The Marginal Benefit of Inpatient Hospital Treatment: Evidence from

The Marginal Benefit of Inpatient Hospital Treatment:
Evidence from Hospital Entries and Exits
Nathan Petek
Chicago Booth
January 12, 2016
Online Appendix
Online appendix figure 1 shows the relationship between Medicare inpatient days at a hospital and
the entry or exit of that hospital as coded in the AHA data.
Online appendix figure 2 shows the relationship between leads and lags of hospital entry/exit
and log population, controlling for state-year fixed effects.
Online appendix figure 3 is a robustness check of the main Medicare specification in levels.
Online appendix figure 4 is a robustness check of the main Medicare specification that uses
hospital entries/exits in the beneficiary’s hospital service area.
Online appendix figures 5 and 6 show the effect of hospital entry/exit on measures of the
quantity of outpatient care using the Medicare data in first differences and levels, respectively.
Online appendix figure 7 is a robustness check of the main HSA specification that adds stateyear fixed effects.
Online appendix figure 8 shows the effect of a hospital entry and exit on the quantity of care at
all hospitals (not just short-term, general hospitals) in the HSA.
Online appendix figure 9 shows the effect of a hospital entry and exit on the five closest hospitals.
Online appendix figure 10 shows the effect of entry and exit on nearby hospitals with different
assumptions about missing data.
Online appendix figure 11 replicates the main quantity results using state inpatient discharge
data with a model that does not include market time trends.
Online appendix figures 12 and 13 plot separately the effect of entry and exit on my main
outcomes using AHA, vital statistics, and Medicare data.
1
Online appendix figures 14 and 15 are robustness checks of the main HSA specification in
levels and with logged outcomes in levels.
Online appendix figure 16 shows the effect of entry and exit on admissions at competing hospitals by the distance to the entering/exiting hospital.
Online appendix figure 17 shows the effect of hospital entry/exit on mortality in levels both
weighted and unweighted.
Online appendix figure 18 shows the effect of hospital entry/exit on mortality by county in the
1982-2010 period overall and for the 0-64 age group, and the over 65 age group.
Online appendix figure 19 shows the effect of entry/exit in rural counties both weighted and
unweighted and by age.
Online appendix figure 20 shows the effect of entry/exit in rural HSAs weighted and unweighted and by age.
Online appendix figure 21 shows the effect of hospital entry/exit on self reported health by
county in the 2002-2010 period in levels.
Online appendix figures 22 and 23 show the effect of a hospital entry and exit on mortality in
the Medicare data in differences and levels.
Online appendix figure 24 separately shows effect of a hospital entry and the effect of hospital
exit on mortality in the Medicare data in levels.
Online appendix figure 25 shows the effect of hospital entry/exit on mortality by county in the
1999-2010 using the unrecoded entries/exits and the recoded entries/exits.
Online appendix figure 26 shows the effect of entry and exit on total Medicare expenditures by
distance to the affected hospital.
Online appendix figure 27 shows the effect of entry and exit on measures of the quantity of care
received by Medicare patients and the probability of dying for entries/exit that affect the distance
to the closest hospital and by the change in distance to the closest hospital.
Online appendix figure 28 shows the effect of entry and exit on measures of the quantity of
care received by Medicare patients by terciles of changes in distance to the closest hospital.
Online appendix figures 29 and 30 shows the effect hospital entry and exit on discharges by
diagnosis where the diagnoses are ordered by a measure of deferability.
Online appendix figures 31-34 are dot plots that show the effect on discharges by diagnosis of
hospital entry and exit of the closest hospital to the beneficiary, of hospitals in the beneficiary’s
county, of hospitals beneficiary’s hospital service area, and of hospitals within 10 miles of the
beneficiary.
Online appendix figure 35 includes estimates of the effect of hospital entry and exit (individually) on births, births per bed, births in rural areas, and births in logs.
Online appendix figure 36 shows estimates of the effect of hospital entry and exit on births at
2
neighboring hospitals.
Online appendix figures 37 and 38 shows estimates of the effect of hospital entry and exit on
births for varying distances around the entering/exiting hospital including the births at the entering/exiting hospital in changes and logs.
Online appendix figures 39 and 40 show the effect hospital entry and exit on measures of the
average severity of diagnosis of admitted patients in differences and levels.
Online appendix figure 41 shows the effect of entry and exit on the probability of dying for
Medicare patients by terciles of changes in distance to the closest hospital.
Online appendix figures 42 and 43 show lead-lag plots the effect of hospital entry and exit on
mortality by cause of death for entries/exits in the beneficiary’s closest hospital.
Online appendix figures 44-47 are dot plots that show the effect on mortality by cause of death
of hospital entry and exit in the beneficiary’s HSA, in the beneficiary’s county, in beneficiary’s
hospital service area, and of hospitals within 10 miles of the beneficiary.
Online appendix table 1 is a robustness check of the main HSA specification where the sample
is limited to rural HSAs.
Online appendix table 2 is a robustness check of the main HSA specification where the independent variable of interest is the change number of beds at entering/exiting hospital in the year of
entry or year before exit.
Online appendix table 3 is a robustness check of the main HSA specification with the outcome
in logs.
Online appendix table 4 is a robustness check of the main HSA specification where the independent variable of interest is the change number of beds at entering/exiting hospital in the year of
entry or year before exit and the sample is limited to rural HSAs.
Online appendix table 5 is a robustness check of the main HSA specification in where the effect
of entry and exit are estimated separately but coded so they are expected to have the same sign.
Online appendix table 6 is a robustness check of the main HSA specification in levels.
Online appendix table 7 shows the effect of hospital entry/exit on the overall mortality rate,
mortality rate by age, and for AMI in counties for the full 1982-2010 period.
Online appendix table 8 shows the effect of hospital entry/exit in a HSA on the average distance
traveled to the hospital in the Medicare data.
Online appendix table 9 shows the effect of hospital entry/exit in a hospital service area on the
measures of the quantity of care in the Medicare data.
Online appendix table 10 is a robustness check of the main HSA specification where the effect
of entry and exit is estimated separately by hospital size where the quartiles are ordered from
smallest to largest by number of beds.
Online appendix table 11 is a robustness check of the main HSA specification that splits the
3
estimates by entry versus exit and by markets with above and below median capacity utilization.
Online appendix table 12 is a robustness check of the main HSA specification where the effect
of entry and exit is estimated separately by number of beds per capita in the market where the
quartiles are ordered from fewest to most by beds per capita.
Online appendix table 13 is a robustness check of the main HSA specification where the effect
of entry and exit is estimated separately by quartiles of the the market share of the entering or
exiting hospital, where the quartiles are ordered from lowest to highest market shares.
4
10000
−5000
−5000
All Hospital Exits on Medicare Days at that Hospital
Medicare Inpatinet Days
0
5000
All Hospital Entries on Medicare Days at that Hospital
Medicare Inpatinet Days
0
5000
10000
Online Appendix Figure 1: Audit of the Timing of Hospital Entry and Exit
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Medicare Inpatinet Days
0
5000
Recoded Hospital Exits on Medicare Days at that Hospital
−5000
−5000
Medicare Inpatinet Days
0
5000
10000
Recoded Hospital Entries on Medicare Days at that Hospital
10000
−5000
Medicare Inpatinet Days
0
5000
10000
Only Partial Year Entries on Medicare Days at that Hospital
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. Unit of analysis is a hospital-year. Plots are of coefficients on leads
and lags of indicators for hospital entry or exit. Coefficients are from regressions of the level of the Medicare
inpatient days at a hospital on leads and lags of indicators for hospital entry or exit of that hospital, hospital
fixed effects, and year fixed effects. Missing values of inpatient days are coded to zero if the hospital ever
had non-zero values of inpatient days. All entries and exits are show in the top row. The middle row uses
only entries where AHA data indicates the hospital was in the market for a partial year. The bottom row
shows entries and exits recoded so the timing of the event exactly matches when the hospital started and
stopped seeing Medicare patients. Error bars show the 95-percent confidence interval. Standard errors are
clustered by state.
5
Online Appendix Figure 2: Hospital Entry and Exit and Log Population Growth
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
0
.001
.002
Hospital Entry/Exit and Population Growth
by HSA
−8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
Year Relative to Entry/Exit
4
5
6
7
8
7
8
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Hospital Entry/Exit and Population Growth
by County
−8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
Year Relative to Entry/Exit
4
5
6
Notes: Data cover the 1982-2010 period. Unit of analysis is the HSA-year or county-year. Plots are of
coefficients on leads and lags of the change in the number of hospitals from entry and exit. Regressions are
of changes in the log population on leads and lags of the change in the number of hospitals from entry and
exit and state-year fixed effects. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA in the HSA level analysis and state in the county-level analysis.
6
Online Appendix Figure 3: Effect of Hospital Entry and Exit on Quantity of Care and Distance for
Medicare Patients - Outcomes in Levels
Effect of Hospital Entry/Exit on Acute Inpatient Expenditures
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Effect of Hospital Entry/Exit on Total Expenditures
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Entries/Exits in Beneficiary’s HSA
−3
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−1
−.5
0
.5
Coefficients on Leads and Lags of Hospital Entry/Exit
−.04
−.02
0
.02
.04
−1
0
Year Relative to Entry/Exit
1
2
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Entry/Exit on Miles to Hospital
Entries/Exits in Beneficiary’s HSA
−2
1
Entries/Exits in Beneficiary’s HSA
Effect of Hospital Entry/Exit on ALOS
−3
−1
0
Year Relative to Entry/Exit
Effect of Hospital Entry/Exit on Acute Inpatient Days
Coefficients on Leads and Lags of Hospital Entry/Exit
−.01
0
.01
.02
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Effect of Hospital Entry/Exit on Acute Inpatient Stays
−2
2
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in
the patient’s HSA of residence, normalized so the coefficient on the one-year lead is zero. The coefficients
are from a regression of the level of the outcome on leads and lags of the change in the number of hospitals
from entry and exit and controls including gender time trends, cohort time trends, race time trends, age fixed
effects, year fixed effects, and fixed effects for each cohort-race-sex-zip code-county bin. The regressions
are weighted by the number of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary
units. Error bars show the 95-percent confidence interval. Standard errors are clustered by HSA.
