hospital price transparency: collusion, search, and ...msaeedi/transparency.pdf · provider price...

Hospital Price Transparency:Collusion, Search, and Equilibrium

Price

Elena Prager, Maryam Saeedi, and Robert Town

Northwestern, CMU, and UT Austin

Ashecon, June 24, 2019

Elena Prager, Maryam Saeedi, and Robert Town Hospital Price Transparency: Collusion, Search, and Equilibrium Price

Lack of Transparency in Health Care Prices

• With a few exceptions, health care prices are notobservable

• They cannot be ascertained even with substantial effort

• Health policy advocates point to this lack of pricetransparency

◦ An impediment to more price competition◦ Proposed Congressional legislation seeking to require

provider price transparency◦ If patients (or referring physicians) are able to compare the

prices, market forces will drive down provider prices




• They cannot be ascertained even with substantial effort• Health policy advocates point to this lack of price

transparency

◦ An impediment to more price competition◦ Proposed Congressional legislation seeking to require






• They cannot be ascertained even with substantial effort• Health policy advocates point to this lack of price

transparency◦ An impediment to more price competition◦ Proposed Congressional legislation seeking to require




Possible Effects of Increased Price Transparency

However, it may have no or the opposite intended effect:

• It also gives other competing providers and insurersinformation about prices.

◦ Changes in information sets can have complex impacts onequilibrium prices in bargaining settings

◦ Grennan and Swanson, (2018)

• Most insurance policies have very little or no co-insurance◦ Patients have little incentive to consider the prices when

selecting a hospital for care◦ Whaley, (2015), Whaley, Brown, Robinson, (2019)

• Low deductibles/ low coinsurance consumers might useprice information as a proxy for quality

◦ May choose a provider with a higher price◦ Hussey, Wertheimer, and Mehrotra (2013)


Possible Effects of Increased Price Transparency

However, it may have no or the opposite intended effect:• It also gives other competing providers and insurers

information about prices.◦ Changes in information sets can have complex impacts on

equilibrium prices in bargaining settings◦ Grennan and Swanson, (2018)

• Most insurance policies have very little or no co-insurance◦ Patients have little incentive to consider the prices when

selecting a hospital for care◦ Whaley, (2015), Whaley, Brown, Robinson, (2019)

• Low deductibles/ low coinsurance consumers might useprice information as a proxy for quality

◦ May choose a provider with a higher price◦ Hussey, Wertheimer, and Mehrotra (2013)


Introduction

• In this paper we have All-Payer Claims Databases(APCDs) from NH and Maine

• NH introduced a public website in 2007◦ price information for a range of common hospital

outpatient procedures◦ hospitals and insurers able to access competitors’ prices

• No contemporaneous regulatory changes for Maine duringthe same period athough they also introduced a pricetransparency portal much later (2015)


Introduction

• We are interested in seeing the impact of this pricetransparency on different parts of market

• We can see different patients with different levels ofco-insurance/deductibles

• Also, some procedures are available for patients to observedirectly and some are not

• The prices change for different insurance companies whichseems to be a function of their size

• Use these to identify the effect of price transparency


Relation to Literature

• Grennan & Swanson (2019)• Brown (2016, 2018)• Whaley (2015), Whaley, Brown, Robinson (2019)• Gowrisankaran, Nevo, Town (2014)


Data

• NH and Maine APCDs (2005-2015)◦ Inpatient and outpatient admissions◦ ID for patients◦ Name of facility, parent insurance company

• NH launched a public website in 2007◦ providers and patients could check prices on certain

outpatient procedures◦ mostly lab work◦ potentially patients can use the website to infer information

on prices of other procedures

• Insurance companies as well as hospitals could alsorequest to get complete pricing data


The Model

• Following GNT (2014)• Insurance and hospitals negotiate on prices using a

Nash-in-Nash framework• Patients get a health shock• Given characteristics and prices, they choose the facility


The Model

• The Nash-in-Nash framework• Maximize weighted value added to the hospital and

insurance company• Weights are proportional to relative bargaining power of

hospital/insurance company• We let these weights to vary across years

◦ This will help us understand the effect of the pricetransparency on bargaining power of insurance companiesvs. hospitals

• Also the demand function for the patients can change• This will lead to different willingness to pay from the

insurance company side


Preliminary Results

• Very concentrated market• Look at the impact of the policy on the top 3 vs top 6

insurance companies• Also the impact on price dispersion for biggest hospitals• The relationship between willingness to pay and prices


Preliminary Results

• Payer enrollment market shares (from NH insurancemarket reports):

◦ 2007: Anthem 57%, Cigna 23%, Harvard Pilgrim 17%◦ 2008: Anthem 55%, Harvard Pilgrim 25%, Cigna 15%◦ 2009: Anthem 51%, Harvard Pilgrim 25%, Cigna 16%◦ 2010: Anthem 44%, Harvard Pilgrim 24%, Cigna 22%◦ 2011: Anthem 44%, Cigna 25%, Harvard Pilgrim 22%


Preliminary Results

DRG weighted prices over time, Top 6 payers0

.000

05.0

001

.000

15D

ensi

ty

0 5000 10000 15000 20000Payer-provider pair mean DRG-weighted price

(winsorized at 95th percentile)

2005 20062007 20082009 20102011


Preliminary Results

DRG weighted prices over time, Top 6 payers0

.000

1.00

02.0

003

0.0

001.

