The Impact of Insurance Status on Hospital Treatment and Outcomes David Card, Carlos Dobkin and Nicole Maestas.

The Impact of Insurance Status on Hospital Treatment and

Outcomes

David Card, Carlos Dobkin and Nicole Maestas

Many Americans Are Uninsured

• 10% of Americans 62-64 are uninsured

• 28% of Minorities with less than a high school education are uninsured

• Understanding the impact of insurance on treatment and outcomes is important given the number of people uninsured.

LITERATURE

• Cross-Sectional– Levy and Meltzer (2001)

• Rand Experiment • Medicaid

– Currie and Gruber (1996)

• Medicare – Lichtenberg (2001) – Dow (2004)– McWilliams et al (2003)– Decker and Rapaport (2002)

• Car Accidents– Doyle (2005)

Goals of this Project

• Determine if Medicare affects the treatment people receive in the hospital

• Determine if Medicare affects outcomes for people admitted to the hospital

Data

• California Hospital Discharge Data 1992-2002

• Census of hospital discharges

Causal Model

,(1) ( )i i i i ij ji

y X f a X C B

Treatment Intensity or Outcome

Observable Characteristics

Age

Type of Insurance

i

i

i

ij

y

X

a

C

1 1 , 1(2 ) ( ) 1( 65)i i i i ia C X g a X Post v

2 2 2 , 2 2(2 ) ( ) 1( 65)i i i i ib C X g a X Post v

1 1 2 2 ,(3) y ( ) ( ) 1( 65)i i i i iX B B h a X Post w

, , 1 1 , 2 2 ,( ) ( ) ( ) ( )i i i i i i i ih a X f a X B g a X B g a X

1 1 2 2B B

1 1 2 2i i iw B v B v

Subsample SubsampleWith Without

Insurance Insurance T-ratioHave Health Insurance 100 0 --Female 51.2 58.3 4.60Married 74.2 58.0 10.65Working 41.2 38.4 1.80Poor (Family Income<Pov. Line) 5.8 19.3 11.37

White Non-Hispanic 83.1 62.6 13.94Black Non-Hispanic 8.4 13.3 4.69Hispanic 5.7 19.1 11.39Other Race Non-Hispanic 2.8 5.0 3.30

High School Dropout 21.3 50.8 19.26High School, No College 35.8 28.4 5.21Some College 21.6 13.1 7.824 Years College or More 21.4 7.7 15.19

Daily Smoker 13.2 20.5Any Exercise in Past 30 Days 71.5 60.9 9.49Overweight (BMI over 25.0) 67.4 69.2 1.19Obese (BMI over 30) 24.7 28.0 2.13

In Good, Very Good, Excellent Health 79.9 74.1 4.2

Why We Should Be Concerned About Omitted Variables Bias

Notes: Except as noted, means are based on data from pooled 1997-2001 NHIS. Percent with any exercise in past month, overweight, and obese are based on data from pooled 1998-2002 BRFSS.

Research Design

• Two main difficulties in getting causal estimates – The uninsured are systematically different from the

insured– Insurance impacts the probability of admission to the

hospital

• Our solutions– Focus on the abrupt change in insurance induced by

Medicare eligibility at age 65– Focus on severely ill patients that will be admitted to

the hospital regardless of insurance status

Figure 1: Health Insurance Coverage By Age and Source, 1996-2004 CPS

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75

Age

Fra

ctio

n w

ith C

ove

rag

e

Any Coverage

Medicare

Private

Medicaid

Military-Related

Microdata Cell-Level Data From 1996-2002 March CPS: from 1992- Married Family Family Family Moved to2001 NHIS: Spouse Income Income Income New House Employed Employed Present <$10,000 <$15,000 <$20,000 in Past Year (1) (2) (3) (4) (5) (6) (7)

All -0.44 -1.27 -0.91 -0.42 -0.78 -0.10 -0.48(0.96) (1.01) (0.70) (0.45) (0.63) (0.63) (0.33)

Table 2: Estimates of Discontinuities at Age 65 in Employment, Marriage, Family Income, and Mobility

Note: Table entries represent estimated coefficient of dummy for age 65 or older in models for outcome listed in column heading. Models in column (1) are fit to NHIS micro data; standard errors (in paretheses) are estimated assuming a cluster structure by age. Models in columns (2)-(7) are estimated using cell level data for ages 50-79 from March 1996-2002 CPS. Models include quadratic in age, fully interacted with a dummy for age 65 or older. Models in column (1) include year and region dummies.

Finding an Unselected Group of Admissions

• Since the probability of hospital admission depends on insurance status the case mix may change discretely at age 65.