7
Online Appendix Figure 4: Effect of Hospital Entry and Exit on Quantity of Care and Mortality
for Medicare Patients - Net Entry in Hospital Service Areas
Effect of Hospital Entry/Exit on Acute Inpatient Expenditures
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Effect of Hospital Entry/Exit on Total Expenditures
Entries/Exits in Beneficiary’s Hospital Service Area
−2
−1
0
Year Relative to Entry/Exit
1
2
Entries/Exits in Beneficiary’s Hospital Service Area
−2
Entries/Exits in Beneficiary’s Hospital Service Area
−2
−1
0
Year Relative to Entry/Exit
1
2
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Entry/Exit on Miles to Hospital
Coefficients on Leads and Lags of Hospital Entry/Exit
−1
−.5
0
.5
Coefficients on Leads and Lags of Hospital Entry/Exit
−.04
−.02
0
.02
.04
−1
1
Entries/Exits in Beneficiary’s Hospital Service Area
Effect of Hospital Entry/Exit on ALOS
Entries/Exits in Beneficiary’s Hospital Service Area
−2
0
Year Relative to Entry/Exit
Effect of Hospital Entry/Exit on Acute Inpatient Days
Coefficients on Leads and Lags of Hospital Entry/Exit
−.01
0
.01
.02
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Effect of Hospital Entry/Exit on Acute Inpatient Stays
−1
2
Entries/Exits in Beneficiary’s Hospital Service Area
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit within
the patient’s hospital service area of residence, normalized so the coefficient on the one-year lead is zero.
The coefficients are from a regression of changes in the outcome on leads and lags of the change in the
number of hospitals from entry and exit and controls including an indicator for gender, age fixed effects,
race fixed effects, year fixed effects, and zip code fixed effects. The regressions are weighted by the number
of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary units. Error bars show the
95-percent confidence interval. Standard errors are clustered by hospital service area.
8
Online Appendix Figure 5: Effect of Hospital Entry and Exit on Quantity of Outpatient Care for
Medicare Patients
Effect of Hospital Entry/Exit on ER Visits
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.04
−.02
0
.02
.04
Effect of Hospital Entry/Exit on Physican Services
Entries/Exits in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Entries/Exits in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−5
0
5
10
Effect of Hospital Entry/Exit on SNF Expenditures
Entries/Exits in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in
the patient’s HSA of residence. The coefficients are from a regression of changes in the outcome on leads
and lags of the change in the number of hospitals from entry and exit and controls including an indicator for
gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects. The regressions are
weighted by the number of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary
units. Error bars show the 95-percent confidence interval. Standard errors are clustered by HSA.
9
Online Appendix Figure 6: Effect of Hospital Entry and Exit on Quantity of Outpatient Care for
Medicare Patients - Outcomes in Levels
Effect of Hospital Entry/Exit on ER Visits
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.004
−.002
0
.002
.004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Effect of Hospital Entry/Exit on Physican Services
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−20
−10
0
10
Effect of Hospital Entry/Exit on SNF Expenditures
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in
the patient’s HSA of residence, normalized so the coefficient on the one-year lead is zero. The coefficients
are from a regression of the level of the outcome on leads and lags of the change in the number of hospitals
from entry and exit and controls including gender time trends, cohort time trends, race time trends, age fixed
effects, year fixed effects, and fixed effects for each cohort-race-sex-zip code-county bin. The regressions
are weighted by the number of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary
units. Error bars show the 95-percent confidence interval. Standard errors are clustered by HSA.
10
Online Appendix Figure 7: Effect of Hospital Entry and Exit on Capacity and Quantity of Treatment with State-Year Fixed Effects
Effect of Hospital Entry/Exit on Admissions
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−2000
0
2000
4000
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
150
Effect of Hospital Entry/Exit on Beds
by HSA
−4
−3
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
−1
0
1
Year Relative to Entry/Exit
2
3
4
3
4
Effect of Hospital Entry/Exit on ALOS
by HSA
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−1
−.5
0
.5
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−20000
0
20000
40000
Effect of Hospital Entry/Exit on Inpatient Days
−2
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Notes: Data cover the 1982-2010 period. Unit of analysis is the HSA-year. Plots are of coefficients on
leads and lags of the change in the number of hospitals from entry and exit from regressions of changes in
each outcome on leads and lags of the change in the number of hospitals from entry and exit, HSA fixed
effects, state-year fixed effects, and demographic and employment controls. Error bars show the 95-percent
confidence interval. Standard errors are clustered by HSA.
11
Online Appendix Figure 8: Effects of Entry and Exit on Capacity and Quantity at All Types of
Hospitals
Effect of Hospital Entry/Exit on Admissions
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−2000
0
2000
4000
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
100
Effect of Hospital Entry/Exit on Beds
by HSA
−4
−3
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−1000
0
1000
2000
Coefficients on Leads and Lags of Hospital Entry/Exit
−20000
0
20000
40000
−2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Effect of Hospital Entry/Exit on Inpatient Surgery
by HSA
−3
2
by HSA
Effect of Hospital Entry/Exit on ER Visits
−4
−1
0
1
Year Relative to Entry/Exit
Effect of Hospital Entry/Exit on ALOS
by HSA
Coefficients on Leads and Lags of Hospital Entry/Exit
−1
0
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−20000
0
20000
40000
Effect of Hospital Entry/Exit on Inpatient Days
−2
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−400
−200
0
200
400
600
Effect of Hospital Entry/Exit on Births at All Hospitals
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. Unit of analysis is the HSA-year. Outcomes in this plot use data
from all hospitals, not just short term, general hospitals. Plots are of coefficients on leads and lags of the
change in the number of hospitals from entry and exit from regressions of changes in each outcome on leads
and lags of the change in the number of hospitals from entry and exit, HSA fixed effects, year fixed effects,
and demographic and employment controls. Error bars show the 95-percent confidence interval. Standard
errors are clustered by HSA.
12
Online Appendix Figure 9: Effect of Hospital Entry and Exit on Admissions at Competing Hospitals
Effects on Closest Competitors − On Impact
Effects on Closest Competitors − After 1 Year
Closest Hospital
Closest Hospital
2nd Closest Hospital
2nd Closest Hospital
3rd Closest Hospital
3rd Closest Hospital
4th Closest Hospital
4th Closest Hospital
5th Closest Hospital
5th Closest Hospital
−200
−100
0
Coefficient on Hospital Entry/Exit
100
−200
−100
0
100
Coefficient on Hospital Entry/Exit
200
Notes: Data cover the 1982-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients
on hospital entry/exit variable from regressions of changes in the number of admissions at each competing
hospital on a hospital’s entry/exit variable controlling for HSA fixed effects, year fixed effects, and demographic and employment controls. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA of the entering/exiting hospital.
13
Online Appendix Figure 10: Effect of Hospital Entry and Exit on Nearby Hospitals - Alternative
Missing Data Assumptions
Effects on Closest Competitors− On Impact
Effects on Closest Competitors − One Lag
No Imputed Data
No Imputed Data
Closest Hospital
Closest Hospital
2nd Closest Hospital
2nd Closest Hospital
3rd Closest Hospital
3rd Closest Hospital
4th Closest Hospital
4th Closest Hospital
5th Closest Hospital
5th Closest Hospital
−200
−100
0
Coefficient on Hospital Entry/Exit
100
−200
Effects on Closest Competitors − On Impact
−100
0
100
Coefficient on Hospital Entry/Exit
200
Effects on Closest Competitors − One Lag
Only Hospitals with No Imputed Data
Only Hospitals with No Imputed Data
Closest Hospital
Closest Hospital
2nd Closest Hospital
2nd Closest Hospital
3rd Closest Hospital
3rd Closest Hospital
4th Closest Hospital
4th Closest Hospital
5th Closest Hospital
5th Closest Hospital
−200
−100
0
100
Coefficient on Hospital Entry/Exit
200
−300
−200
−100
0
Coefficient on Hospital Entry/Exit
100
Notes: Data cover the 1983-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients
on hospital entry/exit variable from regressions of changes in the number of admissions at each competing
hospital on a hospital’s entry/exit variable controlling for HSA fixed effects, year fixed effects, and demographic and employment controls. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA of the entering/exiting hospital.
14
Online Appendix Figure 11: Effects of Entry and Exit on Quantity of Care in the State Inpatient
Databases
Effect of Hospital Entry/Exit on Inpatient Days
by HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−40000 −20000
0
20000 40000 60000
Coefficients on Leads and Lags of Hospital Entry/Exit
−5000
0
5000
10000
Effect of Hospital Entry/Exit on Discharges
−3
−2
−1
0
Year Relative to Entry/Exit
1
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Entry/Exit on Hospital Charges Per Patient
by HSA
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−1000
−500
0
500
1000
Coefficients on Leads and Lags of Hospital Entry/Exit
−2.00e+08
0
2.00e+08
4.00e+08
Effect of Hospital Entry/Exit on Hospital Charges
by HSA
by HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.1
0
.1
.2
Effect of Hospital Entry/Exit on ALOS
by HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Discharge data cover subsets the 1995-2011 period that vary by state for 9 states. Unit of analysis
is the HSA-year. Plots are of coefficients on leads and lags of the change in the number of hospitals from
entry and exit from regressions of changes in each outcome on leads and lags of the change in the number
of hospitals from entry and exit, year fixed effects, and demographic and employment controls. Error bars
show the 95-percent confidence interval. Standard errors are clustered by HSA. Because the sample period
for some of the states is short, the sample of leads and lags of entries/exits is not balanced and HSA fixed
effects are not included.
15
−3
by HSA
−2
2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
Effect of Hospital Entry on Inpatient Days
Effect of Hospital Exit on Inpatient Days
by HSA
by HSA
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Effect of Hospital Entry on ALOS
Effect of Hospital Exit on ALOS
by HSA
by HSA
4
3
4
3
4
3
4
Coefficients on Leads and Lags of Hospital Exit
−1
−.5
0
.5
1
Coefficients on Leads and Lags of Hospital Entry
−1
−.5
0
.5
1
−4
−1
0
1
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Exit
−2000
0
2000
4000
Effect of Hospital Exit on Admissions
by HSA
Coefficients on Leads and Lags of Hospital Exit
−20000
0
20000
40000
−4
Effect of Hospital Entry on Admissions
Coefficients on Leads and Lags of Hospital Entry
−20000
0
20000
40000
Coefficients on Leads and Lags of Hospital Entry
−4000
−2000
0
2000
4000
Online Appendix Figure 12: Effects of Entry versus Exit
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Effect of Hospital Entry on Mortality
Effect of Hospital Exit on Mortality
by HSA
by HSA
Coefficients on Leads and Lags of Hospital Exit
−20
−10
0
10
Coefficients on Leads and Lags of Hospital Entry
−20
−10
0
10
20
−4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Notes: Data cover the 1982-2010 period. Unit of analysis is the HSA-Year. Plots are of coefficients on leads
and lags of the change in the number of hospitals from entry or exit from regressions of changes in each
outcome on leads and lags of the change in the number of hospitals from entry or exit, HSA fixed effects,
year fixed effects, and demographic and employment controls. Error bars show the 95-percent confidence
interval. Standard errors are clustered by HSA.
