does diversity matter for health? experimental evidence from … · 2018-12-06 · pie chart....

Does Diversity Matter for Health?Experimental Evidence from Oakland

Marcella Alsan

Owen Garrick and Grant Graziani

Stanford Medical School and NBERBridge Clinical Research, UC Berkeley

December 2018

This Study

This study examines a recommendation of leading medical institutes, to diversify the physicianworkforce, on health behaviors. Consensus Statements

This Study

I Tests whether African-American men increase their take-up of preventive care whenrandomly assigned to an African-American male doctor.

MotivationRacial Health Disparities

I African-American men have the highest all-cause age-adjusted mortality of any majordemographic group in the U.S. (Ravenell et al. 2008). Pie Chart

I African-American men have the highest all-cause age-adjusted mortality of any majordemographic group in the U.S.(Ravenell et al. 2008). Pie Chart

I Most of the gap between black and white men is driven by cardiovascular disease, cancerand stroke (Harper et al. 2012, JAMA)

I While premature mortality from CVD and cancer somewhat preventable,African-American men are less likely to use preventive care. Prevention

I Lack of insurance does not fully account for utilization differences (Dunlop et al 2002).

I Although many factors influence demand for prevention (e.g. information, liquidity,self-control) medical studies have pointed to beliefs, including mistrust as a potentialfactor. (Powell Hammond et al. 2010).

I African-American men have the highest all-cause age-adjusted mortality of any majordemographic group in the U.S.(Ravenell et al. 2008). Pie Chart

I Most of the gap between black and white men is driven by cardiovascular disease, cancerand stroke (Harper et al. 2012, JAMA)

I While premature mortality from CVD and cancer somewhat preventable,African-American men are less likely to use preventive care. Prevention

I Lack of insurance does not fully account for utilization differences (Dunlop et al 2002).

I Although many factors influence demand for prevention (e.g. information, liquidity,self-control) medical studies have pointed to beliefs, including mistrust as a potentialfactor. (Powell Hammond et al. 2010).

Tuskegee Study of Untreated Syphilis in the Negro Male

I Leverage sharp timing along with other dimensions of variation (demographic,geographic/migratory) - find significant effects of TSUS disclosure on utilization,mortality and mistrust in short-run. (Alsan and Wanamaker, 2017). TSUS Event Study

Tuskegee Study of Untreated Syphilis in the Negro Male

I Leverage sharp timing along with other dimensions of variation (demographic,geographic/migratory) - find significant effects of TSUS disclosure on utilization,mortality and mistrust in short-run. (Alsan and Wanamaker, 2017). TSUS Event Study

Concordance

I Would black men’s demand for preventive care be higher with a racially concordantdoctor?

I Evidence of concordance effect in Labor markets Stoll, Raphael, and Holzer 2004; Giuliano,Levine, and Leonard 2009; Hjort 2014; Glover, Pallais, and Pariente 2017).

I Evidence of concordance effect in Education system (Ehrenberg Goldhaber and Brewer 1995;Dee 2004, 2005; Bettinger and Long 2005; Carrell, Page, and West 2010; Lusher, Campbell, and Carrell2018).

I Findings from Healthcare more mixed (meta-analysis by Meghani et al. 2009).

I May be due to methodological differences, convenient sample of people “in thesystem” (EHR), not focused on prevention. Concordance MEPS

This study uses a randomized design, recruiting subjects from the community and will focusspecifically on the demand for preventive care among African-American men.

Concordance

I Would demand for preventive care be higher with a racially concordant doctor?

I Evidence of concordance in Labor markets (Glover, Pallais, and Pariente 2017; Stoll, Raphael,and Holzer 2004; Giuliano, Levine, and Leonard 2009; Hjort 2014).

I Evidence of concordance in Education system (Ehrenberg 1995; Dee 2004, 2005; Bettinger andLong 2005; Carrell, Page, and West 2010; Lusher, Campbell, and Carrell 2018).

This study uses a randomized design, recruiting subjects from the community and will focusspecifically on preventive care for African-American men.

Overview of Study DesignTwo-stage ‘double-blind’ randomized design at the individual level

I Stage One: Ex AnteI Subject introduced to randomly assigned doctor via photo on tablet.I Subject selects preventive services via tablet.I Random subset of subjects also receive flu vaccination incentive.

I Stage Two: Ex PostI Subject interacts with doctor in person.I Subject revises service selection.I Subject receives services chosen from assigned doctor.

I Hypotheses TestedI Aversion to MD different race → Learning MD black via tablet ↑ demand (Ex Ante).I Better within-race pair interaction → Meeting with black MD ↑ demand (Ex Post).

Overview of Study DesignTwo-stage ‘double-blind’ randomized design at the individual level

I Stage One: Ex AnteI Subject introduced to randomly assigned doctor via photo on tablet.I Subject selects preventive services via tablet.I Random subset of subjects also receive flu vaccination incentive.

I Stage Two: Ex PostI Subject interacts with doctor in person.I Subject revises service selection.I Subject receives services chosen from assigned doctor.

I Hypotheses TestedI Aversion to MD different race → Learning MD black via tablet ↑ demand (Ex Ante).I Better within-race pair interaction → Meeting with black MD ↑ demand (Ex Post).

Preview of Results

I Ex Ante Results (photo only): No impact of doctor race.I Ex Post Results (in-person meeting): Large impact of doctor race.

I Strongest for invasive services.

I Mechanism?I Evidence suggests findings driven by better communication within concordant pair.I Test but find little support for alt. mechanisms such as other chars. correlated with race in

our sample of 14 doctors (i.e. quality, effort) outliers or discrimination.

Preview of Results

I Mechanism?I Evidence suggests findings driven by better communication within concordant pair.

I Test but find little support for alt. mechanisms such as other chars. correlated with race inour sample of 14 doctors (i.e. quality, effort) outliers or discrimination.

Preview of Results

Outline

1. Experimental Design

2. Conceptual Framework

3. Empirical Framework

4. Participation and Balance

5. Results

6. Mechanisms

7. External Validity and Valuation

Outline

1 Experimental Design

2 Conceptual Framework

3 Empirical Framework

4 Participation and Balance

5 Results

6 Mechanisms

7 External Validity and Valuation

Experimental Design Overview

Did not redeem coupon Redeemed coupon

Recruitment

Randomization

Ex Ante Choice

Non-black doctor Black doctor

BMI BP DIA CHO FLU

PATIENT INTERACTS WITH DOCTOR IN PERSON

SUBJECT FEEDBACK

Ex Post Choice

PATIENT SEES DOCTOR PHOTO ON TABLET

$5 $10None $5 $10

Recruitment

I Black men recruited from ˜20 barbershopsand two flea markets around the East Bay.

I Individuals who completed baseline survey(regarding health and demographics)received voucher for free haircut.

I Given a coupon for free health screening.

I Uber donated ride-sharing services.

Coupon

RECRUITMENT

DID NOT REDEEM COUPON REDEEMED COUPON

RANDOMIZED AT CLINIC

Ex Ante Choice –

Select Services

NON-BLACK DOCTOR BLACK DOCTOR

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Redeem Coupon at Clinic

I To facilitate our experiment, set up aclinic.