• We focus on admissions that are not deferrable– First we restrict to unplanned admissions

through the emergency room– We then focus on a subset of non deferrable

conditions

Figure 2: Hospital Admission by Route of Admission (California 1992-2002)

5.6

5.7

5.8

5.9

6

6.1

6.2

6.3

6.4

6.5

6.6

60 61 62 63 64 65 66 67 68 69

Age in Months

Log(

Dai

ly A

dm

issi

ons)

ER Admissions

Non ER Admissions

ER Admission Regression Fitted Values

Non ER admission Regression Fitted Values

ER and Unplanned Non ER or PlannedRD at 65 0.0243 0.1187

0.0047 0.0047

R-Squared 0.807 0.869Observations 3,651 3,651Underlying Obs 1,742,377 1,723,886

Table 3A: Changes in Admissions and Insurance at Age 65 for California Hospital Admissions 1992-2002

Notes: These regressions are run on counts of admission by age where age is measured in days. The sample is restricted to people that are admitted from home to California hospitals between Jan 1, 1992 and November 30, 2002. The regressions include a second order polynomial in age fully interacted with a dummy for age greater than or equal to 65.

Identifying non Deferrable Admissions

• We focus on unplanned admissions through the Emergency Room

• Due to staffing constraints hospitals prefer to admit people on weekdays

• We compute the proportion of admissions for each condition (ICD-9) that occur on the weekend

• The study will focus on a subset of conditions that have a ratio near 2/7

Figure 3: Proportion of Admissions that Occur on the Weekend by ICD-9

0

10

20

30

40

50

60

70

0.17

10.

1780.

1860.

1940.

2010.

2090.

2170.

2240.

2320.

2390.

2470.

2550.

2620.

2700.

2780.

2850.

2930.

3010.

3080.

3160.

3230.

3310.

3390.

3460.

3540.

3620.

3690.

3770.

3850.

3920.

4000.

4080.

4150.

423

Proportion of Admissions on the Weekend

f(x)

All Admissions

ER Admissions

ER Admissions for ICD-9s With No Selection InFavor of Weekday Admissions

Figure 4: Admission Through the ER by Quartile of Weekend Proportion of ICD-9

4.2

4.3

4.4

4.5

4.6

4.7

4.8

4.9

5

5.1

5.2

60 61 62 63 64 65 66 67 68 69

Age in Months

Lo

g (

Da

ily A

dm

issi

on

s) f

or

Qu

art

ile w

ith Z

-Ra

tio (

0.9

6-

2.5

4)

an

d Q

ua

rtile

with

Z-R

atio

(2

.54

-6.6

2)

4.5

4.6

4.7

4.8

4.9

5

5.1

5.2

5.3

5.4

5.5

Lo

g (

Da

ily A

dm

issi

on

s) f

or

Qu

art

ile w

ith Z

-Ra

tio (

0-0

.96

) a

nd

Qu

art

ile w

ith Z

-Ra

tio (

ab

ove

6.6

2)

t-stat 0.96-2.54

t-stat 2.54-6.62

t-stat 0.96-2.54 Regression Fitted Values

t-stat 2.54-6.62 Regression Fitted Values

t-stat 0-0.96

t-stat Above 6.62

t-stat 0-0.96 Regression Fitted Values

t-stat above 6.62 Regression Fitted Values

Study Group

Lowest Weekend Proportion Group

Weekend t-stat > 6.62 Weekend t-stat 2.54-6.62 Weekend t-stat 0.96-2.54 Weekend t-stat < 0.96RD at 65 0.0322 0.0356 0.0271 0.0058

0.0099 0.0091 0.0093 0.0092

R-Squared 0.388 0.533 0.630 0.520Observations 3,651 3,651 3,651 3,651Underlying Obs 397,276 456,745 425,522 424,643

Table 3B: Changes in Admissions and Insurance at Age 65 for California Hospital Admissions 1992-2002

Notes: These regressions are run on counts of admission by age where age is measured in days. The sample is restricted to people that are admitted from home to California hospitals between Jan 1, 1992 and November 30, 2002. The regressions include a second order polynomial in age fully interacted with a dummy for age greater than or equal to 65.

Study Will Focus on ER Admissions With t-stat < .96

• These are the conditions that people are unable to defer

• No discernable change in admissions– Sorting around the discontinuity will be

observable

• Observable covariates are balanced

PercentRegression

Discontinuity65.52 0.095

[0.437]15.61 -0.093

[0.334]9.66 0.241

[0.272]14.49 0.774

[0.324]5.72 0.332

[0.214]4.97 -0.277

[0.200]3.69 0.035

[0.173]2.52 -0.044

[0.144]Sample Size 429,381 429,381

Changes in Characteristics of Study Group at age 65

Notes: Regression discontinuity estimates are by age in days with fully interacted quadratic polynomial in age. The RD estimates have been multiplied by 100 so that they are on the same scale as the percents. The SE are in brackets under the point estimates.

Respiratory failure

Acute myocardial infarction of other inferior wall first episodeAcute myocardial infarction of other anterior wall first episode

Intracerebral hemorrhage

Study Group

Obstructive chronic bronchitis With acute exacerbation

White

Hispanic

Black

ICD-9 AdmissionsLength of

Stay ProceduresList

Charges DiedObstructive chronic bronchitis with acute exacerbation 491.21 61,601 6.25 1.21 23,749 0.030Respiratory failure 518.81 24,376 13.79 3.72 65,315 0.228Acute myocardial infarction of other inferior wall first episode 410.41 21,211 7.29 5.17 52,732 0.071Acute myocardial infarction of other anterior wall first episode 410.11 15,752 7.92 5.37 57,065 0.106Intracerebral hemorrhage 431 10,749 18.03 3.66 62,051 0.338

Table 4: Five Most Common Conditions Included in the Study

Note: Length of stay, procedure count and hospital list charges are totals for all sequential hospital stays.