16
Online Appendix Figure 13: Effects of Entry versus Exit - Medicare Beneficiaries
Effect of Hospital Net Exit on Total Expenditures
Net Entry in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
100
Effect of Hospital Net Entry on Total Expenditures
Net Entry in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
0
.001
.002
.003
Coefficients on Leads and Lags of Hospital Entry/Exit
−.004
−.002
0
.002
.004
0
Year Relative to Entry/Exit
−2
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.01
0
.01
.02
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
−.01
0
.01
.02
0
Year Relative to Entry/Exit
−2
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0005
0
.0005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0005
0
.0005
0
Year Relative to Entry/Exit
1
2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Net Exit on Mortality
Net Entry in Beneficiary’s HSA
−1
0
Year Relative to Entry/Exit
Net Exit in Beneficiary’s HSA
Effect of Hospital Net Entry on Mortality
−2
−1
Effect of Hospital Net Exit on Acute Inpatient Days
Net Entry in Beneficiary’s HSA
−1
2
Net Exit in Beneficiary’s HSA
Effect of Hospital Net Entry on Acute Inpatient Days
−2
1
Effect of Hospital Net Exit on Acute Inpatient Stays
Net Entry in Beneficiary’s HSA
−1
0
Year Relative to Entry/Exit
Net Exit in Beneficiary’s HSA
Effect of Hospital Net Entry on Acute Inpatient Stays
−2
−1
Effect of Hospital Net Exit on Acute Inpatient Expenditures
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Effect of Hospital Net Entry on Acute Inpatient Expenditures
Net Exit in Beneficiary’s HSA
Net Exit in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year. Plots are
of coefficients on leads and lags of the change in the number of hospitals from entry or exit in the patient’s HSA
of residence with the first lead sometimes normalized to zero. Coefficients are from regressions are weighted by the
number of Medicare beneficiaries and outcomes are in per-beneficiary units. Error bars show the 95-percent confidence
interval, clustered by HSA.
17
Online Appendix Figure 14: Effect of Hospital Entry and Exit on Capacity and Quantity of Treatment - Outcomes in Levels
Effect of Hospital Entry/Exit on Admissions
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
Coefficients on Leads and Lags of Hospital Entry/Exit
−5000
0
5000
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Hospital Entry/Exit on Beds
by HSA
−6
−5
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
5
6
Effect of Hospital Entry/Exit on ALOS
by HSA
5
6
Coefficients on Leads and Lags of Hospital Entry/Exit
−.1
0
.1
.2
Coefficients on Leads and Lags of Hospital Entry/Exit
−20000
0
20000
40000
Effect of Hospital Entry/Exit on Days
−4
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the HSA-year. Plots are of coefficients on
leads and lags of the change in the number of hospitals from entry and exit normalized so the one-year lead
is zero. The estimates are from a regression of the level of the outcome within a HSA on leads and lags of
the change in the number of hospitals from entry and exit within that HSA and controls including HSA fixed
effects, HSA time trends, year fixed effects, and demographic and employment variables. Error bars show
the 95-percent confidence interval. Standard errors are clustered by HSA.
18
Online Appendix Figure 15: Effect of Hospital Entry and Exit on Capacity and Quantity of Treatment - Logged Outcomes in Levels
Effect of Hospital Entry/Exit on Log Admissions
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.1
−.05
0
.05
.1
Effect of Hospital Entry/Exit on Log Beds
by HSA
−6
−5
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
5
6
Effect of Hospital Entry/Exit on Log ALOS
by HSA
5
6
Coefficients on Leads and Lags of Hospital Entry/Exit
−.01
0
.01
.02
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Effect of Hospital Entry/Exit on Log Days
−4
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the HSA-year. Plots are of coefficients on
leads and lags of the change in the number of hospitals from entry and exit normalized so the one-year lead
is zero. The estimates are from a regression of the level of the log outcome within a HSA on leads and lags
of the change in the number of hospitals from entry and exit within that HSA and controls including HSA
fixed effects, HSA time trends, year fixed effects, and demographic and employment variables. Error bars
show the 95-percent confidence interval. Standard errors are clustered by HSA.
19
Online Appendix Figure 16: Effect of Hospital Entry and Exit on Competing Hospitals by Distance
- Admissions in Logs
Effect of Hospital Entry/Exit on Admissions
for 0−5 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.04
−.02
0
.02
.04
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
0
.02
.04
Effect of Hospital Entry/Exit on Admissions
for 5−10 miles
−4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Effect of Hospital Entry/Exit on Admissions
for 10−20 miles
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
0
.02
.04
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
0
.02
.04
Effect of Hospital Entry/Exit on Admissions
−3
for 20−50 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
0
.02
.04
Effect of Hospital Entry/Exit on Admissions
for 50−150 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients
on hospital entry/exit variable from regressions of changes in the log number of admissions at each competing hospital on a hospital’s entry/exit variable controlling for HSA fixed effects, year fixed effects, and
demographic and employment controls. Error bars show the 95-percent confidence interval. Standard errors
are clustered by HSA of the entering/exiting hospital.
20
Online Appendix Figure 17: Effect of Hospital Entry and Exit on Mortality in Levels
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
Effect of Hospital Entry/Exit on Mortality − Weighted
Coefficients on Leads and Lags of Hospital Entry/Exit
−20
−10
0
10
Coefficients on Leads and Lags of Hospital Entry/Exit
−20
−10
0
10
Effect of Hospital Entry/Exit on Mortality − Unweighted
by HSA
−6
−5
−4
−3
−2
−1
0
1
2
Year Relative to Entry/Exit
3
4
5
6
Notes: Data cover the 1982-2010 period. The unit of analysis is the HSA-year. The estimates are from
a regression of the level of the outcome within a HSA on leads and lags of the change in the number
of hospitals from entry and exit within that HSA and controls including HSA fixed effects, HSA time
trends, year fixed effects, and demographic and employment variables. Estimates are normalized so the
coefficient on the one-year lead is zero. Error bars show the 95-percent confidence interval. Standard errors
are clustered by HSA.
21
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality for Ages 0−64 − Weighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality for Ages 65 plus − Weighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
−5
0
5
10
−4
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
−5
0
5
10
Mortality − Weighted
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
Coefficients on Leads and Lags of Hospital Entry/Exit
−20
0
20
40
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
−5
0
5
10
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
−5
0
5
10
Online Appendix Figure 18: Effect of Hospital Entry and Exit on the Mortality Rate in Counties
Mortality − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
3
4
Mortality for Ages 0−64 − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Mortality for Ages 65 plus − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. Unit of analysis is the county-year. Plots are of coefficients on
leads and lags of the change in the number of hospitals from entry and exit from regressions of changes in
the mortality rate on leads and lags of the change in the number of hospitals from entry and exit, county
fixed effects, year fixed effects, and demographic and employment controls. Error bars show the 95-percent
confidence interval. Standard errors are clustered by state.
22
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 0−64 − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 0−64 − Weighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
−4
Coefficients on Leads and Lags of Hospital Entry/Exit
−200
0
200
400
Mortality in Rural Areas − Weighted
Coefficients on Leads and Lags of Hospital Entry/Exit
−200
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Online Appendix Figure 19: Effect of Hospital Entry and Exit on the Mortality Rate in Rural
Counties
Mortality in Rural Areas − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 65 plus − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 65 plus − Weighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period for rural areas. Unit of analysis is the county-year. Plots are of
coefficients on leads and lags of the change in the number of hospitals from entry and exit from regressions
of changes in the mortality rate on leads and lags of the change in the number of hospitals from entry and
exit, county fixed effects, year fixed effects, and demographic and employment controls. Error bars show
the 95-percent confidence interval. Standard errors are clustered by state.
23
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 0−64 − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Mortality in Rural Areas − Weighted
Coefficients on Leads and Lags of Hospital Entry/Exit
−200
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
−50
0
50
Online Appendix Figure 20: Effect of Hospital Entry and Exit on the Mortality Rate in Rural HSAs
Mortality in Rural Areas − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Mortality in Rural Areas for Ages 65 plus − Unweighted
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period for rural areas. Unit of analysis is the HSA-year. Plots are of
coefficients on leads and lags of the change in the number of hospitals from entry and exit from regressions
of changes in the mortality rate on leads and lags of the change in the number of hospitals from entry and
exit, HSA fixed effects, year fixed effects, and demographic and employment controls. Error bars show the
95-percent confidence interval. Standard errors are clustered by HSA.
24
Online Appendix Figure 21: Effect of Hospital Entry and Exit on Self-Reported Health in Levels
Effect of Hospital Entry/Exit on Poor Physical Health
by County
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Effect of Hospital Entry/Exit on Poor Health
by County
−2
−2
−1
0
Year Relative to Entry/Exit
1
by County
−2
−1
0
Year Relative to Entry/Exit
−1
0
Year Relative to Entry/Exit
1
Effect of Hospital Entry/Exit on Use Equipment
by County
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Effect of Hospital Entry/Exit on Unable to Work
−2
1
Effect of Hospital Entry/Exit on Limited Activities
by County
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
.1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
Effect of Hospital Entry/Exit on Poor Mental Health
−1
0
Year Relative to Entry/Exit
by County
−2
−1
0
Year Relative to Entry/Exit
1
Notes: Data cover the 2002-2010 period. The unit of analysis is the county-year. Plots are of coefficients
the change in the number of hospitals from entry and exit from regressions of levels in each outcome on the
change in the number of hospitals from entry and exit, county fixed effects, county time trends, year fixed
effects, and demographic and employment controls. Error bars show the 95-percent confidence interval.
Standard errors are clustered by state.