RECRUITMENT

DID NOT REDEEM COUPON REDEEMED

COUPON

Ex Ante Choice –

Select Services

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Redeem Coupon at Clinic

I To facilitate our experiment, set up aclinic.

I Hired 14 doctors and about 25field/clinic staff.

I Double blind. Oakland Men’s HealthDisparities Project.

I Doctors role: provide information onbenefits of preventive care and provideselected services.

I Doctors instructed to encourage allpatients to obtain all services.

I Doctors compensated for shift- workedon their “off” Saturdays.

RECRUITMENT

DID NOT REDEEM COUPON REDEEMED

COUPON

Ex Ante Choice –

Select Services

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Randomization

I Subjects entered clinic if had validcoupon, received a wait slip.

I Escorted to waiting room until numbercalled.

I Escorted to private patient room.I Given incentive payment for showing up.I Received tablet which did in-form

randomization (SurveyCTO).

RECRUITMENT

RANDOMIZATION

Ex Ante Choice –

Select Services

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Ex Ante Stage

I Ex ante — patient introduced to assigneddoctor on tablet.

I Selects from following services:I Weight and heightI Blood pressureI Diabetes (required blood sample)I Cholesterol (required blood sample)I Flu vaccination (required

injection/incentivized, screen forcontraindication if selected )

I None of the above also an option

Tablet

RECRUITMENT

Ex Ante Choice –

Select Services

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Ex Post Stage

I Ex post — patient interacts with doctorin person.

I Revises choices.I Receives chosen services.I Fill out feedback form privately.I Escorted out of clinic.

RECRUITMENT

Ex Ante Choice –

Select Services

NONE $5 $10 NONE $5 $10

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice –

Revise Selection,

Receive Services

Design and Flow

Not Redeem (667) Redeem Coupon (707)

Recruitment (1,374)

Randomization (637)

Ex Ante Choice

Non-black doctor (324) Black doctor (313)

None (96)

BMI BP DIA CHO FLU

SUBJECT FEEDBACK

Ex Post Choice

$5 (106) $10 (111) $5 (96) $10 (108)

12 Not Self-identify Af-Am

02 Women

06 Missing Consent

50 Attrit

None (120)

Outline

5 Results

6 Mechanisms

Framework and Hypotheses Tested

I Develop a simple model to formalize hypotheses tested and facilitate interpretation of theresults.

I Patients may hold false beliefs about the value of preventive health benefits, b,discounting by βi (Pauly and Blavin 2008; Baicker, Mullainathan and Schwartzstein 2015). Beliefs

I Patients may have a strong aversion for doctors of a given race, dr , which is additive todisutility c and would manifest as differential take-up across doctor race ex ante (Becker1957).

I Doctors provide accurate information so subjects can update their beliefs upon interactingwith them. Yet whether information is effectively communicated can depend on socialdistance, δr , and would manifest as differential take-up across doctor race ex post (Tabellini2008).

I Patients may have a strong aversion for doctors of a given race, dr , which is additive todisutility c and would manifest as differential take-up across doctor race in the ex antestage (Becker 1957).

I Doctors (in our experiment) provide accurate medical information that would allowsubjects to update their beliefs upon interacting with their assigned MD. Yet whether suchinformation is effectively communicated can depend on social distance δr (Tabellini 2008).

I Patients may have a strong aversion for doctors of a given race, dr , which is additive todisutility c and would manifest as differential take-up across doctor race in the ex antestage (Becker 1957).

I Doctors (in our experiment) provide accurate medical information that would allowsubjects to update their beliefs upon interacting with their assigned MD. Yet whether suchinformation is effectively communicated can depend on social distance δr (Tabellini 2008).

Ex Ante Stage (0)Patient Assigned Doctor on Tablet

I Ex ante utility:

U0i = βi · b− c − drj . (1)

I where d is the non-negative psychic cost associated with the assignment of doctor j fromrace group {black,white}, as rj=b and rj=w β ∼ U [0, 1] and c + d ≤ b.

I Yields three cases based on whether aversion for different race, same race (internalizedracism) or absent/equal across all races. Ex Ante Stage - Cases

Ex Post Stage (1)Patient Interacts with Doctor in Person

I Doctors provide medical information to disabuse patients of false beliefs, ε∗i , defined asε∗i = (1− βi )b.

I Yet whether the information is credible/comprehensible may depend on social distance,∆rji , which reflects the difference between the race of assigned doctor j and race ofpatient i (i.e. | rj − ri |), with rj=b = ri=b = 1 and rj=w = 0.

I Ex post utility:

U1i = βi · b− c + (1− δ1

∆rij )ε∗i − drj . (2)

I Yields three cases based on whether communication facilitated across or within race pair.

Ex Post Stage - Cases

Summary of Predictions

I If there is aversion for doctors of a particular race (i.e. drj ) it will be reflected by choicesin the ex ante stage.

I Aversion could be augmented or offset by patient-doctor interaction in the second stage.

I If there is no differential effect ex ante, differences in the ex post stage will primarilyreflect differences in patient-doctor relation (i.e. δ1

∆rij ).

Outline

5 Results

6 Mechanisms

Empirical Framework: Model

Yi = α + β1 · 1BlackMDi + β2 · 1$5

i + β3 · 1$10i + Γ′Xi + εi (3)

where:

I i represents an individual subject

I Yi is the selection of preventive services

I 1BlackMDi is an indicator for black MD

I 1$5i is an indicator for a $5 incentive for the flu vaccination

I 1$10i is an indicator for a $10 incentive for the flu vaccination

I Xi are subject characteristics (included in some specifications)

I β1 is the ITT/TOT (given perfect compliance).

I multiple inference - control for FDR (Anderson 2008); make an “index”

Identification: E (εi |1Ti ) = 0

Outline

5 Results

6 Mechanisms

Participation

1,374 recruited

At Oakland barbershops & flea markets

Did not redeem clinic coupon

Redeemed clinic coupon

ParticipationDemographic Characteristics

Uninsured Age Married Unemployed≤ High School

EducationLow Income SSI/DI/UI

0.038 3.411*** -0.058*** 0.129*** 0.190*** 0.198*** 0.113***(0.027) (0.811) (0.022) (0.025) (0.029) (0.027) (0.024)

Mean 0.24 41.06 0.20 0.18 0.44 0.25 0.18Observations 1,074 1,241 1,201 1,176 1,141 1,171 1,198

Clinic Presentation

ParticipationHealth Characteristics

Self-Reported Health

Any Health Problem

Hospital Visits

ER VisitsNights

HospitalMedical Mistrust

Has Primary MD

-0.126*** 0.033 0.244 0.513*** -0.332 -0.011 -0.072**(0.025) (0.028) (0.469) (0.183) (0.746) (0.042) (0.029)

Mean 0.81 0.57 4.74 1.24 1.93 1.64 0.69Observations 1,148 1,241 935 1,031 1,041 1,232 1,096

Clinic Presentation

Results on Balance and Attrition

I Only 2 of 72 tests significant: non-black - $10 less likely to be insured and have lowerself-reported health.

I Attrition not different across arms.

Mean (S.D.)

