Where Would You Go for Your Next Hospitalization? Kyoungrae Jung Penn State University Roger Feldman University of Minnesota Dennis Scanlon Penn State.

Where Would You Go for Your Next Hospitalization?

Kyoungrae JungPenn State University

Roger FeldmanUniversity of Minnesota

Dennis ScanlonPenn State University

1

Introduction

Providers vary in diverse dimensions of quality (clinical skills, interpersonal quality, outcomes of patients)

Consumers are likely to value each of these dimensions

It is often difficult for consumers to observe provider quality in health care markets

Improve consumer information for better decisions and better-performing health care markets

2

Introduction (II)

Recent efforts have focused on public release of comparative quality information

Often including clinical quality information Response has been small

Consumers may not value such information or they already know about it prior to public reporting

What types of quality information do consumers value and use in making health care decisions?

The answer can help to devise effective strategies to increase consumer information (Feldman et al., 2000; Harris & Buntin, 2008)

Few studies have examined this question

3

Research objective

To examine the effects of different dimensions of hospital quality in the context of a future hospital choice.

We focus on:

1) consumers’ perceptions of unobservable (to researchers) quality attributes, such as reputation

2) hospital clinical quality, whose indicators are often included in public reporting programs

3) consumers’ satisfaction ratings from their own recent experiences with hospitals

4

Literature on hospital choice

Consumers choose closer hospitals (Porell and Adams, 1995)

After release of report cards, consumers chose hospitals with mortality rates lower than expected (Mukamel et al., 2004/2005; Dranove and Sfekas, 2008)

Financial incentive for using “safer” hospitals has mixed effects on hospital choice (Scanlon et al., 2008)

5

Hospital quality and choice

Consumer’ perceptions about quality may play a significant role in hospital choice

Goal has been to obtain unbiased estimates of the impact of public quality information on choices

Research has relied on hospital fixed effects to control for consumers’ beliefs about unmeasured hospital quality

Captures multiple unmeasured quality attributes with a single variable

Can’t examine the relative contributions of different attributes to choice

6

Our approach

We utilize stated preference data to infer consumers’ perceptions about unobservable hospital attributes (e.g. reputation)

We estimate parameters representing consumers’ perceptions about several unmeasured hospital attributes

Allows us to measure the amount of each unobserved attribute offered by each hospital

Enables us to examine their relative contributions to consumer utility

7

Additional contribution

We introduce individual-level satisfaction ratings from their own experience to the hospital choice model

Certain important features of hospital quality can be evaluated only by experience

Consumers report they use experience to make health care decisions (Feldman et al., 2000; Schultz et al., 2001)

“Bad experience” or dissatisfaction did not motivate consumers to switch health plans (Abraham et al., 2006)

We estimate the impact of individual satisfaction ratings on hospital choice in terms of a driving time trade-off

8

Study setting and data Survey of employees at a large self-insured employer

Administered twice: April/May 2004 and Spring 2005 Random sample stratified by union status and recent

hospitalization Observations on 16 hospitals chosen by more than 15 people 969 hospitalized and 790 non-hospitalized people

Key variables from the survey Future hospital choice from a hypothetical question Stated preferences for four unobserved attributes: Overall

reputation, medical services, amenities, and OOP costs Future choice and preference data available for both users and

non-users Satisfaction with hospital for users on a 1 to 10 scale

9

Stated preferences

Example of reputation: “On a scale of 1 to 10, with 1 being not at all valuable and 10 being extremely valuable, please rate each item.”

“The next time you decide which hospital to use for inpatient services, how valuable would you find: The hospital's overall reputation?”

Out-of-pocket cost Specialty medical services offered (e.g. cardiac bypass

surgery) Amenities (e.g. private rooms and convenient parking)

10

Other data sources

2005 Hospital Quality Initiative (HQI)

Hospitals’ clinical quality scores released in April 2005

Not publicly available during our study period

Compliance with Leapfrog safety standards (CPOE, IPS)

Publicly available on the Leapfrog website; posted on the TPA website during the 2nd round of the survey

Mapquest.com: driving time

AHA: profit, teaching status

11

Conceptual Model Based on the expected utility theory of decision making

(1)

Rj - hospital j’s clinical quality

Eij – consumer i’s beliefs about hospital j’s unobservable attributes

Dij - driving time

Xj – observable hospital characteristics

(2)

Ej – hospital-specific beliefs before experience Sij – satisfaction rating Iij – indicator of use (1 if consumer i used hospital j; 0 otherwise) h – weight given to experiential signal

12

ijjijijjij XDERU

)()1( ijijijjij IIShEhE

Model (2)

By substitution,

(3)

A consumer will choose hospital j in the future if

(4) for all k ≠ j

Estimate equation (3) by conditional logit analysis

13

ikij UU

ijjijij

ijijjjij

XDIh

IhSEhRU

)1(

Estimation

Two-stage estimation process:

Stage 1: Estimate hospital-specific beliefs (E j) about unobserved hospital attributes, using stated preference data from non-hospitalized (naïve) people

Stage 2: Estimate equation (3) -- how different dimensions of hospital quality influence future choice among hospitalized people, using parameters of consumers’ beliefs obtained from first stage

14

Estimating consumers’ perceptions (1st stage) Based on approach developed by Harris & Keane (1999);

used by Harris, Schultz & Feldman (2002)

Estimate choice model for non-hospitalized people that includes interactions between preference weights and hospital dummies

Preference weights on reputation, medical services, amenities, and OOP cost are variables

Coefficients represent average perceived amounts of unmeasured attributes possessed by each choice, relative to reference hospital