25
Online Appendix Figure 22: Effect of Hospital Entry and Exit on Mortality for Medicare Patients
- Outcomes in Levels
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Effect of Hospital Entry/Exit on Deaths
Entries/Exits in Beneficiary’s County
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
0
.001
.002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0004
−.0002
0
.0002
Effect of Hospital Entry/Exit on Deaths
Entries/Exits in Beneficiary’s Hospital Service Area
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Entry/Exit within 10 Miles on Mortality
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit
in markets defined using the patient’s residence. The coefficients are from a regression of changes in the
outcome on leads and lags of the change in the number of hospitals from entry and exit and controls including
an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects.
The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes are all
in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are clustered
by HSA except the in the hospitals service area specification where they are clustered by hospital service
area.
26
Online Appendix Figure 23: Effect of Hospital Entry and Exit on Mortality for Medicare Patients
- Outcomes in Levels
Effect of Hospital Entry/Exit on Deaths
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Effect of Hospital Entry/Exit on Mortality
Entries/Exits in Beneficiary’s County
−3
−3
−2
−1
0
Year Relative to Entry/Exit
1
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Entry/Exit on Deaths
Entries/Exits in Beneficiary’s Hospital Service Area
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
−.001
0
.001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0015
−.001
−.0005
0
.0005
Effect of Hospital Entry/Exit on Deaths
−2
Entries/Exits within 10 Miles of Beneficiary
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in the
patient’s HSA, county, hospital service area, or within 10 miles of the beneficiary’s residence, normalized so
the coefficient on the one-year lead is zero. The coefficients are from a regression of the level of the outcome
on leads and lags of the change in the number of hospitals from entry and exit and controls including gender
time trends, cohort time trends, race time trends, age fixed effects, year fixed effects, and fixed effects
for each cohort-race-sex-zip code-county bin. The regressions are weighted by the number of Medicare
beneficiaries in each bin and the outcomes are all in per-beneficiary units. Error bars show the 95-percent
confidence interval. Standard errors are clustered by HSA or hospital service area.
27
Online Appendix Figure 24: Effect of Hospital Entry vs. Exit on Mortality for Medicare Patients
- in Levels
Effect of Hospital Net Exit on Mortality
Net Entry in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Effect of Hospital Net Entry on Mortality
Net Exit in Beneficiary’s HSA
−3
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
−.001
0
.001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
−1
0
Year Relative to Entry/Exit
2
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Hospital Net Exit on Mortality
Net Entry in Beneficiary’s Hospital Service Area
−2
1
Net Exit in Beneficiary’s County
Effect of Hospital Net Entry on Mortality
−3
−1
0
Year Relative to Entry/Exit
Effect of Hospital Net Exit on Mortality
Net Entry in Beneficiary’s County
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0015
−.001
−.0005
0
.0005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
0
.001
Effect of Hospital Net Entry on Mortality
−2
Net Exit in Beneficiary’s Hospital Service Area
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in
the patient’s HSA, county, or hospital service area of residence, normalized so the coefficient on the oneyear lead is zero. The coefficients are from a regression of the level of the outcome on leads and lags of the
change in the number of hospitals from entry and exit and controls including gender time trends, cohort time
trends, race time trends, age fixed effects, year fixed effects, and fixed effects for each cohort-race-sex-zip
code-county bin. The regressions are weighted by the number of Medicare beneficiaries in each bin and
the outcomes are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard
errors are clustered by HSA or hospital service area.
28
Online Appendix Figure 25: Comparing Medicare Recoded Entry/Exit to Aggregate Entry/Exit Effect on the Mortality Rate in Counties from 1999-2010
Medicare Sample Entry/Exit − Weighted
by County
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
0
10
20
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
0
10
20
Aggregate Sample Entry/Exit − Weighted
by County
−2
−2
−1
0
Year Relative to Entry/Exit
1
2
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
0
Year Relative to Entry/Exit
by County
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
Aggregate Sample Entry/Exit − Over 65 Population − Unweighted
−2
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Medicare Sample Entry/Exit − Over 65 − Weighted
by County
−1
1
by County
Aggregate Sample Entry/Exit − Over 65 − Weighted
−2
0
Year Relative to Entry/Exit
Medicare Sample Entry/Exit − Unweighted
by County
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
0
10
20
Coefficients on Leads and Lags of Hospital Entry/Exit
−10
0
10
20
Aggregate Sample Entry/Exit − Unweighted
−1
by County
−2
−1
0
Year Relative to Entry/Exit
1
2
Medicare Sample Entry/Exit − Over 65 − Unweighted
by County
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2010 period. Unit of analysis is the county-year. Plots are of coefficients on
leads and lags of the change in the number of hospitals from entry and exit from regressions of changes in
the mortality rate on leads and lags of the change in the number of hospitals from entry and exit, county
fixed effects, year fixed effects, and demographic and employment controls. Error bars show the 95-percent
confidence interval. Standard errors are clustered by state. The figures in the left column use the main sample
of entries/exits and the figures in the right column sample of entries/exits recoded using the Medicare data.
29
−1
0
Year Relative to Entry/Exit
1
2
Effect of Entry/Exit within 15 Miles on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Entry/Exit within 30 Miles on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Entry/Exit within 5 Miles on Total Expenditures
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Online Appendix Figure 26: Effect of Hospital Entry and Exit on Total Medicare Spending by
Distance to Event
Effect of Entry/Exit within 10 Miles on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Entry/Exit within 20 Miles on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Entry/Exit within 50 Miles on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit within
a certain distance of the patient’s zip code of residence. The coefficients are from a regression of changes
in the outcome on leads and lags of the change in the number of hospitals from entry and exit and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
30
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−20
−10
0
10
Entry/Exit of Closest Hospital on Total Expenditures
2
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.001
−.0005
0
.0005
Entry/Exit of Closest Hospital on Acute Inpatient Stays
2
Entry/Exit of Closest Hospital on Mortality
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
0
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
−.001
0
.001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.005
0
.005
.01
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
100
150
Online Appendix Figure 27: Effect of Hospital Entry/Exit on Quantity of Care and Deaths by
Distance to Closest Hospital
2
One Mile Increase in Distance on Total Expenditures
−2
−1
0
Year Relative to Entry/Exit
1
2
One Mile Increase in Distance on Acute Inpatient Stays
−2
−1
0
Year Relative to Entry/Exit
1
2
One Mile Increase in Distance on Mortality
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of either a variable that indicates a change in distance to the closest
hospital or the change in distance to the closest hospital caused by an entry or an exit. The coefficients are
from a regression of changes in the outcome on leads and lags of the independent variable of interest and
controls including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip
code fixed effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and
the outcomes are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard
errors are clustered by HSA.
31
Online Appendix Figure 28: Effect of Hospital Entry and Exit on Expenditures and Admissions
by Change in Distance Terciles
Effect of Distance to the Hospital on Acute Inpatient Stays
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
−.01
0
.01
Coefficients on Leads and Lags of Hospital Entry/Exit
−400
−200
0
200
Effect of Distance to the Hospital on Total Expenditures
1st Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
2nd Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
3rd Tercile of Increase in Distance
−1
0
Year Relative to Entry/Exit
1
−1
0
Year Relative to Entry/Exit
1
2
2nd Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect of Distance to the Hospital on Acute Inpatient Stays
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
−.01
0
.01
Coefficients on Leads and Lags of Hospital Entry/Exit
−400
−200
0
200
Effect of Distance to the Hospital on Total Expenditures
−2
−2
Effect of Distance to the Hospital on Acute Inpatient Stays
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
−.01
0
.01
Coefficients on Leads and Lags of Hospital Entry/Exit
−400
−200
0
200
Effect of Distance to the Hospital on Total Expenditures
1st Tercile of Increase in Distance
2
3rd Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of a variable that indicates a change in distance to the closest
hospital for terciles of the change distance caused by an entry or an exit. The coefficients are from a
regression of changes in the outcome on leads and lags of the independent variable of interest and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
32
Effect on UTI
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on GI Hemorrhage
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Respiratory Failure
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Stroke
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Asthma
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Congestive Heart Failure
−2
−1
0
Year Relative to Entry/Exit
1
2
0
Year Relative to Entry/Exit
1
2
Effect on Septicemia
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Pneumonia
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on COPD
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Fluids Electrolytes
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
2
Effect on AMI
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
0
Year Relative to Entry/Exit
1
Effect on Aspiration Pneumonitis
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
0
.0002
.0004
−2
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
.0002
Effect on TIA
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
2
Effect on Hip Fracture
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Effect on Fainting
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002 −.0001
0
.0001
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
Online Appendix Figure 29: Effect of Hospital Entry and Exit on Discharges by Diagnosis - Part 1
Effect on Intestinal Obstruction
−2
−1
0
Year Relative to Entry/Exit
1
2
1
2
1
2
Effect on Chest Pain
−2
−1
0
Year Relative to Entry/Exit
Effect on Other Fractures
−2
−1
0
Year Relative to Entry/Exit
Effect on Diverticulosis Diverticulitis
−2
−1
0
Year Relative to Entry/Exit
1
2
1
2
Effect on Renal Failure
−2
−1
0
Year Relative to Entry/Exit
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show 20 of the top 40 diagnoses during the sample period ordered by the share of admissions
that occur on weekends. The plotted coefficients are from a regression of changes in the outcome on leads
and lags of the change in the number of hospitals from entry and exit within the patients HSA and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
33
Effect on Cognitive Disorders
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Lung Cancer
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Atherosclerosis
−2
−1
0
Year Relative to Entry/Exit
1
2
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Anemia
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Complications Device
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Back Problems
−2
−1
0
Year Relative to Entry/Exit
1
2
0
Year Relative to Entry/Exit
1
2
Effect on Complications Procedure
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Mood Disorders
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Heart Disease
−2
−1
0
Year Relative to Entry/Exit
1
2
Effect on Narrowing Precerebral Arteries
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
2
Effect on Cardiac Dysrhythmias
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0004
−.0002
0
.0002
.0004
0
Year Relative to Entry/Exit
1
Effect on Skin Infection
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
−2
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
.00015
Effect on Other GI
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
.0002
2
Effect on Biliary Tract Disease
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00004
−.00002
0
.00002
.00004
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
−1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Effect on Hypertension Complication
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00004 −.00002
0
.00002
.00004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00002
0
.00002
.00004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Online Appendix Figure 30: Effect of Hospital Entry and Exit on Discharges by Diagnosis- Part 2
Effect on Diabetes with Comp
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Secondary Malignancies
−2
−1
0
Year Relative to Entry/Exit
1
2
1
2
1
2
1
2
Effect on Blood Clot
−2
−1
0
Year Relative to Entry/Exit
Effect on Rehabilitation
−2
−1
0
Year Relative to Entry/Exit
Effect on Osteoarthritis
−2
−1
0
Year Relative to Entry/Exit
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show 20 of the top 40 diagnoses during the sample period ordered by the share of admissions
that occur on weekends. The plotted coefficients are from a regression of changes in the outcome on leads
and lags of the change in the number of hospitals from entry and exit within the patients HSA and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
34
2
Online Appendix Figure 31: Effect of Hospital Entry and Exit on Discharges by Diagnosis - Net
Entry of Closest Hospital
Less Deferrable in Diagnoses in Top 40
Fainting
Hip fracture
Aspiration pneumonitis
Intestinal obstruction
TIA
AMI
Septicemia
Chest pain
UTI
Stroke
Pneumonia
Other fractures
GI hemorrhage
Asthma
COPD
Diverticulosis diverticulitis
Respiratory failure
Congestive heart failure
Fluids electrolytes
Renal failure
−.001
0
.001
Coefficient on Hospital Entry/Exit
.002
More Deferrable in Diagnoses in Top 40
Hypertension complication
Biliary tract disease
Skin infection
Diabetes with comp
Other GI
Cardiac dysrhythmias
Complications procedure
Secondary malignancies
Cognitive disorders
Anemia
Mood disorders
Blood clot
Lung cancer
Complications device
Heart disease
Rehabilitation
Atherosclerosis
Back problems
Narrowing precerebral arteries
Osteoarthritis
−.001
0
.001
Coefficient on Hospital Entry/Exit
.002
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 40 diagnoses during the sample period ordered by the share of admissions that occur
on weekends. The plotted coefficients are from a regression of changes in the outcome on two leads and one
lag of the change in the number of hospitals from entry and exit that affect the distance to the closest hospital
and controls including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and
zip code fixed effects. The regressions are weighted by the number of Medicare beneficiaries in each bin
and the outcomes are all in per-beneficiary units. The plotted coefficients are the linear combination of the
sum of the on impact effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars
show the 95-percent confidence interval. Standard errors are clustered by HSA.