Non-Black MD - $5

Non-Black MD - $10

Black MD - $0

Black MD - $5

Black MD - $10

Attrition 0.03 0.022 0.045 0.031 0.015 -0.029(0.18) (0.033) (0.034) (0.034) (0.031) (0.025)

I F-test: 1.715 (p-value: 0.129)

Full Table

Outline

5 Results

6 Mechanisms

Ex Ante Preventives - Non-Black Doctors

Non-Black MD

0 20 40 60 80 100

Blood Pressure (56%)

Non-Black MD

0 20 40 60 80 100

BMI (50%)

Non-Black MD

0 20 40 60 80 100

Diabetes (37%)

Non-Black MD

0 20 40 60 80 100

Cholesterol (35%)

Ex Ante Preventives - Non-Black Doctors

Non-Black MD

0 20 40 60 80 100

Flu Vaccine: Non-Incentivized (20%)

Non-Black MD

0 20 40 60 80 100

Flu Vaccine: Incentivized (43%)

Ex Ante Preventives - Black Doctors

Non-Black MD

Black MD

0 20 40 60 80 100

Blood Pressure (58%)

Non-Black MD

Black MD

0 20 40 60 80 100

BMI (52%)

Non-Black MD

Black MD

0 20 40 60 80 100

Diabetes (43%)

Non-Black MD

Black MD

0 20 40 60 80 100

Cholesterol (36%)

Ex Ante Preventives - Black Doctors

Non-Black MD

Black MD

0 20 40 60 80 100

Flu Vaccine: Non-Incentivized (17%)

Non-Black MD

Black MD

0 20 40 60 80 100

Flu Vaccine: Incentivized (43%)

Effects on Ex Ante Preventives

Blood Pressure

BMI Diabetes Cholesterol Flu

VaccinationShare of 4

0.025 0.023 0.050 0.010 -0.009 0.027(0.039) (0.040) (0.039) (0.038) (0.037) (0.030)

0.028 -0.059 0.085* 0.067 0.192*** 0.030

(0.048) (0.049) (0.048) (0.047) (0.043) (0.037)

-0.023 -0.009 0.028 -0.014 0.299*** -0.004

(0.048) (0.048) (0.047) (0.045) (0.043) (0.036)

$5 = $10 p -value 0.295 0.300 0.238 0.083 0.026 0.366

Control Mean 0.56 0.50 0.37 0.35 0.20 0.44

Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Effects on Ex Ante Preventives

Blood Pressure

0.025 0.023 0.050 0.010 -0.009 0.027(0.039) (0.040) (0.039) (0.038) (0.037) (0.030)

0.028 -0.059 0.085* 0.067 0.192*** 0.030

(0.048) (0.049) (0.048) (0.047) (0.043) (0.037)

-0.023 -0.009 0.028 -0.014 0.299*** -0.004(0.048) (0.048) (0.047) (0.045) (0.043) (0.036)

$5 = $10 p -value 0.295 0.300 0.238 0.083 0.026 0.366

Control Mean 0.56 0.50 0.37 0.35 0.20 0.44

Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Ex Post Preventives - Non-Black Doctors

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

Blood Pressure (56% → 72%)

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

BMI (50% → 60%)

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

Diabetes (37% → 42%)

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

Cholesterol (35% → 36%)

Ex Post Preventives - Non-Black Doctors

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

Flu: Non-Incentivized (20% → 18%)

Non-Black MD

Black MD

Non-Black MD

0 20 40 60 80 100

Flu: Incentivized (43% → 40%)

Ex Post Preventives - Black Doctors

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

Blood Pressure (58% → 82%)

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

BMI (52% → 76%)

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

Diabetes (43% → 63%)

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

Cholesterol (35% → 62%)

Ex Post Preventives - Black Doctors

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

Flu: Non-Incentivized (17% → 28%)

Non-Black MD

Black MD

Non-Black MD

Black MD

0 20 40 60 80 100

Flu: Incentivized (43% → 50%)

I Milkman et al. Vaccination rate 33% among employees at large firm - increased by 1.5ppt with prompt to write down date and 4 ppt. to write down both date/time.

Flu Vaccination - Ex Post, by MD Race

Effects on Ex Post Preventives

Blood Pressure

BMI Diabetes CholesterolFlu

0.107*** 0.161*** 0.204*** 0.256*** 0.100*** 0.182***(0.033) (0.036) (0.039) (0.038) (0.038) (0.029)

0.044 0.019 0.110** 0.065 0.221*** 0.059*(0.040) (0.045) (0.048) (0.048) (0.045) (0.035)

-0.026 -0.010 0.054 -0.004 0.219*** 0.003(0.041) (0.044) (0.047) (0.047) (0.044) (0.035)

$5 = $10 p -value 0.082 0.521 0.240 0.139 0.974 0.120Control Mean 0.72 0.60 0.42 0.36 0.18 0.53Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Effects on Ex Post Preventives

Blood Pressure

0.107*** 0.161*** 0.204*** 0.256*** 0.100*** 0.182***(0.033) (0.036) (0.039) (0.038) (0.038) (0.029)

0.044 0.019 0.110** 0.065 0.221*** 0.059*(0.040) (0.045) (0.048) (0.048) (0.045) (0.035)

-0.026 -0.010 0.054 -0.004 0.219*** 0.003(0.041) (0.044) (0.047) (0.047) (0.044) (0.035)

$5 = $10 p -value 0.082 0.521 0.240 0.139 0.974 0.120Control Mean 0.72 0.60 0.42 0.36 0.18 0.53Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Effect of Black MD for Non-Invasive Tests

BP BMI

Non-invasive

Effect of Black MD for Invasive Tests

BP BMI Diabetes Flu No $ Chol

Non-invasive

Invasive

Persuasion Rate

Ex Ante Ex Post

0.022 0.132***(0.034) (0.030)

-0.167*** -0.269***

(0.023) (0.024)

0.010 0.099***(0.034) (0.033)

Control Mean 0.53 0.66Observations 2,548 2,548

Black Doctor

Invasive Test

Black MD * Invasive Test

Ex Ante Ex Post

0.022 0.132***(0.034) (0.030)

-0.167*** -0.269***

(0.023) (0.024)

0.010 0.099***(0.034) (0.033)

Control Mean 0.53 0.66Observations 2,548 2,548

Black Doctor

Invasive Test

Black MD * Invasive Test

Effect of Black MD on Delta Preventives∆ = Shareex post

/∈ flu− Shareex ante

/∈ flu

-1 -.5 0 .5 1Delta Excluding Flu

Non-Black doctor Black doctor

Effect of Black MD on Delta Preventives

Blood Pressure

VaccineDelta Share

0.082** 0.138*** 0.154*** 0.246*** 0.108*** 0.155***(0.034) (0.033) (0.029) (0.032) (0.033) (0.022)

0.017 0.078* 0.024 -0.002 0.029 0.029

(0.043) (0.043) (0.036) (0.037) (0.039) (0.028)

-0.003 -0.001 0.026 0.010 -0.080* 0.008

(0.040) (0.038) (0.035) (0.038) (0.041) (0.026)

$5 = $10 p -value 0.646 0.053 0.975 0.762 0.010 0.449

Control Mean 0.16 0.11 0.05 0.01 -0.02 0.08

Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Robustness of Results

I Fixed Effects and Alternative Samples Alt. Fixed Effects and Samples

I Additional Subject-level Covariates Add Covariates

I False Discovery Rate Adjusted q-values Adjusted q-values

I Other Concordance Other Concordance

Interpretation of ResultsSample of Doctors

I Doctors received same pay and malpractice coverage, had same instructions and wereblind to fact that race concordance main aspect of study.