0002

.000

3

0 5000 10000 15000 20000 0 5000 10000 15000 20000 0 5000 10000 15000 20000

1 - Anthem 2 - Cigna 3 - Harvard Pilgrim

4 - Aetna 5 - United 6 - MVP

2005 2006 2007 20082009 2010 2011

Den

sity

Payer-provider pair mean DRG-weighted price(winsorized at 95th percentile)

Graphs by ParentCompanyRank


Preliminary Results

Average DRG weighted prices over time, Top 3 vs. 4-680

0010

000

1200

014

000

Aver

age

DR

G W

eigh

ted

Pric

es

2004 2006 2008 2010 2012 2014Year

Top 3 4 to 6


Preliminary Results

Standard Deviation Over Time, Top 3 vs. 4-620

0025

0030

0035

0040

00St

anda

rd D

evia

tion

2004 2006 2008 2010 2012 2014Year

Top 3 4 to 6


Preliminary Results

Standard Deviation Over DRG Weighted Prices, Top 3 vs. 4-6.2

.25

.3.3

5St

anda

rd D

evia

tion

over

Ave

rage

DR

G P

rice

2004 2006 2008 2010 2012 2014Year

Top 3 4 to 6


Preliminary Results

Kurtosis, Top 3 vs. 4-62

46

8Ku

rtosi

s

2004 2006 2008 2010 2012 2014Year

Top 3 4 to 6


Preliminary Results

Average DRG weighted prices over time, Top 380

0010

000

1200

014

000

1600

0M

ean

DR

G W

eigh

ted

Pric

es

2004 2006 2008 2010 2012 2014Year

Anthem CignaHarvard Pilgrim


Preliminary Results

Standard Deviation Over Time, Top 315

0020

0025

0030

0035

0040

00St

anar

d D

evia

tion

2004 2006 2008 2010 2012 2014Year



Preliminary Results

Standard Deviation Over DRG Weighted Prices, Top 3.1

5.2

.25

.3.3

5.4

Stan

dard

Dev

iatio

n ov

er M

ean

2004 2006 2008 2010 2012 2014Year



Preliminary Results

Biggest 3 Hospitals, Average DRG weighted prices over time,Top 6

8000

1000

012

000

1400

016

000

1800

0M

ean

DR

G W

eigh

ted

Pric

e, T

op 6

Insu

ranc

e C

ompa

nies

2004 2006 2008 2010 2012 2014Year

Concortd DartmouthElliot


Preliminary Results

Biggest 3 Hospitals, Standard Deviation Over Time, Top 610

0020

0030

0040

0050

0060

00St

anda

rd D

evia

tion,

Top

6 In

sura

nce

Com

pani

es

2004 2006 2008 2010 2012 2014Year

Concortd DartmouthElliot


Preliminary Results

Biggest 3 Hospitals, Standard Deviation Over DRG WeightedPrices, Top 6

.1.2

.3.4

Stan

ard

Dev

iatio

n o

ver M

ean,

Top

6 In

sura

nce

Com

pani

es

2004 2006 2008 2010 2012 2014Year

Concortd DartmouthElliotElena Prager, Maryam Saeedi, and Robert Town Hospital Price Transparency: Collusion, Search, and Equilibrium Price

Willingness to Pay

• Estimate Willingness to Pay• Using the demand estimations• Assuming that insurance companies maximize a weighted

sum of patients’ utility minus cost


Willingness to Pay and Price

Relationship between WTP and Price, before and after policy

Anthem Cigna Harvard PilgrimWTP 0.44 18.29 -0.8

(0.25) (9.15) (4.16)WTP*Post Dummy 0.83 -13.34 5.48

(0.36) (9.32) (4.33)Year Fixed Effects Y Y Y

Constant 9101 3573 8319(678) (2481) (2782)

R2 0.35 0.59 0.56


Next Steps

• Bring in outpatient and Maine data. Allows for cleaneridentification

• Estimate full GNT model• Explore estimating a bargaining model with asymmetric

information• Counterfactuals with different benefit designs, market and

information structures.


hospital price transparency: collusion, search, and ...msaeedi/transparency.pdf · provider price...

Documents