley, New
Krishnan, R., 1999. Market Restructuring and Pricing in the Hospital Industry. Katz Graduate School
of Business, University of Pittsburgh.
Lagnado, L., Sharpe, A., Jaffe, G., 1997. How ColumbiarHCA changed health care, for better or
worse. The Wall Street Journal, A1–A4, August 1.
Liu, L.L., Clemens, C.J., Shay, D.K., Davis, R.L., Novack, A.H., 1997. The safety of newborn early
discharge: the Washington state experience. The Journal of the American Medical Association 278,
Luft, H.S., Bunker, J., Enthoven, A., 1979. Should operations be regionalized? An empirical study of
the relation between surgical volume and mortality. The New England Journal of Medicine 301,
Luft, H.S., Robinson, J.C., Garnick, D.W., Maerki, S.C., McPhee, S.J., 1986. The role of specialized
clinical services in competition among hospitals. Inquiry 23, 83–94.
Lutz, S., 1996. Merger, acquisition activity hits record in 1st quarter, report says. Modern Healthcare,
2–3, May 27.
Makuc, D.M., Haglund, B., Ingram, D.D., Kleinman, J.C., Anonymous, 1991. Vital and Health
Statistics: Health Service Areas for the United States, DHHS Publication No. PHS 92-1386;
National Center for Health Statistics; Series 2, No. 112. Centers for Disease Control.
McGinley, L., 1998. These doctors, sick of Columbia, want their hospital back. The Wall Street
Journal, A1–A8, March 13.
Monroe, S.M., 1999. The Hospital Acquisition Report. Irving Levin Associates.
Rather, J., 1997. Facing intense pressures, hospitals step up mergers. The New York Times, Late
Edition-Final ed, November 2, 14LI.1-1.
Salant, S., Switzer, S., Reynolds, R., 1983. Losses due to merger: the effect of an exogenous change in
industry structure on Cournot-Nash equilibrium. Quarterly Journal of Economics 98, 185–199.
United States of America, Plaintiff, v. Mercy Health Services and Finley Tri-States Health Group,
1994. US District Court Northern District of Iowa Eastern Division.United States of America.
Vandewater, J., 1998. Tenet merger in Poplar Bluff opposed by FTC. St. Louis Post-Dispatch,