I Yet still concern that black doctors “better” in some way.I Show that doctors similar on observables in training, expertise etc.I Show that black doctors perform worse on non-black subjects (see Section on Mechanisms).I Similar study-specific measures of effort and quality across doctor arms (see Section on

Mechanisms).

I Results might be driven by outlier doctors given the sample size.I Test for robustness to leave-one-doctor-out (including best black and worst non-black).I Inference by permutation test randomly re-assigning race, wild-cluster bootstrap, clustering

with DoF adjustment.I Inference using robust standard errors (Athey et al. clarify when to cluster - when estimand is for

population - i.e. two-stage sampling, first sampling clusters than units within clusters, or when treatment isnot assigned at individual level).

I Yet still concern that black doctors “better” in some way.

I Show that doctors similar on observables in training, expertise etc.I Show that black doctors perform worse on non-black subjects (see Section on Mechanisms).I Similar study-specific measures of effort and quality across doctor arms (see Section on

Mechanisms).

I Results might be driven by outlier doctors given the sample size.

I Test for robustness to leave-one-doctor-out (including best black and worst non-black).I Inference by permutation test randomly re-assigning race, wild-cluster bootstrap, clustering

Mechanisms).

Similar on Observables: Rank of Medical School, Internists, Experience

Medical School Rank: Research

Medical School Rank: Primary Care

Internist Experience

Black Mean 24 23 0.67 15.17

Non-Black Mean 11 16 1.00 12.25

p -value .846 .559 .089 .741

Observations 14 14 14 14

Not Driven by Outliers: Leave-One-Out

Leave-One-Doctor-Out

I Leave out“best” black and “worst” non-black doctor: βinvasive1 0.11 (0.02) vs. 0.17 (0.02) with all.

Exact Test, Ex Post ShareI Randomly assign doctor race and plot β1’s.

I 214 = 16384 possibilities*

Treatment Coefficient, >97%

-.4 -.2 0 .2 .4Coefficient on Share Take-Up

Delta Exact Test Other Modes of Inference

Outline

5 Results

6 Mechanisms

Mechanisms

1. Communication.

2. Quality.

3. Effort.

4. Discrimination.

CommunicationExperimental Evidence

Subject Talk to MD

Doctor NotesAbout Subject

Non-Preventive Notes

0.100*** 0.111*** 0.089***(0.039) (0.038) (0.026)

-0.072 0.055 0.001

(0.048) (0.047) (0.033)

-0.085* 0.016 -0.016

(0.047) (0.046) (0.031)

Control Mean 0.35 0.32 0.08

Observations 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Subject Talk to MD

0.100*** 0.111*** 0.089***(0.039) (0.038) (0.026)

-0.072 0.055 0.001

(0.048) (0.047) (0.033)

-0.085* 0.016 -0.016

(0.047) (0.046) (0.031)

Control Mean 0.35 0.32 0.08

Black Doctor

$5 Incentive

$10 Incentive

Subject Talk to MD

0.100*** 0.111*** 0.089***(0.039) (0.038) (0.026)

-0.072 0.055 0.001(0.048) (0.047) (0.033)

-0.085* 0.016 -0.016(0.047) (0.046) (0.031)

Control Mean 0.35 0.32 0.08Observations 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Controls for Tests Doctor Comments

Further Evidence on Concordance from a Large Scale Survey

I To complement experimental results, we conducted a survey to understand concordancein the general population.

I 1490 self-identify black and white male respondents.

I Respondents matched educational characteristics of study sample (i.e. 50% had highschool education or less).

I Questions regarding World Health Organization (2003) domains of a responsive healthsystem — quality, access and communication.

I Asked respondents which doctor was most likely to meet certain characteristics.

Quality and ConcordanceResponses near 50%

give you appropriate treatment?

be the most qualified?

35 40 45 50 55 60 65 70 75

Percentage selecting MD of same race

Black Respondent White Respondent

Which doctor would...

Quality

Concordance Table

Communication and ConcordanceResponses shift right

understand your concerns best?

you be comfortable discussing concerns with?

35 40 45 50 55 60 65 70 75

Communication

Concordance Table

Heterogeneous EffectsInteraction among those who lack health care experience or are mistrustful of medical field

X = ER VisitsNo Recent Screening

Medical Mistrust

0.012** 0.141** 0.061**(0.006) (0.066) (0.031)

-0.0004 -0.030 -0.017(0.003) (0.040) (0.019)

0.133*** 0.123*** 0.056(0.028) (0.024) (0.053)

Black Doctor * X

Black Doctor

Medical Mistrust

0.012** 0.141** 0.061**(0.006) (0.066) (0.031)

-0.0004 -0.030 -0.017(0.003) (0.040) (0.019)

0.133*** 0.123*** 0.056(0.028) (0.024) (0.053)

Black Doctor * X

Black Doctor

Medical Mistrust

0.012** 0.141** 0.061**(0.006) (0.066) (0.031)

-0.0004 -0.030 -0.017(0.003) (0.040) (0.019)

0.133*** 0.123*** 0.056(0.028) (0.024) (0.053)

Black Doctor

Black Doctor * X

Demographics Hassle Costs Null Heterogeneity Observational Evidence Communication

Utilization and Concordance in MEPSAdult Male Sample

Go To Doctor for Preventive Care

Doctor Listens Understand Doctor

Black Respondent -0.008* -0.013 -0.015(0.005) (0.012) (0.014)

Black MD -0.012 -0.064** -0.066*(0.009) (0.025) (0.040)

Black Resp * Black MD 0.020** 0.082*** 0.080*(0.009) (0.026) (0.041)

Any Insurance 0.004 0.051*** 0.022(0.003) (0.010) (0.013)

Age Categories Yes Yes YesIncome Categories Yes Yes YesEducation Categories Yes Yes YesOther Ethnic/Race Groups Yes Yes YesObservations 32,189 22,118 7,649Years 2005–2015 2005–2015 2011–2015

Full table

Are Black Doctors Higher Quality/Exerting More Effort?

I Proxies for QualityI Rank of medical school, experience in adult primary care.I Legal action (or lack thereof).I Ratings (in and out of experiment).I Errors with testing devices.I Performance on subjects not meeting study criteria.I Predicting doctor fixed effects.

I Proxies for EffortI Time spent.I Targeting.

Rank of Medical School, Internists, Experience

Medical School Rank: Research

Medical School Rank: Primary Care

Internist Experience

Black Mean 24 23 0.67 15.17

Non-Black Mean 11 16 1.00 12.25

p -value .846 .559 .089 .741

Legal Action

I All doctors were vetted for malpractice suits by Stanford.I 94% of doctors have no paid malpractice claims (Studdert et al. 2016).

Ratings

Subject Rating of Experience

Subject Recommend MD

-0.019 -0.0005(0.048) (0.010)

0.029 0.009(0.065) (0.013)

0.078 0.010(0.056) (0.012)

Control Mean 4.80 0.99Observations 574 597

Black Doctor

$5 Incentive

$10 Incentive

Subject Comments

I Vitals.com ratings: black doctor average = 4.32; non-black average = 4.56I Net promoter scores: black doctor average = .987; non-black average = .980I No difference in error rates on devices.

