Mortality Among Very Low Birthweight Infants in Hospitals Serving Minority Populations Support Provided by: UCLA-Drew EXCEED Program (AHRQ), RWJ Minority.

Mortality Among Very Low Birthweight Infants

in Hospitals Serving Minority Populations

Support Provided by: UCLA-Drew EXCEED Program (AHRQ), RWJ Minority Medical Faculty Development Program, and UCLA-RCMAR Center (NIA)

Leo Morales, M.D., Ph.D.Assistant Professor, UCLA

AcademyHealthJune 7, 2004

Morales-2 03/18/04

Key Collaborators

• Jeanette Rogowski, Ph.D., University of Medicine and Dentistry of New Jersey and RAND

• Douglas Staiger, Ph.D., Dartmouth University

• Jeffery Horbar, M.D., The Vermont Oxford Network (VON)

• Joe Carpenter, M.S., VON

• Mike Kenny, M.A., VON

• Jeff Geppert, M.A., National Bureau of Economic Research

Morales-3 03/18/04

I. BACKGROUND

Morales-4 03/18/04

Hospital Characteristics and Patient Outcomes

• Worse outcomes are associated with:

– Rural hospitals (Kahn, 1994)

– Non-teaching hospitals (Kuhn, 1994; Polanczyk, 2002; Taylor, 1999; Kahn, 1994)

– For-profit hospitals (Hartz, 1989; Haas, 2003)

– Lower expenses per admission (Burstin, 1993)

– Minority-serving hospitals (Brennan, 1991)

– Low volume and lower level of care (Phibbs, 1996)

• Little is known about the relationship of minority-serving hospital status to infant mortality

Morales-5 03/18/04

Trends in Infant Mortality

• Overall, infant mortality is decreasing for black and white infants

• However, the disparity between black and white infant mortality remains constant and maybe increasing (MacDorman, 2002)

– Black infant mortality 14.1 per 1,000 live births

– White infant mortality 5.7 per 1,000 live births

• Eliminating the racial disparity in infant mortality is one of six target areas in the Health People 2010 initiative

Morales-6 03/18/04

Very Low Birthweight Infants

• Definitions

– Low birthweight (LBW): <2500 grams

– Very low birthweight (VLBW): <1500 grams

• Small but high risk infant population

– LBW infants account for 7.6% of live births but 66% of all infant deaths (MacDorman, 2002)

– VLBW infants account for 1.4% of live births but 52% of all infant deaths (MacDorman, 2002)

Morales-7 03/18/04

Research Questions

• Do VLBW infants treated by minority-serving hospitals have similar neonatal mortality as those treated by other hospitals?

• Do hospital characteristics and process of care variables explain differences in neonatal mortality between minority-serving hospitals and other hospitals?

• Are black and white infants treated by minority-serving hospitals at similar risk for neonatal mortality?

Morales-8 03/18/04

II. METHODS

Morales-9 03/18/04

Primary Data Source

• 1995-2000 Vermont Oxford Network (VON)– 332 hospitals– 40% of US hospitals with NICUs– 50% of VLBW infants in US

• Abstracted medical records– Mortality outcomes– Case-mix variables– Process of care

• Institutional survey of participating hospitals– NICU level of care

Morales-10 03/18/04

Additional Data Sources

• American Hospital Association Annual Survey of Hospitals

– Hospital characteristics

• 1990 United States Census

– Maternal income and education

Morales-11 03/18/04

Study Subjects

• VLBW infants between 500g and 1500g

– White infants (n= 49,132)

– Black (n=24,918)

• Inborn infants only

Morales-12 03/18/04

Outcome Variable

• Neonatal mortality

– Mortality in the first 28 days after birth

– Mortality ascertained through transfers until discharge home

Morales-13 03/18/04

Main Explanatory Variable:Hospital Minority-Serving Status

• Hospitals assigned to 1 of 3 categories based on the proportion of infants treated between 1995 and 2000 who were Black

• % VLBW black infants = VLBW black infants / VLBW black and white infants

Category Number of Hospitals

Proportion of Hospitals

<15% Black Infants 113 34%

15%-35% Black Infants 121 36%

>35% Black Infants 98 30%

Morales-14 03/18/04

Explanatory Variables:Case-Mix Variables

• Gestational age (+ gestational age squared)*

• Birthweight*

• Small for gestational age

• Congenital malformation

• Multiple birth

• Any prenatal care

• 1-minute APGAR

• Sex

• Race

• Vaginal delivery

• Maternal income and education based on census

Morales-15 03/18/04

Explanatory Variables:Hospital Variables

• Geography– Region– Urban setting of >1,000,000

• NICU characteristics– Level of care– Volume

• Hospital characteristics– Ownership– Teaching status– Percent Medicaid admissions– Expenditures per admission– Average maternal income and education based on census