35
Online Appendix Figure 32: Effect of Hospital Entry and Exit on Discharges by Diagnosis - Net
Entry in the County
Less Deferrable in Diagnoses in Top 40
Fainting
Hip fracture
Aspiration pneumonitis
Intestinal obstruction
TIA
AMI
Septicemia
Chest pain
UTI
Stroke
Pneumonia
Other fractures
GI hemorrhage
Asthma
COPD
Diverticulosis diverticulitis
Respiratory failure
Congestive heart failure
Fluids electrolytes
Renal failure
−.0005
0
Coefficient on Hospital Entry/Exit
.0005
More Deferrable in Diagnoses in Top 40
Hypertension complication
Biliary tract disease
Skin infection
Diabetes with comp
Other GI
Cardiac dysrhythmias
Complications procedure
Secondary malignancies
Cognitive disorders
Anemia
Mood disorders
Blood clot
Lung cancer
Complications device
Heart disease
Rehabilitation
Atherosclerosis
Back problems
Narrowing precerebral arteries
Osteoarthritis
−.0005
0
Coefficient on Hospital Entry/Exit
.0005
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 40 diagnoses during the sample period ordered by the share of admissions that occur
on weekends. The plotted coefficients are from a regression of changes in the outcome on two leads and
one lag of the change in the number of hospitals from entry and exit within the patients county and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. The plotted coefficients are the linear combination of the sum of the on
impact effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars show the
95-percent confidence interval. Standard errors are clustered by HSA.
36
Online Appendix Figure 33: Effect of Hospital Entry and Exit on Discharges by Diagnosis - Net
Entry in the Hospital Service Area
Less Deferrable in Diagnoses in Top 40
Fainting
Hip fracture
Aspiration pneumonitis
Intestinal obstruction
TIA
AMI
Septicemia
Chest pain
UTI
Stroke
Pneumonia
Other fractures
GI hemorrhage
Asthma
COPD
Diverticulosis diverticulitis
Respiratory failure
Congestive heart failure
Fluids electrolytes
Renal failure
−.0005
0
.0005
Coefficient on Hospital Entry/Exit
.001
More Deferrable in Diagnoses in Top 40
Hypertension complication
Biliary tract disease
Skin infection
Diabetes with comp
Other GI
Cardiac dysrhythmias
Complications procedure
Secondary malignancies
Cognitive disorders
Anemia
Mood disorders
Blood clot
Lung cancer
Complications device
Heart disease
Rehabilitation
Atherosclerosis
Back problems
Narrowing precerebral arteries
Osteoarthritis
−.001
0
.001
Coefficient on Hospital Entry/Exit
.002
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 40 diagnoses during the sample period ordered by the share of admissions that occur
on weekends. The plotted coefficients are from a regression of changes in the outcome on two leads and one
lag of the change in the number of hospitals from entry and exit within the patients hospital service area and
controls including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip
code fixed effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and
the outcomes are all in per-beneficiary units. The plotted coefficients are the linear combination of the sum
of the on impact effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars show
the 95-percent confidence interval. Standard errors are clustered by hospital service area.
37
Online Appendix Figure 34: Effect of Hospital Entry and Exit on Discharges by Diagnosis - Net
Entry within 10 Miles
Less Deferrable in Diagnoses in Top 40
Fainting
Hip fracture
Aspiration pneumonitis
Intestinal obstruction
TIA
AMI
Septicemia
Chest pain
UTI
Stroke
Pneumonia
Other fractures
GI hemorrhage
Asthma
COPD
Diverticulosis diverticulitis
Respiratory failure
Congestive heart failure
Fluids electrolytes
Renal failure
−.0005
0
.0005
Coefficient on Hospital Entry/Exit
.001
More Deferrable in Diagnoses in Top 40
Hypertension complication
Biliary tract disease
Skin infection
Diabetes with comp
Other GI
Cardiac dysrhythmias
Complications procedure
Secondary malignancies
Cognitive disorders
Anemia
Mood disorders
Blood clot
Lung cancer
Complications device
Heart disease
Rehabilitation
Atherosclerosis
Back problems
Narrowing precerebral arteries
Osteoarthritis
−.001
−.0005
0
Coefficient on Hospital Entry/Exit
.0005
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 40 diagnoses during the sample period ordered by the share of admissions that occur
on weekends. The plotted coefficients are from a regression of changes in the outcome on two leads and
one lag of the change in the number of hospitals from entry and exit within 10 miles of the beneficiary and
controls including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip
code fixed effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and
the outcomes are all in per-beneficiary units. The plotted coefficients are the linear combination of the sum
of the on impact effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars show
the 95-percent confidence interval. Standard errors are clustered by HSA.
38
Effect of Hospital Entry on Births
Effect of Hospital Exit on Births
by HSA
by HSA
Coefficients on Leads and Lags of Hospital Exit
−400
−200
0
200
400
600
Coefficients on Leads and Lags of Hospital Entry
−1500
−1000
−500
0
500
Online Appendix Figure 35: Effect of Hospital Entry and Exit on Births - Robustness Checks
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.02
0
.02
.04
Coefficients on Leads and Lags of Hospital Entry/Exit
−5
0
5
10
Effect of Hospital Entry/Exit on Births − Per Bed
Coefficients on Leads and Lags of Hospital Entry/Exit
−50
0
50
Effect of Hospital Entry/Exit on Births − Rural Areas
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
2
3
4
Effect of Hospital Entry/Exit on Births − in Logs
by HSA
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
Within HSA
Within CI
2
3
4
Cross HSA
Cross CI
Notes: Data cover the 1982-2010 period. Unit of analysis is the HSA-year. Plots are of coefficients on leads
and lags of the change in the number of hospitals from entry and exit from regressions of changes in each
outcome on leads and lags of the change in the number of hospitals from entry and exit, HSA fixed effects,
year fixed effects, and demographic and employment controls. Error bars show the 95-percent confidence
interval. Standard errors are clustered by HSA.
39
Online Appendix Figure 36: Effect of Hospital Entry and Exit on Nearby Hospitals
Effect of Hospital Entry/Exit on Births
on closest hospital
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Hospital Entry/Exit on Births
on 2nd closest hospital
−4
−3
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
−1
0
1
Year Relative to Entry/Exit
2
3
4
3
4
Effect of Hospital Entry/Exit on Births
on 3rd closest hospital
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Hospital Entry/Exit on Births
−2
on 4th closest hospital
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Hospital Entry/Exit on Births
on 5th closest hospital
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients on
hospital entry/exit variable from regressions of changes in the number of births at each competing hospital
on a hospital’s entry/exit variable controlling for HSA fixed effects, year fixed effects, and demographic and
employment controls. Error bars show the 95-percent confidence interval. Standard errors are clustered by
HSA of the entering/exiting hospital.
40
Online Appendix Figure 37: Effect of Hospital Entry and Exit on Births by Distance
Effect of Hospital Entry/Exit on Births
for 0−5 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−200
−100
0
100
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−100
0
100
200
Effect of Hospital Entry/Exit on Births
for 0−10 miles
−4
−3
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
−1
0
1
Year Relative to Entry/Exit
2
3
4
3
4
Effect of Hospital Entry/Exit on Births
for 0−20 miles
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−400
−200
0
200
Coefficients on Leads and Lags of Hospital Entry/Exit
−200
−100
0
100
200
Effect of Hospital Entry/Exit on Births
−2
for 0−50 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−500
0
500
Effect of Hospital Entry/Exit on Births
for 0−150 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients
on hospital entry/exit variable from regressions of changes in the total number of births within the specified
distance of the entering or exiting hospital on a hospital’s entry/exit variable controlling for the entry/exit
of other hospitals in that area, HSA fixed effects, year fixed effects, and demographic and employment
controls. Error bars show the 95-percent confidence interval. Standard errors are clustered by HSA of the
entering/exiting hospital.