Subjects Not Meeting Study CriteriaI Twelve subjects did not identify as black (still given services as were consented).I 14 percentage points less likely to choose services from black doctors.I This finding is more extreme than 93% of bootstrap coefficients.

Non-Criteria Sample, <93%

-.6 -.3 0 .3 .6Coefficient

Difference-in-differences

Doctor Fixed Effects

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

R-squared 0.033 0.062 0.417Observations 14 14 14

Black Doctor

Experience

Medical School Rank

Constant

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

Black Doctor

Experience

Medical School Rank

Constant

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

Black Doctor

Experience

Medical School Rank

Constant

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

R-squared 0.033 0.062 0.417

Black Doctor

Experience

Medical School Rank

Constant

I Race explains 85% of the R2

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

Black Doctor

Experience

Medical School Rank

Constant

I Equivalent to moving from the 80th ranked to the top ranked medical school.

0.161**(0.064)

0.002 0.001(0.004) (0.002)

-0.001 -0.002*(0.001) (0.001)

0.138*** 0.109 0.072(0.044) (0.065) (0.060)

Black Doctor

Experience

Medical School Rank

Constant

I Similar with LASSO using doctor chars (black MD coef +/- black MD* yrs experienceselected). Distribution

Proxy for Effort: Time Spent with Patient

I What does time spent represent?

I More tests (Rx effect)?

I Efficiency (or lack thereof)?

I Communication?

Effort: Time Spent with Patient

Length Visit,Minutes

Length Visit,Test Controls

4.384*** 1.016(0.897) (0.718)

3.275*** 0.973(1.126) (0.887)

0.617 -0.369(1.088) (0.861)

Black Doctor

$5 Incentive

$10 Incentive

Effort: Time Spent with Patient Similar Once Controlling for Main Effect

Length Visit,Minutes

Length Visit,Test Controls

4.384*** 1.016(0.897) (0.718)

3.275*** 0.973(1.126) (0.887)

0.617 -0.369(1.088) (0.861)

Black Doctor

$5 Incentive

$10 Incentive

Effort: No Evidence of Targeting by Black Doctors

0.039 0.024 -0.016 -0.160 -0.154 0.006(0.088) (0.090) (0.075) (0.184) (0.192) (0.140)

0.018 0.047 0.030 0.031 -0.015 -0.046(0.062) (0.062) (0.048) (0.129) (0.129) (0.095)

-0.022 0.234*** 0.256*** 0.058 0.202*** 0.144***(0.076) (0.078) (0.065) (0.043) (0.043) (0.032)

Observations 620 620 620 627 627 627

At-Risk, Cholesterol At-Risk, Diabetes

Black Doctor * X

Black Doctor

Subject values

Discrimination: No Evidence of Differing Thresholds

0.039 0.024 -0.016 -0.160 -0.154 0.006(0.088) (0.090) (0.075) (0.184) (0.192) (0.140)

0.018 0.047 0.030 0.031 -0.015 -0.046(0.062) (0.062) (0.048) (0.129) (0.129) (0.095)

-0.022 0.234*** 0.256*** 0.058 0.202*** 0.144***(0.076) (0.078) (0.065) (0.043) (0.043) (0.032)

Observations 620 620 620 627 627 627

At-Risk, Cholesterol At-Risk, Diabetes

Black Doctor * X

Black Doctor

I ‘Outcome’ test - if threshold to screen higher, then non-black doctors would be predictedto pick up more disease (see Chandra and Staiger 2017).

I No differences in means nor are there differences in distributions.

I Consistent with study doctors following instructions as well as importance of patientautonomy in decisions about preventive healthcare.

Subject values

Subject Discrimination? : Ex Ante ResultsEx Ante Results

Blood Pressure

0.025 0.023 0.050 0.010 -0.009 0.027(0.039) (0.040) (0.039) (0.038) (0.037) (0.030)

0.028 -0.059 0.085* 0.067 0.192*** 0.030

(0.048) (0.049) (0.048) (0.047) (0.043) (0.037)

-0.023 -0.009 0.028 -0.014 0.299*** -0.004

(0.048) (0.048) (0.047) (0.045) (0.043) (0.036)

$5 = $10 p -value 0.295 0.300 0.238 0.083 0.026 0.366

Control Mean 0.56 0.50 0.37 0.35 0.20 0.44

Observations 637 637 637 637 637 637

Black Doctor

$5 Incentive

$10 Incentive

Reviewing Evidence on Mechanisms1. Communication

I More likely to talk with doctors about other health and personal matters.I Concordance strongest for healthcare-related communication questions.I Black MD effect greater for those with ↑ mistrust who might be skeptical of information.

2. QualityI Rank of medical school: Black doctors’ schools ranked lower.I Experience: Black doctors slightly more (estimated) years in field, but less likely to be

internists.I Legal action (or lack thereof): No malpractice suits.I Ratings: High for both sets of doctors.I Error rates: Low for both sets of doctors.I Performance on subjects not meeting study criteria: Lower for black doctors.I Doctor fixed effects: Race most important.

3. EffortI Time with patient: similar across treatments after controlling for additional testing.I Targeting: no evidence of targeting by disease presence/severity.

4. DiscriminationI No differences in preventive selections ex ante.I Very high ratings for both sets of doctors ex post.

2. Quality

I Rank of medical school: Black doctors’ schools ranked lower.I Experience: Black doctors slightly more (estimated) years in field, but less likely to be

3. Effort

I Time with patient: similar across treatments after controlling for additional testing.I Targeting: no evidence of targeting by disease presence/severity.

4. Discrimination

I No differences in preventive selections ex ante.I Very high ratings for both sets of doctors ex post.

Outline

5 Results

6 Mechanisms

Sample and Population Characteristics

U.S., 2016Study

Sample

Age 43.21 43.04

≤ High School Education 0.58 0.63

Uninsured 0.17 0.28

Unemployed 0.07 0.31

Source: U.S. averages are from 2016 ACS

Disease Prevalence

33%41%

29%37%

35%33%

15%18%

Hypertension (BP)

Obesity (BMI)

High Cholesterol

Diabetes

0 10 20 30 40 50Prevalence of Condition (%)

White Male Pop. Black Male Pop. Black Male Sample

Back-of-the-Envelope Health Valuation

I Change prob. of flu vaccine take-up by same amount if give patient ≈ $5 or a blackdoctor.

I This valuation calculation neglects effect on other services.I In setting of misperceptions, demand curve questionable for welfare calculations.

I Use studies that estimate the value of preventive services to estimate health gainassociated with intervention. (Kahn et al. 2010, Dehmer et al. 2017).

I Calculations suggest intervention could lead to a 19% reduction in the black-white gap inmale mortality rates for cardiovascular disease.

I Does not take into account diseases not included in the study (e.g. screening for HIV,prostate cancer).

Issues with Back-of-the Envelope Approach

I Based off studies that assume those who screen positive obtain and followguideline-recommended care.

I Assumes supply of black doctors to treat black patients.

Issues with Back-of-the Envelope Approach

I Based off studies that assume those who screen positive obtain and followguideline-recommended care.

I Assumes supply of black doctors to treat black patients.