Morales-16 03/18/04

Explanatory Variables:Process of Care Variables

• Indicator variables:

– Treatment with surfactants

– Treatment with antenatal steroids

Morales-17 03/18/04

Statistical Models

• Descriptive analysis– Infants by case-mix, hospital, and process of

care variables

• Stratified regression by race– %black + case-mix

• Pooled regression– Model 1: %black + case-mix– Model 2: %black + case-mix + hospital– Model 3: %black + case-mix + hospital +

process of care

Morales-18 03/18/04

Estimation Methods

• Maximum-likelihood logistic regression models

• Robust standard errors

• Clustering of infants within hospitals

• STATA 8.0

Morales-19 03/18/04

III. RESULTS

Morales-20 03/18/04

Infants by Case-Mix Variables

  All Infants(n=74,050)

Neonatal Infant Morality 11%

Birth Weight (grams) 1048.5

Gestational Age (weeks) 28.5

1-Minute APGAR Score 5.4

Male Sex 51%

Small for Gestational Age 21%

Multiple Birth 29%

Congenital Malformation 4%

Vaginal Delivery 38%

Maternal Black Race 34%

Had Prenatal Care 96%

Maternal Education (years) 12.41

Maternal Income ($1000s) 36.01

Antenatal Steroids 74%

Surfactants 60%

Morales-21 03/18/04

Infants by Hospital Variables

All Infants (n=74,050)

Minority Serving Status <15% 31% 15% to 35% 36% >35% 33%

Urban Hospital 53% Region

Northeast 19% Midwest 27% South 37% West 16%

Volume <40 Infants per Year 8% Level C NICU 28% Hospital Ownership

Government 11% Not For-Profit 83% For-Profit 7%

Member, Council of Teaching Hospitals 48% Percent Medicaid Admissions 16% Expense per Hospital Admission ($1000s) 11.7

Morales-22 03/18/04

Stratified Logistic Regressions:Neonatal Mortality on Case-Mix Variables

Note. Models include year dummies. *p<0.05 **p<0.01

White Infants Black Infants

Odds Ratio 95% CI Odds Ratio 95% CI

Minority-Serving Status

<15% Black Infants 1.00 1.00

15%-35% Black Infants 1.10 (0.97 – 1.27) 1.16 (0.91 – 1.47)

>35% Black Infants 1.30 (1.09 – 1.56)** 1.29 (1.01 – 1.64)*

Gestational Age 0.04 (0.04 – 0.53)** 0.04 (0.03 – 0.50)**

Gestational Age Squared 1.05 (1.05 – 1.05)** 1.05 (1.05 – 1.06)**

1-Minute APGAR Score 0.73 (0.71 – 0.74)** 0.75 (0.72 – 0.77)**

Small for Gestational Age 2.44 (2.14 – 2.77)** 2.28 (1.93 – 2.69)**

Multiple Birth 1.32 (1.20 – 1.45)** 1.22 (1.07 – 1.39)**

Congenital Malformation 16.50 (14.46 – 18.82)** 14.74 (11.48 – 18.9)**

Vaginal Delivery 1.36 (1.25 – 1.49)** 1.36 (1.21 – 1.52)**

Prenatal Care 0.94 (0.77 – 1.15) 0.85 (0.72 – 1.01)

Maternal Income 1.00 (1.00 – 1.01) 1.00 (0.99 – 1.01)

Maternal Education 1.01 (0.95 – 1.06) 0.99 (0.93 – 1.07)

Male 1.25 (1.17 – 1.34)** 1.28 (1.15 – 1.43)**

Morales-23 03/18/04

Pooled Regressions:Neonatal Mortality on Hospital and Case-Mix Variables

Model 1 Model 2 Model 3 Odds Ratio 95% CI Odds Ratio 95% CI Odds Ratio 95% CI Minority-serving Status

0-15% Black Infants 1.00 1.00 1.00 15-35 Black Infants 1.12 (0.98 - 1.27) 1.11 (0.98 - 1.27) 1.11 (0.98 - 1.27) >35% Black Infants 1.28 (1.10 - 1.50)** 1.25 (1.04 - 1.51)* 1.26 (1.04 - 1.52)*

Urban Setting 1.10 (0.98 - 1.23) 1.10 (0.97 - 1.24) Regional Location

Northeast 1.00 1.00 Midwest 1.09 (0.90 - 1.31) 1.09 (0.89 - 1.32) South 1.05 (0.90 - 1.23) 1.05 (0.90 - 1.24) West 1.21 (1.00 - 1.45)* 1.19 (0.98 - 1.44)