41
Online Appendix Figure 38: Effect of Hospital Entry and Exit on Births by Distance - Log Outcomes
Effect of Hospital Entry/Exit on Births
for 0−5 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.2
0
.2
.4
.6
Coefficients on Leads and Lags of Hospital Entry/Exit
−.5
0
.5
1
Effect of Hospital Entry/Exit on Births
for 0−10 miles
−4
−3
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
−1
0
1
Year Relative to Entry/Exit
2
3
4
3
4
Effect of Hospital Entry/Exit on Births
for 0−20 miles
3
4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.2
0
.2
.4
Coefficients on Leads and Lags of Hospital Entry/Exit
−.2
0
.2
.4
Effect of Hospital Entry/Exit on Births
−2
for 0−50 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.05
0
.05
Effect of Hospital Entry/Exit on Births
for 0−150 miles
−4
−3
−2
−1
0
1
Year Relative to Entry/Exit
2
3
4
Notes: Data cover the 1982-2010 period. The unit of analysis is the hospital-year. Plots are of coefficients
on hospital entry/exit variable from regressions of changes in log births at each competing hospital on a
hospital’s entry/exit variable controlling for HSA fixed effects, year fixed effects, and demographic and
employment controls. Error bars show the 95-percent confidence interval. Standard errors are clustered by
HSA of the entering/exiting hospital.
42
Online Appendix Figure 39: Effect of Hospital Entry and Exit on Severity of Diagnosis
Effect of Hospital Entry/Exit on Death Rate of Diagnosis
Entries/Exits in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.005
0
.005
Effect of Hospital Entry/Exit on Charlson Index
Entries/Exits in Beneficiary’s HSA
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in
the patient’s HSA of residence. The coefficients are from a regression of changes in the outcome on leads
and lags of the change in the number of hospitals from entry and exit within the patients HSA and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
43
Online Appendix Figure 40: Effect of Hospital Entry and Exit on Diagnosis Severity - Outcome in
Levels
Effect of Hospital Entry/Exit on Death Rate of Diagnosis
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.01
−.005
0
.005
Effect of Hospital Entry/Exit on Charlson Index
Entries/Exits in Beneficiary’s HSA
−3
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of the change in the number of hospitals from entry and exit in the
patient’s HSA of residence, normalized so the coefficient on the one-year lead is zero. The coefficients are
from a regression of the level of the outcome on leads and lags of the change in the number of hospitals from
entry and exit within the patients HSA and controls including gender time trends, cohort time trends, race
time trends, age fixed effects, year fixed effects, and fixed effects for each cohort-race-sex-zip code-county
bin. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes are
all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are clustered
by HSA.
44
Online Appendix Figure 41: Effect of Hospital Entry and Exit on Deaths by Change in Distance
Terciles
Effect of Distance to the Hospital on Mortality
1st Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Effect of Distance to the Hospital on Mortality
2nd Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.002
0
.002
.004
Effect of Distance to the Hospital on Mortality
3rd Tercile of Increase in Distance
−2
−1
0
Year Relative to Entry/Exit
1
2
Notes: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
Plots are of coefficients on leads and lags of a variable that indicates a change in distance to the closest
hospital for terciles of changes in distance that are caused by an entry or an exit. The coefficients are from a
regression of changes in the outcome on leads and lags of the independent variable of interest and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
45
1
−2
−1
0
Year Relative to Entry/Exit
1
−1
0
Year Relative to Entry/Exit
1
Effect on All Other Accidents and Adverse Effects
−2
−1
0
Year Relative to Entry/Exit
Effect on Chronic Lower Respiratory Diseases
−2
1
−1
0
Year Relative to Entry/Exit
1
Effect on Alzheimer’s Disease
−2
.
Effect on Diabetes Mellitus
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
Effect on Other Cancers
−1
0
Year Relative to Entry/Exit
1
Effect on Colorectal Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Prostate Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0004
−.0002
0
.0002
.0004
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
.0002
−1
0
Year Relative to Entry/Exit
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
.0002
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002 −.0001
0
.0001
.0002
Effect on Cerebrovascular Diseases
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
0
.0002
1
Effect on All Other Diseases Residual
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0004
−.0002
0
.0002
.0004
Effect on Ischemic Heart Diseases
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0003
−.0002
−.0001
0
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
0
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
0
.0001
.0002
.0003
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
0
.0002
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0005
0
.0005
Online Appendix Figure 42: Effect of Hospital Entry and Exit that Affects the Distance to the
Closest Hospital on Mortality for Medicare Beneficiaries’ Top 15 Causes of Death
Effect on Other Diseases of the Heart
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Lung Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Influenza and Pneumonia
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Kidney Disease
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Breast Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 15 causes of death during the sample period ordered by the frequency. The plotted
coefficients are from a regression of changes in the outcome on two leads and one lag of the change in the
number of hospitals from entry and exit that affects the distance to the closest hospital and controls including
an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects.
The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes are all
in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are clustered
by HSA.
46
Effect on Non−Hodgkin’s Lymphoma
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Atherosclerosis
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Motor Vehicle Accidents
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Hypertension and Related Kidney Disease
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Cervical, Uterine, Ovarian Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Chronic Liver Disease and Cirrhosis
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Suicide
−2
−1
0
Year Relative to Entry/Exit
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001 −.00005
0
.00005
.0001
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
1
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
−1
0
Year Relative to Entry/Exit
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
−2
Effect on Hypertensive Heart Disease
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00004
−.00002
0
.00002
Effect on Urinary Tract Cancer
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
1
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0002
−.0001
0
.0001
−1
0
Year Relative to Entry/Exit
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
−2
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
Effect on Pancreatic Cancer
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00004 −.00002
0
.00002
.00004
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
Coefficients on Leads and Lags of Hospital Entry/Exit
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001
−.00005
0
.00005
.0001
Coefficients on Leads and Lags of Hospital Entry/Exit
−.0001 −.00005
0
.00005
.0001
Online Appendix Figure 43: Effect of Hospital Entry and Exit that Affects the Distance to the
Closest Hospital on Mortality for Medicare Beneficiaries’ Next 15 Causes of Death
Effect on Other Diseases of the Circulatory System
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Unclassified Abnormal Findings
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Leukemia
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Stomach Cancer
−2
−1
0
Year Relative to Entry/Exit
1
Effect on Peptic Ulcer
−2
−1
0
Year Relative to Entry/Exit
1
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the next 15 leading causes of death during the sample period ordered by the frequency. The
plotted coefficients are from a regression of changes in the outcome on two leads and one lag of the change
in the number of hospitals from entry and exit that affects the distance to the closest hospital and controls
including an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed
effects. The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes
are all in per-beneficiary units. Error bars show the 95-percent confidence interval. Standard errors are
clustered by HSA.
47
Online Appendix Figure 44: Effect of Hospital Entry and Exit on Mortality by Cause of Death Net Entry in the HSA
Top 15 Causes of Death
Ischemic heart diseases
All other diseases Residual
Other diseases of the heart
Cerebrovascular diseases
Chronic lower respiratory diseases
Lung cancer
Other cancers
Alzheimer’s disease
Influenza and pneumonia
Diabetes mellitus
Colorectal cancer
Kidney disease
All other accidents and adverse effects
Prostate Cancer
Breast cancer
−.0002
−.0001
0
.0001
Coefficient on Hospital Entry/Exit
Next 15 Causes of Death
Pancreatic cancer
Hypertensive heart disease
Other diseases of the circulatory system
Urinary tract cancer
Hypertension and related kidney disease
Unclassified abnormal findings
Non−Hodgkin’s lymphoma
Cervical, Uterine, Ovarian Cancer
Leukemia
Atherosclerosis
Chronic liver disease and cirrhosis
Stomach cancer
Motor vehicle accidents
Suicide
Peptic ulcer
−.00004 −.00002
0
.00002 .00004
Coefficient on Hospital Entry/Exit
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 30 causes of death during the sample period ordered by the frequency. The plotted
coefficients are from a regression of changes in the outcome on two leads and one lag of the change in the
number of hospitals from entry and exit within the beneficiary’s HSA and controls including an indicator for
gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects. The regressions are
weighted by the number of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary
units. The plotted coefficients are the linear combination of the sum of the on impact effect plus the oneyear lag minus the sum of the one- and two-year leads. Error bars show the 95-percent confidence interval.
Standard errors are clustered by HSA.
48
Online Appendix Figure 45: Effect of Hospital Entry and Exit on Mortality by Cause of Death Net Entry in the County
Top 15 Causes of Death
Ischemic heart diseases
All other diseases Residual
Other diseases of the heart
Cerebrovascular diseases
Chronic lower respiratory diseases
Lung cancer
Other cancers
Alzheimer’s disease
Influenza and pneumonia
Diabetes mellitus
Colorectal cancer
Kidney disease
All other accidents and adverse effects
Prostate Cancer
Breast cancer
−.0002
−.0001
0
.0001
Coefficient on Hospital Entry/Exit
Next 15 Causes of Death
Pancreatic cancer
Hypertensive heart disease
Other diseases of the circulatory system
Urinary tract cancer
Hypertension and related kidney disease
Unclassified abnormal findings
Non−Hodgkin’s lymphoma
Cervical, Uterine, Ovarian Cancer
Leukemia
Atherosclerosis
Chronic liver disease and cirrhosis
Stomach cancer
Motor vehicle accidents
Suicide
Peptic ulcer
−.00005
0
.00005
.0001
Coefficient on Hospital Entry/Exit
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 30 causes of death during the sample period ordered by the frequency. The plotted
coefficients are from a regression of changes in the outcome on two leads and one lag of the change in the
number of hospitals from entry and exit within the beneficiary’s county and controls including an indicator
for gender, age fixed effects, race fixed effects, year fixed effects, and HSA fixed effects. The regressions are
weighted by the number of Medicare beneficiaries in each bin and the outcomes are all in per-beneficiary
units. The plotted coefficients are the linear combination of the sum of the on impact effect plus the oneyear lag minus the sum of the one- and two-year leads. Error bars show the 95-percent confidence interval.
Standard errors are clustered by HSA.
49
Online Appendix Figure 46: Effect of Hospital Entry and Exit on Mortality by Cause of Death Net Entry in the Hospital Service Area
Top 15 Causes of Death
Ischemic heart diseases
All other diseases Residual
Other diseases of the heart
Cerebrovascular diseases
Chronic lower respiratory diseases
Lung cancer
Other cancers
Alzheimer’s disease
Influenza and pneumonia
Diabetes mellitus
Colorectal cancer
Kidney disease
All other accidents and adverse effects
Prostate Cancer
Breast cancer
−.0004
−.0002
0
.0002
Coefficient on Hospital Entry/Exit
Next 15 Causes of Death
Pancreatic cancer
Hypertensive heart disease
Other diseases of the circulatory system
Urinary tract cancer
Hypertension and related kidney disease
Unclassified abnormal findings
Non−Hodgkin’s lymphoma
Cervical, Uterine, Ovarian Cancer
Leukemia
Atherosclerosis
Chronic liver disease and cirrhosis
Stomach cancer
Motor vehicle accidents
Suicide
Peptic ulcer
−.0001
0
.0001
.0002
Coefficient on Hospital Entry/Exit
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 30 causes of death during the sample period ordered by the frequency. The plotted
coefficients are from a regression of changes in the outcome on two leads and one lag of the change in the
number of hospitals from entry and exit within the beneficiary’s hospital service area and controls including
an indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects.
The regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes are
all in per-beneficiary units. The plotted coefficients are the linear combination of the sum of the on impact
effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars show the 95-percent
confidence interval. Standard errors are clustered by hospital service area.
50
Online Appendix Figure 47: Effect of Hospital Entry and Exit on Mortality by Cause of Death Net Entry within 10 Miles of Beneficiary
Top 15 Causes of Death
Ischemic heart diseases
All other diseases Residual
Other diseases of the heart
Cerebrovascular diseases
Chronic lower respiratory diseases
Lung cancer
Other cancers
Alzheimer’s disease
Influenza and pneumonia
Diabetes mellitus
Colorectal cancer
Kidney disease
All other accidents and adverse effects
Prostate Cancer
Breast cancer
−.0004
−.0002
0
.0002
Coefficient on Hospital Entry/Exit
Next 15 Causes of Death
Pancreatic cancer
Hypertensive heart disease
Other diseases of the circulatory system
Urinary tract cancer
Hypertension and related kidney disease
Unclassified abnormal findings
Non−Hodgkin’s lymphoma
Cervical, Uterine, Ovarian Cancer
Leukemia
Atherosclerosis
Chronic liver disease and cirrhosis
Stomach cancer
Motor vehicle accidents
Suicide
Peptic ulcer
−.00005
0
.00005
.0001
Coefficient on Hospital Entry/Exit
Notes: Data cover the 1999-2008 period. The unit of analysis is the cohort-race-sex-zip code-county-year.
The plots show the top 30 causes of death during the sample period ordered by the frequency. The plotted
coefficients are from a regression of changes in the outcome on two leads and one lag of the change in
the number of hospitals from entry and exit within 10 miles of the beneficiary and controls including an
indicator for gender, age fixed effects, race fixed effects, year fixed effects, and zip code fixed effects. The
regressions are weighted by the number of Medicare beneficiaries in each bin and the outcomes are all
in per-beneficiary units. The plotted coefficients are the linear combination of the sum of the on impact
effect plus the one-year lag minus the sum of the one- and two-year leads. Error bars show the 95-percent
confidence interval. Standard errors are clustered by HSA.
51
Online Appendix Table 1: Effects of Hospital Entry and Exit in Rural Health Service Areas
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality
(5)
Mortality
(6)
Number of Hospitals
27
(3)
469
(68)
2,531
(842)
0.11
(0.21)
3.88
(8.40)
-0.09
(7.59)
Lag Number of Hospitals
-6
(2)
-49
(56)
156
(609)
0.09
(0.20)
-14.93
(9.90)
-12.09
(6.74)
On-Impact + First Lag
21
(4)
420
(70)
2,687
(1099)
0.20
(0.30)
-11.05
(9.41)
-12.18
(9.02)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
0.06
189
5,061
0.24
189
5,061
0.10
189
5,061
0.03
189
5,047
0.05
189
5,057
0.07
189
5,057
Note: Data cover the 1982-2010 period. Unit of analysis is HSA-Year. The sample is limited to HSAs where all of its
constituent counties have an urban population under 20,000 and do not border a metropolitan area. Numbers reported
in the Number of Hospitals rows are coefficients and standard errors (in parenthesis, clustered by HSA) from
regressions of changes in each of the dependent variables on the change in the number of hospitals from entry and
exit and its first lag, HSA fixed effects, year fixed effects, and demographic and employment controls (see section
4.1). The On-Impact + First Lag row is the sum of the number of hospitals coefficient and its first lag.
52
Online Appendix Table 2: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- Per Bed
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality
(5)
Mortality
(6)
Number of Hospitals
1.16
(0.14)
28
(5)
247
(42)
0.0007
(0.0002)
0.010
(0.005)
0.005
(0.003)
Lag Number of Hospitals
-0.10
(0.09)
-7
(4)
-56
(25)
0.0001
(0.0001)
-0.003
(0.006)
-0.001
(0.002)
On-Impact + First Lag
1.06
(0.16)
22
(6)
191
(49)
0.0007
(0.0002)
0.007
(0.006)
0.004
(0.003)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
0.24
805
21,516
0.15
805
21,516
0.29
805
21,516
0.02
805
21,494
0.05
805
21,508
0.17
805
21,508
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the Number
of Beds rows are coefficients and standard errors (in parenthesis, clustered by HSA) from regressions of changes in
each of the dependent variables on the change in the number of hospital beds from entry and exit and its first lag,
HSA fixed effects, year fixed effects, and demographic and employment controls (see section 4.1). The On-Impact +
First Lag row is the sum of the number of beds coefficient and its first lag.
53
Online Appendix Table 3: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- Outcomes in Logs
Dependent Variable:
Log Beds
(1)
Log Admissions
(2)
Log Inpatient Days
(3)
Log ALOS
(4)
Log Deaths
(5)
Number of Hospitals
0.049
(0.005)
0.028
(0.004)
0.035
(0.005)
0.0057
(0.0025)
0.0008
(0.0009)
Lag Number of Hospitals
-0.005
(0.003)
-0.004
(0.002)
0.000
(0.004)
0.0032
(0.0024)
-0.0007
(0.0014)
On-Impact + First Lag
0.045
(0.006)
0.025
(0.004)
0.035
(0.006)
0.0089
(0.0033)
0.0001
(0.0013)
HSA and Year FE
X
X
X
X
X
Additional Controls
X
X
X
X
X
0.03
805
21,516
0.04
805
21,516
0.03
805
21,516
0.02
805
21,494
0.08
805
21,508
R2
# HSAs
# Observations
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the Number
of Hospitals rows are coefficients and standard errors (in parenthesis, clustered by HSA) from regressions of changes
in each of the dependent variables on the change in the number of hospitals from entry and exit and its first lag, HSA
fixed effects, year fixed effects, and demographic and employment controls (see section 4.1). The On-Impact + First
Lag row is the sum of the number of beds coefficient and its first lag.
54
Online Appendix Table 4: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- Per Bed in Rural Areas
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality
(5)
Mortality
(6)
Number of Beds
0.78
(0.10)
16
(2)
87
(34)
0.0089
(0.0058)
0.145
(0.204)
-0.057
(0.135)
Lag Number of Beds
-0.16
(0.07)
-2
(1)
-6
(16)
0.0042
(0.0039)
-0.383
(0.201)
-0.346
(0.127)
On-Impact + First Lag
0.63
(0.16)
14
(3)
81
(43)
0.0132
(0.0072)
-0.238
(0.263)
-0.403
(0.133)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
0.07
189
5,061
0.25
189
5,061
0.10
189
5,061
0.03
189
5,047
0.05
189
5,057
0.07
189
5,057
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. The sample is limited to HSAs
where all of its constituent counties have an urban population under 20,000 and do not border a metropolitan area.
Numbers reported in the Number of Beds rows are coefficients and standard errors (in parenthesis, clustered by HSA)
from regressions of changes in each of the dependent variables on the change in the number of hospital beds from
entry and exit and its first lag, HSA fixed effects, year fixed effects, and demographic and employment controls (see
section 4.1). The On-Impact + First Lag row is the sum of the number of beds coefficient and its first lag.
55
Online Appendix Table 5: Effects of Entries versus Exits
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality
(5)
Mortality
(6)
-1,104
(8,351)
-0.03
(0.04)
-0.94
(3.30)
1.67
(1.89)
Number of Hospitals + First Lag:
Net Entries
43
1,250
(18)
(1,054)
Net Exits
89
(16)
2,067
(546)
14,459
(4,744)
0.06
(0.03)
-0.56
(1.10)
0.29
(0.40)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
Population Weighted
p-value
R2
# HSAs
# Observations
0.09
0.20
805
21,516
0.52
0.14
805
21,516
0.15
0.27
805
21,516
0.09
0.02
805
21,494
0.91
0.05
805
21,508
0.46
0.17
805
21,508
Note: Data cover the 1982-2010 period. Unit of observation is HSA-year. Numbers reported in the Number of
Hospitals rows are the sum the number of hospitals coefficient and it’s first lag from regressions of changes in each of
the dependent variables on the change in the number of hospitals and its first lag if there is net entry, the change in the
number of hospitals and its first lag if there is net exit, HSA fixed effects, year fixed effects, and demographic and
employment controls (see section 4.1). The p − value is from a test of the equality of the Net Entry and Net Exit
coefficients. Standard errors, in parenthesis, are clustered by HSA.
56
Online Appendix Table 6: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- In Levels
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Deaths
(5)
Number of Hospitals
100
(18)
1,982
(502)
16,065
(4,681)
0.03
(0.02)
4.96
(7.42)
HSA and Year FE
X
X
X
X
X
Additional Controls
X
X
X
X
X
1.00
805
22,357
1.00
805
22,357
1.00
805
22,357
0.54
805
22,339
1.00
805
22,352
R2
# HSAs
# Observations
Note: Data cover the 1983-2010 period. The unit of observation is the HSA-year. Numbers reported in the Number
of Hospitals rows are coefficients and standard errors (in parenthesis, clustered by HSA) from regressions of each of
the dependent variables on a running total of the change in number of hospitals due to entry and exit, HSA fixed
effects, year fixed effects, and demographic and employment controls (see section 4.1).