Non-Experimental Survey Respondents

understand your concerns best?

you be comfortable discussing concerns with?

be available near you?

35 40 45 50 55 60 65 70 75

Access

Concordance Table Consensus Statements

Conclusion

I Black men randomized to black male doctors increase their uptake of preventive care.

I Results seem to be driven by better communication during the patient-doctor interaction.

I Findings suggest policies that increase the supply of African-American doctors could helpnarrow racial health gaps.

Conclusion

Gap Largely Explained by Cancer and CVD

Cardiovascular(e.g. heart disease)

Cancers(e.g. lung, prostate)

Communicable(e.g. HIV)

Injuries(e.g. homicide)

Miscellaneous Diseases(e.g. Alzheimer's)

Motivation

Source: Harper et al. (2012) JAMA

TSUS Event StudyAlsan and Wanamaker (QJE 2017)

Motivation

Utilization in MEPSAdult Male Sample

(1) (2) (3) (4) (5)

Black Respondent -0.008*** -0.007*** -0.005*** -0.005** -0.005**(0.002) (0.002) (0.002) (0.002) (0.002)

Asian Respondent 0.002 0.002 0.003 0.003 0.003(0.003) (0.003) (0.003) (0.003) (0.003)

Hispanic Respondent -0.000 0.001 0.004* 0.004** 0.004**(0.002) (0.002) (0.002) (0.002) (0.002)

Age Categories No Yes Yes Yes YesInsurance No No Yes Yes YesIncome Categories No No No Yes YesEducation Categories No No No No YesObservations 76280 76280 76280 76280 76280

Motivation

Race/Ethnicity of Patients and Doctors in MEPS

White MD Black MD Hispanic MD Asian MDWhite Patient 0.851 0.017 0.039 0.093Black Patient 0.527 0.257 0.065 0.151Hispanic Patient 0.381 0.029 0.439 0.151Asian Patient 0.254 0.009 0.027 0.710

Concordance

I African-Americans make up 12% of population but are only 3.5% of physician workforce.

I 73% (42%) of black doctors seen by black men (women) are male (female).

I Sample includes individuals 18+. Other race is omitted.

Sex of Patients and Doctors in MEPS

Female Male

Female 0.34 0.66

Male 0.17 0.83

Doctor

Concordance

Coupon

Coupon for Free Men's Health Screening

• See a doctor about a free health screeningand receive $50

• Receive free health screening for:1. Diabetes2. Cholesterol3. Height and Weight (Body Mass Index)

4. Blood Pressure

Clinic Address:(See Map on back)

Clinic Hours:11am-5pm

Saturdays only (List dates here)

Subject ID

Redeem Coupon

Recruitment Photo

Redeem Coupon

Tablet Screenshots

Ex Post Stage Flow

Beliefs

I “Flu shot makes me sick.”

I “Fear of being experimented on.”

I Diagnosed with diabetes in the past but,“refused to believe it.”

I Nutritional or other remedies can ward off illness - no need for screening.Framework and Hypotheses Tested

Ex Ante Stage - CasesI Case I Ex Ante: d > 0 if rj=w and d = 0 otherwise

I Fraction of subjects that demand preventives will be strictly greater for those randomized toblack versus white doctors.

I Pr(βi >c+drj=w

b |rj=w ) = 1−(c+drj=w

b < 1− cb = Pr(βi >

cb |rj=b)

I Case II Ex Ante: d > 0 if rj=b and d = 0 otherwiseI Black men discriminate against doctors of their own race.

I Pr(βi >cb |rj=w ) > Pr(βi >

c+drj=b

b |rj=b).

I Case III Ex Ante: d = 0 ∀ rj or d > 0 ∀ rjI No aversion to doctors based on their race, or the same level of aversion to doctors

regardless of their race.

I Pr(βi >c+db |rj=w ) = Pr(βi >

c+db |rj=b).

I Pr(βi >cb |rj=w ) = Pr(βi >

cb |rj=b).

Ex Ante Stage (0)

Ex Post Stage - Cases

I Case I Ex Post: 1 =

{1 if ∆rji = 1

0 if ∆rji = 0and δ ∈ (0, 1)

I If patients self-identify as black, then minimizing social distance by pairing such patients withblack doctors dominates pairings with white doctors.

I E[U1|rj=w ] = b− c − δb2 < b− c = E[U1|rj=b].

I Case II Ex Post: 1 =

{0 if ∆rji = 1

1 if ∆rji = 0and δ ∈ (0, 1)

I White doctors are viewed as more credible sources of information than black doctors.

I E[U1|rj=w ] > E[U1|rj=b].

I Case III Ex Post: δ = 0 or δ = 1 for all rjI Either no discounting of information by social distance or the information is discounted fully

from both black or white doctors.

Ex Post Stage (1)

Inference: Exact Test, Delta Share

-.4 -.2 0 .2 .4Coefficient on Share Take-Up

Exact Test

Other Modes of Inference

Heteroskedastic robust SEs

Wild cluster bootstrap

Doctor-date cluster robust

0 .005 .01 .015 .02 .025p-value

Ex Post Share Delta Share Invasive

Exact Test

Balance TableMean (S.D.)

Non-Black MD - $5

Non-Black MD - $10

Black MD - $0

Black MD - $5

Black MD - $10

F-test p-value N

Self-Reported Health 0.72 -0.033 -0.181*** 0.007 -0.016 0.004 2.075 0.067 563(0.45) (0.066) (0.067) (0.065) (0.064) (0.063)

Any Health Problem 0.62 -0.026 0.036 -0.015 -0.025 -0.021 0.250 0.940 614(0.49) (0.068) (0.065) (0.069) (0.067) (0.066)

ER Visits 1.69 -0.149 0.867 -0.212 0.145 -0.391 1.336 0.247 511(3.54) (0.434) (0.609) (0.443) (0.558) (0.419)

Nights Hospital 1.20 -0.392 0.839 1.956 -0.214 0.230 1.332 0.249 511(3.52) (0.415) (0.734) (1.490) (0.466) (0.663)

Has Primary MD 0.63 -0.042 0.033 -0.059 0.008 -0.019 0.415 0.838 537(0.49) (0.074) (0.070) (0.073) (0.070) (0.071)

Medical Mistrust 1.61 0.162 -0.046 0.032 0.016 -0.034 0.979 0.430 611(0.74) (0.105) (0.100) (0.105) (0.105) (0.100)

Age 44.96 -1.051 -0.100 -0.261 -1.109 -0.495 0.109 0.990 620(14.76) (1.973) (2.001) (1.982) (2.048) (1.944)

Married 0.14 0.043 -0.037 0.069 -0.015 0.024 1.120 0.348 586(0.35) (0.052) (0.045) (0.055) (0.047) (0.050)

Unemployed 0.32 -0.045 -0.008 -0.051 0.008 0.025 0.394 0.853 570(0.47) (0.066) (0.066) (0.065) (0.065) (0.065)

High School Education 0.62 0.006 -0.006 -0.029 0.055 0.034 0.344 0.886 556(0.49) (0.070) (0.070) (0.072) (0.068) (0.068)

Low Income 0.47 -0.026 -0.033 -0.043 0.022 -0.042 0.258 0.936 571(0.50) (0.072) (0.071) (0.072) (0.070) (0.069)

Uninsured 0.22 0.042 0.146** 0.112 0.057 0.010 1.398 0.223 517(0.42) (0.066) (0.067) (0.070) (0.064) (0.062)

Attrition 0.03 0.022 0.045 0.031 0.015 -0.029 1.715 0.129 684(0.18) (0.033) (0.034) (0.034) (0.031) (0.025)

Attrition

Flu Vaccination - Ex Ante

Black Doctor

Non-Black Doctor

$10 $5 $0Incentive Amount

Ex ante flu

Flu Vaccination - Delta, by MD Race

Black Doctor

Non-Black Doctor

Ex post flu

Persuasion Rate

BP BMI Diabetes Flu No $ Chol

I f = 100 · yT−yCeT−eC ·1

1−y0(DellaVigna and Gentzkow 2010).