NICU Volume ?40 Infants 1.00 1.00 <40 Infants 1.29 (1.12 - 1.49)** 1.27 (1.10 - 1.46)**

Level of Care Levels A and B 1.00 1.00 Level C 0.86 (0.76 - 0.98)* 0.86 (0.76 - 0.98)*

Hospital Ownership Private Not-For-Profit 1.00 1.00 Private For-Profit 0.83 (0.67 - 1.02) 0.79 (0.65 - 0.95)* Government 1.07 (0.91 - 1.26) 1.07 (0.90 - 1.26)

Teaching Hospital 1.02 (0.91 - 1.15) 1.03 (0.91 - 1.17) % Medicaid Admissions 1.04 (0.55 - 1.97) 0.96 (0.51 - 1.84) Expenses per Admission ($1000s) 1.01 (0.99 - 1.02) 1.01 (0.99 - 1.03) Average Education (Hospital Level)

0.91 (0.80 – 1.03) 0.95 (0.83 - 1.09)

Average Income (Hospital Level) 1.00 (0.99 – 1.02) 1.00 (0.98 - 1.01) Treatment with Antenatal Steroids (Infant Level) 0.54 (0.50 - 0.58)** Treatment with Surfactants (Infant Level) 0.56 (0.50 - 0.62)**

Note. Models include case-mix model and year dummies. *p<0.05 **p<0.01

Morales-24 03/18/04

Thought Experiment-1

• What if black infants were treated by the three categories of hospitals we studied (e.g., <15% black, 15% to 35% black, >35% black) in the same proportions as white infants?

– Black infant mortality would drop by 8.5%

0%

20%

40%

60%

80%

100%

White Infants Black Ifants

<15%

15%-35%

>35%

Morales-25 03/18/04

Thought Expereiment-2

• What if neonatal mortality at hospitals where 15% or more of the treated infants were black were the same as hospitals where <15% of the infants treated were black?

– 10% lower for white infants

– 22% lower for black infants

Morales-26 03/18/04

IV. CONCLUSIONS

Morales-27 03/18/04

Conclusions

• Minority-serving hospitals had higher neonatal mortality than other hospitals

• The difference in neonatal mortality between minority-serving and other hospitals was not explained by the hospital variables or process of care variables

• Neonatal morality was similarly elevated for black and white VLBW infants treated by minority-serving hospitals

Morales-28 03/18/04

Implications

• Minority-serving hospitals may provide worse quality of care than other hospitals

• Hospital-level factors may be more important in understanding disparities in care than individual characteristics such as infant race per se, at least among VLBW infants

• Disparities in infant mortality nationally might be reduced by improving care for VLBW infants at minority-serving hospitals

Contact Information:

[email protected] or

310-794-2296

Morales-30 03/18/04

Potential Explanations for Results

• Staffing patterns

– Nurse-to-patient ratio

– Board-certified specialists

• Maternal characteristics: smoking, drug and alcohol use

– Mediated by infant severity of illness

– Infant characteristics in minority-serving similar to other hospitals

– Maternal income and education had no effect

• Unmeasured severity of illness

– VON risk adjustment ROC=0.88

– SNAP ROC=0.73 to 0.91

Morales-31 03/18/04

Study Generalizability

• Compared with all US hospitals, VON hospitals are:

– More likely to be private non-profit

– Teaching hospitals

– Children’s hospitals

– More NICU beds

• Compared with all VLBW infants in US, VLBW infants treated by VON hospitals are:

– Differed in terms of birthweight but not gestational age

Morales-32 03/18/04

Hospital Characteristics

All (n=74,050)

White (n=49,132)

Black (n=24,918)

Minority-Serving Status <15% Black Infants 31% 39% 8% 15% to 35% Black Infants 36% 43% 35% >35% Black Infants 33% 18% 57%

Urban Hospital 53% 48% 64% Region

Northeast 19% 19% 20% Midwest 27% 29% 23% South 37% 32% 48% West 16% 20% 10%

Volume<40 Admits per Year 8% 8% 8% Level C (Highest Level of Care) 28% 30% 24% Hospital Ownership

Government 11% 9% 14% Not For-Profit 83% 84% 80% For-Profit 7% 7% 6%

Member, Council of Teaching Hospitals 48% 46% 52% Medicaid Admissions 16% 15% 19% Expense per Admission ($1000s) 11.7 11.8 11.4

Morales-33 03/18/04

Secondary Analyses

• Do the effects of risk-adjustment vary by race?

– Insignificant case-mix*race interactions

• Do the effects of minority-serving status vary by race?

– Insignificant percent black*race interactions

• Do the effects of hospital variables vary by percentage black infants treated?

– Insignificant percent black*hospital characteristic interactions