15

Other explanatory variables (1st stage) Driving time

Teaching status; Profit status

Observed HQI quality score

Compliance with the Leapfrog safety standards

HQI score * 2005 survey

Compliance * 2005 survey

Compliance * 2005 survey * union

16

Estimation of future hospital choice (2nd stage) Estimate conditional logit model among hospitalized

people as function of:

Parameters of consumers’ beliefs about unobserved attributes

HQI score Satisfaction with hospital used Indicator for use of hospital Other covariates as in 1st stage Compliance* 2005*union*used non-compliant hospital

Standard errors based on bootstrapping

17

Descriptive statistics

Driving time: 41.2 min (24.5 min) HQI scores: 72.1 Preference weights among non-hospitalized people

Reputation: 8.41 Medical services: 8.78 Amenities: 6.62 Out-of-pocket cost: 7.55

Satisfaction rating (among users): 8.11 Intention to use another hospital (among users): 31.7%

18

1st stage model for non-users: Preference weights for unobserved attributes

19

Coefficients Standard Errors

Reputation

Hospital 11 22.720*** 7.601

Hospital 12 19.748** 10.266

Hospital 14 31.814*** 11.351

Medical services

Hospital 5 41.985*** 12.902

Out-of-pocket costs

Hospital 1 -25.361*** 8.610

Hospital 2 -15.301** 7.180

Hospital 9 -13.370** 6.429

Hospital 11 -14.938*** 5.373

**: p<0.05; **:p<0.01

2nd stage future choice model for usersSpecification: I

Measured Hosp. Char.

Driving time -0.074***

For-profit -1.985***

Teaching 0.587***

Compliance 0.304**

Observed quality

HQI score 0.105***

Perceptual parameters

Reputation

Medical services

Out-of-pocket cost

Experiential measures

Satisfaction rating

Use indicator

20

II

-0.079***

-0.094

-2.147***

-1.047***

0.046***

0.824***

1.148***

0.465***

Impact of HQI score (Model II)

Consumers perceive clinical quality as a distinctive feature of hospital quality

Consumers correctly infer clinical quality before public disclosure.

Marginal effect of HQI score:

4%-point increase in market share for 1 SD increase in score

Much smaller than those of beliefs about medical services (13% points) or reputation (18.4% points).

21

2nd stage future choice model for users

22

III

-0.044***

0.075

-1.945***

-0.815***

0.030

0.725***

0.833***

0.458***

0.817***

3.227***

Specification: I

Measured Hosp. Char.

Driving time -0.074***

For-profit -1.985***

Teaching 0.587***

Compliance 0.304**

Observed quality

HQI score 0.105***

Perceptual parameters

Reputation

Medical services

Out-of-pocket cost

Experiential measures

Satisfaction rating

Use indicator

II

-0.079***

-0.094

-2.147***

-1.047***

0.046***

0.824***

1.148***

0.465***

Our approach vs. Fixed effects

23

IIIHospital

Fixed Effects

Driving time -0.044*** -0.047***

Perceptual parameters

Reputation 0.725*** -

Medical services 0.833*** -

Out-of-pocket cost 0.458*** -

Experiential measures

Satisfaction rating 0.817*** 0.818***

Use indicator 3.227*** 3.245***

Large impacts of consumers’ perceptions, satisfaction, and use (Model IV)1. 1 SD increases in medical services and reputation would

increase a hospital’s market share from 20% to 33.3% and 31.6%, respectively

2. 1 SD increase in satisfaction (2.17 on 1-10 scale) would increase the hospital’s market share from 20% to 33%

Consumer would drive 18.6 minutes farther to use a hospital with 1 SD better satisfaction rating

3. Marginal effect of use: 64 %-points

24

25

Weights on physician recommendation

Admission throughemergency room

Top tertile Bottom Yes No

Perceptual parameters

Reputation 0.776*** 0.873*** 0.883*** 0.574***

Medical services 1.085*** 0.812*** 0.807*** 0.862***

Out-of-pocket costs 0.399*** 0.628*** 0.632*** 0.304***

Experiential measures

Satisfaction rating 0.794*** 1.036*** 0.883*** 0.772***

Use indicator 3.528*** 3.163*** 3.086*** 3.318***

Sensitivity Analysis

Summary of results: future hospital choice Naïve consumers perceive differences in reputation, medical

services, and OOP costs across hospitals

Large effects of consumers’ beliefs about unobservable attributes are consistent with a study on health plan choice

Consumers may already know about hospital clinical quality prior to public reporting; its contribution to hospital choice is small

Positive effect of satisfaction is inconsistent with a study of health plan choice

Large effect of prior use is consistent with prior literature

26

Limitations

Hypothetical future choices may differ from actual choices But we avoid potential bias associated with repeated

hospitalizations

Stated preference data were collected from non-hospitalized people during a survey window Did not control for “ever” use May not be as naïve as we think Data were obtained by prospective questions

We can only estimate average, hospital-specific beliefs We obtain average beliefs for each attribute Use indicator may partially capture individual heterogeneity

27

Discussion

Recent trend for “report cards” to include information about satisfaction CMS initiated Hospital CAHPS in 2008

Inexperienced consumers may turn to report cards that contain quality measures based on others’ experience

Consider publicizing information about hospital reputation or medical services May represent what consumers would like to see May increase consumers’ responses to “report cards.”

28

Discussion (II)

Efforts are needed to increase awareness and use of public quality information to overcome effects of “use” and consumers’ beliefs Employers’ initiatives

Ensure that physicians are informed about hospital quality and incorporate it in their recommendations

What are effective strategies to increase consumer information and improve performance of health care markets?

29