Invasive Tests

Persuasion Rate Comparison

Chol.Dia.

0 10 20 30 40Persuasion Rate (%)

DG Studies Oakland StudyInvasive Tests

Flu Vaccination - Ex Post, by MD Race

Black Doctor

Non-Black Doctor

Expost Last

Difference-in-Differences, Non-Criteria SubjectsI Delta invasive coefficient on black subject * black MD = .267

-.75 -.5 -.25 0 .25 .5 .75Coefficient on Share Take-Up

Non-Criteria

Alternative Fixed EffectsMain coefficients: βex ante

1 = 0.027, βex post1 = 0.182, βdelta

1 = 0.155

Ex Ante Ex Post Delta Ex Ante Ex Post Delta Ex Ante Ex Post Delta

0.036 0.191*** 0.154*** 0.032 0.178*** 0.146*** 0.035 0.184*** 0.149***(0.031) (0.029) (0.022) (0.030) (0.029) (0.022) (0.030) (0.029) (0.022)

0.027 0.062* 0.034 0.032 0.049 0.017 0.026 0.047 0.022(0.037) (0.035) (0.027) (0.036) (0.035) (0.027) (0.037) (0.036) (0.027)

-0.007 0.011 0.018 -0.005 -0.004 0.001 -0.011 -0.007 0.004(0.036) (0.034) (0.025) (0.036) (0.034) (0.025) (0.036) (0.035) (0.026)

Control Mean 0.44 0.53 0.08 0.44 0.53 0.08 0.44 0.53 0.08Observations 637 637 637 637 637 637 618 618 618

$5 Incentive

$10 Incentive

Reception Officer Study Date Recruitment Location

Black Doctor

Robustness

Alternative SamplesMain coefficients: βex ante

1 = 0.027, βex post1 = 0.182, βdelta

1 = 0.155

Ex Ante Ex Post Delta Ex Ante Ex Post Delta Ex Ante Ex Post Delta

0.023 0.176*** 0.153*** 0.016 0.177*** 0.161*** 0.031 0.179*** 0.148***(0.030) (0.028) (0.022) (0.031) (0.029) (0.023) (0.032) (0.030) (0.023)

0.038 0.066* 0.028 0.027 0.064* 0.038 0.033 0.070* 0.037(0.037) (0.035) (0.028) (0.038) (0.036) (0.028) (0.039) (0.037) (0.028)

-0.002 0.006 0.008 -0.009 0.005 0.014 -0.023 -0.005 0.018(0.036) (0.034) (0.025) (0.037) (0.035) (0.026) (0.038) (0.037) (0.026)

Control Mean 0.44 0.53 0.08 0.45 0.53 0.08 0.45 0.52 0.08Observations 651 651 651 623 623 623 578 578 578

$10 Incentive

All Subjects Without Assisted Subjects Strict Specification

Black Doctor

$5 Incentive

Robustness

Add CovariatesEx Ante Share, Excluding Flu

Ex Post Share, Excluding Flu

Delta Share, Excluding Flu

Delta Share, Including Flu

0.028 0.181*** 0.153*** 0.144***(0.029) (0.028) (0.022) (0.019)

0.030 0.056 0.026 0.026

(0.035) (0.035) (0.027) (0.024)

0.002 -0.001 -0.003 -0.020

(0.035) (0.034) (0.025) (0.022)

-0.003 -0.005 -0.002 -0.002

(0.006) (0.006) (0.005) (0.004)

0.00003 0.00007 0.00004 0.00002

(0.00007) (0.00007) (0.00005) (0.00005)

-0.106*** -0.060* 0.046* 0.030

(0.035) (0.033) (0.025) (0.022)

-0.166*** -0.093*** 0.073*** 0.053**

(0.033) (0.031) (0.025) (0.022)

0.021 -0.003 -0.023 -0.018

(0.035) (0.034) (0.026) (0.023)

-0.069** -0.107*** -0.038 -0.048**

(0.034) (0.032) (0.027) (0.023)

0.001 -0.017 -0.017 -0.021

(0.040) (0.038) (0.028) (0.025)

Month Fixed Effects Yes Yes Yes Yes

Control Mean 0.44 0.53 0.08 0.04

Self-Assessed Health

Has Primary MD

Uninsured

Black Doctor

$5 Incentive

$10 Incentive

Age Squared

≤ High School Education

Low Income

Robustness

Adjusting for False Discovery RateOutcome = βex post

0.107 0.161 0.204 0.256 0.100 0.182{0.001} {<0.001} {<0.001} {<0.001} {0.008} {<0.001}{0.005} {0.001} {0.001} {0.001} {0.027} {0.001}

0.105 0.221{0.028} {<0.001}{0.080} {0.001}

0.219{<0.001}{0.001}

Share 4—

Black MD

Flu—

Black MD

Flu—

Diabetes—

Black MD

Diabetes—$5

Chol.—

Black MD

BMI—

Black MD

Robustness

Other Types of ConcordanceOutcome = Share of non-incentivized screenings

-0.011 -0.021 -0.003 0.017(0.023) (0.026) (0.052) (0.089)

-0.007 -0.080(0.044) (0.110)

0.153*** 0.153***(0.030) (0.025)

Control Mean 0.16 0.09 0.15 0.08Observations 620 620 556 556

X * Black Doctor

Black Doctor

Age, 10 Years Education

Robustness

Distribution of Doctor Fixed Effects: Delta Share

0 .1 .2 .3 .4Coefficient

Non-Black Doctor Black Doctor

Distribution of Doctor Fixed Effects: Invasive Services

-.1 0 .1 .2 .3Coefficient

Non-Black Doctor Black Doctor

Communication, Controls for Tests

(1) (2) (3)

Subject Talk to MD

0.095** 0.093** 0.070**(0.045) (0.044) (0.028)

-0.082 0.058 0.007(0.054) (0.054) (0.038)

-0.080 0.017 -0.049(0.054) (0.053) (0.035)

Control Mean 0.35 0.32 0.08Observations 498 498 498

Black Doctor

$5 Incentive

$10 Incentive

Communication

Sample of Doctor Comments

I Black doctor commentsI “needs food, shelter, clothing, job”I “‘flu shot makes you sick,’ he got one.”I “subject yelled at me but then agreed to get flu shot because I recommended it.”

I White doctor commentsI “?Homeless”I “Very appreciative. ‘I decided not to get any shots years ago’”I “Unhappy with what was told.”