Measures of Treatment Intensity

• Length of Stay

• Diagnostic and Surgical Procedures Received

• Hospital List Charges

Figure 5A: Length of Stay in Days

0

1

2

3

4

5

6

7

8

9

10

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Leng

th o

f Sta

y in

Day

s

Age Over 65 0.37381 0.37275 0.36978[0.23644] [0.23653] [0.23390]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 7.964 7.964 7.964Observations 424,643 424,643 424,643R-squared 0.000 0.000 0.030

Length of Stay

Figure 5B: Counts of Procedures

2.3

2.35

2.4

2.45

2.5

2.55

2.6

2.65

2.7

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

cedu

re C

ount

s

Age Over 65 0.08904 0.08866 0.09780[0.03005] [0.02985] [0.02701]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 2.507 2.507 2.507Observations 424,643 424,643 424,643R-squared 0.000 0.010 0.200

Procedure Count

Figure 5C: Hospital List Charges in Logs

9.7

9.72

9.74

9.76

9.78

9.8

9.82

60 61 62 63 64 65 66 67 68 69 70

Age in Months

List

Cha

rges

in L

ogs

Age Over 65 0.02464 0.02489 0.02633[0.01078] [0.01060] [0.00910]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 9.754 9.754 9.754Observations 367,571 367,571 367,571R-squared 0.000 0.030 0.290

Log Charges

Hospital Transfers

• Federal and California State law require that hospitals admit and stabilize patients

• Hospitals may be discharging uninsured patients earlier than they would discharge an insured patient

Figure 6A: Proportion of Admissions Discharged to their Home

0.55

0.57

0.59

0.61

0.63

0.65

0.67

0.69

0.71

0.73

0.75

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

port

ion

Dis

char

ge H

ome

Age Over 65 -0.01525 -0.01503 -0.01681[0.00444] [0.00443] [0.00401]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 0.639 0.639 0.639Observations 424,643 424,643 424,643R-squared 0.010 0.010 0.190

Discharged Home

Figure 6B: Proportion With an Across Hospital Transfer

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

port

ion

With

an

Acr

oss

Hos

pita

l Tra

nsfe

r

Age Over 65 0.00219 0.00225 0.00354[0.00238] [0.00238] [0.00227]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 0.071 0.071 0.071Observations 424,643 424,643 424,643R-squared 0.000 0.000 0.090

Transferred to Another Hospital

Figure 6C: Proportion of Patients with a Within Hospital Transfer

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

port

ion

With

A W

ithin

Hos

pita

l Tra

nsfe

r

Age Over 65 0.00921 0.00913 0.00881[0.00190] [0.00190] [0.00186]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 0.046 0.046 0.046Observations 424,643 424,643 424,643R-squared 0.000 0.010 0.050

Transferred Within Hospital

Outcomes

• Readmission to the hospital

• Within hospital mortality

Figure 7: Readmission to Hospital Within 28 Days of Discharge

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

port

ion

with

a R

eadm

issi

on W

ithin

28

Day

s of

Dis

char

ge

Age Over 65 -0.00747 -0.00753 -0.00777[0.00325] [0.00325] [0.00321]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 0.141 0.141 0.141Observations 424,643 424,643 424,643R-squared 0.000 0.000 0.030

Readmission within One Month

Figure 8: Proportion of Patients Dying Within the Hospital

0

0.02

0.04

0.06

0.08

0.1

0.12

60 61 62 63 64 65 66 67 68 69 70

Age in Months

Pro

port

ion

Dyi

ng

Age Over 65 -0.00426 -0.00422 -0.00285[0.00252] [0.00252] [0.00232]

Year /Month/Sat/Sun N Y YRace and Gender N Y YCondition FE N N YMean Dep. Var. 0.081 0.081 0.081Observations 424,643 424,643 424,643R-squared 0.000 0.000 0.160

Died in Hospital

Summary of Results• No evidence that selection is a problem

– Covariates balanced– No change in counts of admission at age 65– Regressions not effected by inclusion of covariates– Pattern of outcomes not consistent with selection

• Find an increase in treatment intensity – 4.6% increase in LOS (not statistically significant)– 3.9% increase in procedures– 2.6% increase in list charges

• Find increase in transfer probabilities– 19% increase in within hospital transfers– 5% increase in across hospital transfers (not statistically significant)

• Outcomes– 5.5% reduction in readmissions to the hospital within a month– 3.5% decrease in within hospital mortality (not statistically significant)

Conclusion

• Medicare increases treatment intensity and probability of transfer within the hospital

• The increased treatment reduces probability of readmission