57
Online Appendix Table 7: Effect of Hospital Entry and Exit in Counties on Mortality
All Deaths
Mortality Rate
AMI Deaths
Age 0-64
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Number of Hospitals
2.03
(3.13)
0.04
(0.31)
-0.53
(0.57)
-0.04
(0.17)
-0.88
(1.18)
-0.16
(0.14)
12.14
(10.60)
0.83
(2.55)
Lag Number of Hospitals
-3.10
(3.36)
-0.38
(0.61)
-1.09
(0.74)
-0.18
(0.10)
-0.55
(0.94)
-0.46
(0.29)
-6.46
(11.51)
1.14
(2.92)
On-Impact + First Lag
-1.07
(3.22)
-0.34
(0.65)
-1.62
(0.82)
-0.22
(0.17)
-1.43
(1.09)
-0.62
(0.34)
5.68
(11.70)
1.97
(4.54)
HSA and Year FE
X
X
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
X
Dependent Variable:
X
Population Weighted
R2
# States
# Observations
0.02
49
68,880
0.07
49
68,880
X
0.02
49
57,026
0.02
49
57,026
Age 65+
X
0.01
49
55,628
0.03
49
55,628
X
0.03
49
67,163
0.07
49
67,163
Note: Data cover the 1982-2010 period. The unit of observation is country-year. Numbers reported in the Number of
Hospitals rows are coefficients and standard errors (in parenthesis, clustered by State) from regressions of changes in
each of the dependent variables on the change in the number of hospitals from entry and exit and its first lag, county
fixed effects, year fixed effects, and demographic and employment controls (see section 4.1). The On-Impact + First
Lag row is the sum of the number of hospitals coefficient and its first lag.
58
Online Appendix Table 8: Effect of Hospital Entry and Exit on Distance to the Hospital
Dependent Variable:
Miles to Hospital
(1)
Log Miles to Hospital
(2)
Admit w/in 10 Miles
(3)
Admit w/in 20 Miles
(4)
Number of Hospitals
-0.18
(0.14)
-0.0043
(0.0041)
0.0015
(0.0009)
0.0013
(0.0006)
Lag Number of Hospitals
-0.05
(0.09)
-0.0011
(0.0024)
0.0017
(0.0006)
0.0005
(0.0003)
On-Impact + First Lag
-0.23
(0.11)
-0.0053
(0.0027)
0.0032
(0.0011)
0.0018
(0.0006)
HSA and Year FE
X
X
X
X
Additional Controls
X
X
X
X
0.00101
805
13,883,838
0.00109
805
13,883,838
0.00900
805
13,883,838
0.00395
805
13,883,838
R2
# HSAs
# Observations
Note: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year. Numbers
reported in the rows are coefficients and standard errors (in parenthesis, clustered by HSA) from a regression of
changes in the dependent variable on the change in the number of hospitals from entry and exit in the beneficiary’s
HSA of residence, one lag of that variable, an indicator for gender, age fixed effects, race fixed effects, year fixed
effects, and zip code fixed effects (see section 4.1). The regressions are weighted by the number of Medicare
beneficiaries in each bin. The On-Impact + One Lag row displays the linear combination of the two coefficient
estimates.
59
Online Appendix Table 9: Effect of Hospital Entry and Exit on Hospital Service Areas
Dependent Variable:
Total
Expenditures
(1)
Expenditures
(2)
Number of Hospitals
8
(14)
-2
(7)
0.0007
(0.0006)
Lag Number of Hospitals
25
(15)
15
(8)
On Impact + First Lag
33
(14)
Zip Code and Year FE
Additional Controls
R2
# HSAs
# Observations
Acute Hospital
Admissions
Days
(3)
(4)
ALOS
(5)
Mortality
(6)
-0.0006
(0.0030)
-0.0114
(0.0061)
-0.00018
(0.00023)
0.0020
(0.0007)
0.0077
(0.0041)
-0.0086
(0.0059)
0.00001
(0.00016)
13
(13)
0.0027
(0.0011)
0.0071
(0.0055)
-0.0200
(0.0077)
-0.00017
(0.00016)
X
X
X
X
X
X
X
X
X
X
X
X
0.0021
3436
30,870,674
0.0015
3436
30,870,674
0.0013
3436
30,870,674
0.0016
3436
30,870,674
0.0016
3436
13,937,579
0.0308
3436
30,870,674
Note: Data cover the 1999-2011 period. The unit of analysis is the cohort-race-sex-zip code-county-year. Numbers
reported in the rows are coefficients and standard errors (in parenthesis, clustered by hospital service area) from a
regression of changes in the dependent variable on the change in the number of hospitals from entry and exit in the
beneficiary’s hospital service area of residence, one lag of that variable, an indicator for gender, age fixed effects, race
fixed effects, year fixed effects, and zip code fixed effects (see section 4.1). The regressions are weighted by the
number of Medicare beneficiaries in each bin. The On-Impact + One Lag row displays the linear combination of the
two coefficient estimates.
60
Online Appendix Table 10: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- By Hospital Size
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality Rate
(5)
Mortality Rate
(6)
Per-Bed Estimates, On-Impact + First Lag:
1st Quartile of Hospitals
2
(14)
483
(363)
933
(3,585)
-0.087
(0.090)
-2.149
(3.278)
0.774
(2.419)
2nd Quartile of Hospitals
16
(10)
778
(352)
2,011
(2,783)
0.018
(0.039)
-2.651
(2.948)
0.058
(1.636)
3rd Quartile of Hospitals
54
(11)
479
(349)
2,650
(4,423)
0.267
(0.075)
0.570
(2.422)
-0.802
(1.241)
4th Quartile of Hospitals
162
(29)
3,796
(944)
27,733
(8,752)
0.043
(0.018)
0.333
(0.947)
0.491
(0.368)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
X
0.22
805
21,516
0.15
805
21,516
0.28
805
21,516
0.02
805
21,494
0.05
805
21,508
0.17
805
21,508
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the Quartile
of Hospitals rows are coefficients and standard errors (in parenthesis, clustered by HSA) of the sum of the on-impact
and first lag of coefficients on the number of beds from regressions of changes in each of the dependent variables on
the change in the number of hospitals from entry and exit and its first lag in each quartile, HSA fixed effects, year
fixed effects, and demographic and employment controls (see section 4.1). Quartiles are ordered from smallest to
largest by number of beds.
61
Online Appendix Table 11: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- By Utilized Capacity
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality Rate
(5)
Mortality Rate
(6)
Per-Bed Estimates, On-Impact + First Lag:
High Capacity Utilization:
Entering Hospital’s Beds
0.34
(0.44)
0
(21)
-330
(174)
-0.0013
(0.0008)
-0.024
(0.055)
0.007
(0.033)
-1.21
(0.21)
-24
(8)
-268
(65)
-0.0006
(0.0001)
-0.008
(0.007)
-0.002
(0.003)
0.74
(0.33)
1
(28)
133
(148)
0.0009
(0.0009)
0.030
(0.057)
0.106
(0.052)
-0.66
(0.20)
-16
(10)
60
(47)
-0.0016
(0.0007)
0.002
(0.017)
-0.013
(0.007)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
X
X
0.05
805
21,487
0.17
805
21,487
Exiting Hospital’s Beds
Low Capacity Utilization:
Entering Hospital’s Beds
Exiting Hospital’s Beds
Population Weighted
R2
# HSAs
# Observations
0.24
805
21,495
0.15
805
21,495
0.30
805
21,495
0.01
805
21,491
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the Number
of Beds rows are coefficients and standard errors (in parenthesis, clustered by HSA) of the sum of the on-impact and
first lag of coefficients on the number of beds from regressions of changes in each of the dependent variables on the
number of beds of hospital entry or exit in markets with above or below median levels of capacity utilization and their
first lag, HSA fixed effects, year fixed effects, and demographic and employment controls (see section 4.1).
62
Online Appendix Table 12: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- By Beds Per Capita
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality Rate
(5)
Mortality Rate
(6)
Per Hospital Estimates, On-Impact + First Lag:
1st Quartile of Hospitals
0.91
(0.15)
5.9
(7.0)
114
(46)
0.0003
(0.0002)
0.010
(0.010)
0.012
(0.005)
2nd Quartile of Hospitals
0.85
(0.19)
22.7
(9.8)
192
(57)
0.0004
(0.0001)
0.014
(0.008)
0.007
(0.005)
3rd Quartile of Hospitals
1.58
(0.30)
25.2
(9.6)
287
(112)
0.0008
(0.0002)
0.010
(0.014)
0.014
(0.014)
4th Quartile of Hospitals
0.86
(0.29)
25.4
(14.3)
106
(74)
0.0019
(0.0008)
-0.025
(0.017)
-0.027
(0.013)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
X
0.24
805
21,495
0.16
805
21,495
0.29
805
21,495
0.01
805
21,491
0.05
805
21,487
0.17
805
21,487
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the rows are
coefficients and standard errors (in parenthesis, clustered by HSA) of the sum of the on-impact and first lag of
coefficients on the number of hospitals from regressions of changes in each of the dependent variables on the change
in the number of hospitals from entry and exit and its first lag for each quartile of the bed to population ratio, HSA
fixed effects, year fixed effects, and demographic and employment controls (see section 4.1). The quartiles are
ordered from the lowest to highest bed to population ratios.
63
Online Appendix Table 13: Effect of Hospital Entry and Exit on Capacity, Quantity, and Mortality
- By Entrant/Exit Market Share
Dependent Variable:
Beds
(1)
Admissions
(2)
Inpatient Days
(3)
ALOS
(4)
Mortality Rate
(5)
Mortality Rate
(6)
Per Hospital Estimates, On-Impact + First Lag:
1st Quartile Events
59
(28)
-19
(756)
-5,504
(8,515)
-0.027
(0.030)
-0.17
(1.55)
0.62
(1.09)
2nd Quartile Events
83
(35)
2,779
(1,087)
13,623
(9,277)
0.006
(0.020)
1.48
(1.23)
1.12
(0.75)
3rd Quartile Events
89
(18)
2,053
(877)
18,666
(5,963)
0.039
(0.027)
-0.73
(1.64)
-2.34
(0.92)
4th Quartile Events
102
(20)
2,460
(755)
19,679
(6,617)
0.182
(0.085)
-3.20
(3.01)
1.32
(1.54)
HSA and Year FE
X
X
X
X
X
X
Additional Controls
X
X
X
X
X
X
Population Weighted
R2
# HSAs
# Observations
X
0.20
805
21,516
0.14
805
21,516
0.27
805
21,516
0.02
805
21,494
0.05
805
21,508
0.17
805
21,508
Note: Data cover the 1982-2010 period. The unit of observation is the HSA-year. Numbers reported in the rows are
coefficients and standard errors (in parenthesis, clustered by HSA) of the sum of the on-impact and first lag of
coefficients on the number of hospitals from regressions of changes in each of the dependent variables on the change
in the number of hospitals from entry and exit and its first lag for each quartile of the bed to population ratio, HSA
fixed effects, year fixed effects, and demographic and employment controls (see section 4.1). The quartiles are
ordered from the lowest to highest bed to population ratios.
64