Communication

Sample of Subject Comments

I Black men who saw black doctorsI “cool doctor”I “The doctor gave me some great information, friendly staff as well, thank you again.”I “very informative learning experience”

I Black men who saw white doctorsI “It was a great and fast experience, doctor was great as well.”I “very informative, very appreciated.”I “Great experience went by pretty quick. The doctor has provided me with a lot of great

information.”

Ratings

Concordance Table

Black MD White MD Race Match Black MD White MD Race Match Black MD White MD Race Match Race Match

0.350*** -0.055* 0.531*** -0.001 0.241*** -0.255*** -0.047(0.025) (0.030) (0.024) (0.029) (0.024) (0.029) (0.030)

0.273*** 0.479*** 0.175***

(0.029) (0.027) (0.030)

0.144***(0.023)

Mean 0.11 0.27 0.54 0.12 0.19 0.69 0.11 0.43 0.62 0.48R-squared 0.12 0.08 0.03 0.23 0.24 0.04 0.09 0.04 0.07 0.06Observations 1,490 1,490 1,490 1,490 1,490 1,490 1,490 1,490 1,490 2,980

Quality AccessCommunication

Black Respondent

White Respondent

Communication

Which MD most qualified? Which MD understands me? Which MD available near me?

Communication vs. Quality

Communication and Concordance Conclusion

Heterogeneous Effects: Hassle Costs

X = Long Wait Time High Congestion Long Commute

0.181*** 0.142*** 0.090*(0.055) (0.050) (0.046)

-0.005 0.014 0.023(0.030) (0.027) (0.026)

0.109*** 0.100*** 0.108***(0.029) (0.031) (0.028)

Black Doctor * X

Black Doctor

Heterogeneous Effects

Heterogeneous Effects: Demographics

X = Low IncomeHigh School Education

Younger than 40

0.087* -0.075 0.033(0.047) (0.049) (0.047)

0.053* 0.106*** -0.031(0.029) (0.028) (0.028)

0.112*** 0.192*** 0.136***(0.029) (0.037) (0.028)

Black Doctor * X

Black Doctor

Null Heterogeneous Effects

X = Medical Mistrust Legal Mistrust

0.061** 0.043(0.031) (0.036)

-0.017 -0.030(0.019) (0.021)

0.056 0.040(0.053) (0.091)

Observations 611 527

Black Doctor * X

Black Doctor

Observational Evidence CommunicationInformation is giving out - communication is getting through. Sydney J. Harris

I Cooper et al. (2016) record and score outpatient visits. Speech speed, patient centeredinterviewing and physician verbal dominance all scored the same. However, ethnicdiscordant visits were characterized by less social talk and lower global ratings of“physician positive affect” which includes interest, friendliness and sympathy.

I Elliot et al. (2015) studied doctors’ interactions with patients at the end of life usingstandardized patients. Although the doctors said similar things to both black and whiteactors posing as patients, they stood closer to the white patients, made more eye contact,and touched them more often.

Utilization and Concordance in MEPSAdult Male Sample

Doctor Listens Understand Doctor

Black Respondent -0.008* -0.013 -0.015(0.005) (0.012) (0.014)

Hispanic Respondent -0.002 -0.010 -0.012(0.005) (0.013) (0.016)

Asian Respondent -0.004 -0.020 -0.035(0.007) (0.019) (0.024)

Black MD -0.012 -0.064** -0.066*(0.009) (0.025) (0.040)

Black Resp * Black MD 0.020** 0.082*** 0.080*(0.009) (0.026) (0.041)

Hispanic MD -0.001 -0.028* 0.004(0.005) (0.016) (0.017)

Hisp Resp * Hisp MD 0.001 0.032* -0.004(0.006) (0.018) (0.025)

Asian MD -0.004 -0.015 0.002(0.005) (0.011) (0.014)

Asian Resp * Asian MD 0.003 -0.007 -0.002(0.008) (0.021) (0.028)

White Resp * White MD -0.003 -0.004 0.004(0.005) (0.013) (0.016)

Any Insurance 0.004 0.051*** 0.022(0.003) (0.010) (0.013)

Age Categories Yes Yes YesIncome Categories Yes Yes YesEducation Categories Yes Yes YesObservations 32,661 22,473 7,837Years 2005–2015 2005–2015 2011–2015

MEPS Concordance

Predicting Education/Race through Consumer Behavior

Education Race

Source: Bertrand and Kamenica (2018) Racial Concordance

Results from Clinic Encounter: BMI

Overweight Obese

10 20 30 40 50 60BMI (kg/m2)

Non-Black Doctor Black DoctorThreshold

Kolmogorov-Smirnov p-value: 0.638

Results from Clinic Encounter: Blood Pressure

Pre-Hyp. Hypertension Hypertensive Crisis

100 120 140 160 180 200Systolic Blood Pressure (mm Hg)

Results from Clinic Encounter: Cholesterol

High Choletserol

100 150 200 250 300 350 400Total Cholesterol (mg/dL)

Results from Clinic Encounter: Diabetes

Pre-Diabetes Diabetes

4 6 8 10 12Hemoglobin A1c (%)

Statements from Leading Medical Institutes

Introduction Conclusion

African-American Trends

Share of Medical School Graduates, Black

Share of Population, Black

2004 2006 2008 2010 2012 2014

Other trends

Source: AAMC, Census Bureau Population Estimates

Racial and Ethnic Differences in Medical School Representation

Hispanic

0 1 2 3 4Ratio of Share of Medical School Graduates to Share of Total Population, 2014

Source: AAMC, Census Bureau Population Estimates

Physician Shortages in Minority Communities

I In California, Black/Hispanic communities 4x more likely to be designated physicianshortage areas (PSA) regardless of community income.

Source: Komaromy et al. NEJM

Practice Location Choice

I African-American and Hispanic physicians more likely to practice in MUAs.

Latino

0 5 10 15 20 25 30Percentage of Physicians in Medically Underserved Areas, California

Source: Walker, Moreno, and Grumbach (2012)

Occupational Choice

I African-American and Hispanic physicians more likely to work in primary care.

Latino

0 10 20 30 40 50 60Percentage of Physicians Working in Primary Care, 2008 (%)

Conclusion

Source: AAMC (2008)

does diversity matter for health? experimental evidence from … · 2018-12-06 · pie chart....

Documents

ph201x: health and society racial/ethnic...

action plan to reduce racial and ethnic health disparities...

racial disparities in health care access among pediatric

hhs action plan to reduce racial and ethnic health...

racial and ethnic health disparities in miami-dade

racial disparities among black women in maternal health: a

reducing racial and ethnic disparities in health care

racial/ethnic and socioeconomic disparities of sars-cov-2...

hhs action plan to reduce racial and ethnic health...

racial and ethnic disparities in medicare: what the ... ·...

racial/ethnic disparities and health policy

racial and ethnic health disparities workforce diversity ·...

racial and ethnic disparities in u.s. health care: a...

racial disparities & breast cancer

racial and ethnic disparities in u.s. health care: a...

racial, ethnic, and gender disparities in health care in...

hhs action plan to reduce racial and ethnic health · pdf...

racial, ethnic, and socioeconomic disparities in health and...

racial and ethnic disparities in healthcare · eliminating...

racial and ethnic health disparities among people …