Key Collaborators Mortality Among Very Low Birthweight Infants in Hospitals Serving Minority Populations • Jeanette Rogowski, Ph.D., University of Medicine and Dentistry of New Jersey and RAND • • • • • Leo Morales, M.D., Ph.D. Assistant Professor, UCLA 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 AcademyHealth June 7, 2004 Support Provided by: UCLA-Drew EXCEED Program (AHRQ), RWJ Minority Medical Faculty Development Program, and UCLA-RCMAR Center (NIA) Morales-2 03/18/04 Hospital Characteristics and Patient Outcomes I. BACKGROUND • 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 minorityserving hospital status to infant mortality Morales-3 03/18/04 Trends in Infant Mortality Morales-4 03/18/04 Very Low Birthweight Infants • Overall, infant mortality is decreasing for black and white infants • Definitions – Low birthweight (LBW): <2500 grams • However, the disparity between black and white infant mortality remains constant and maybe increasing (MacDorman, 2002) – Very low birthweight (VLBW): <1500 grams • Small but high risk infant population – 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 – 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) one of six target areas in the Health People 2010 initiative Morales-5 03/18/04 Morales-6 03/18/04 1 Research Questions II. METHODS • 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 minorityserving hospitals at similar risk for neonatal mortality? Morales-7 03/18/04 Primary Data Source Morales-8 03/18/04 Additional Data Sources • 1995-2000 Vermont Oxford Network (VON) • American Hospital Association Annual Survey of Hospitals – 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 – Hospital characteristics • 1990 United States Census – Maternal income and education Morales-9 03/18/04 Study Subjects Morales-10 03/18/04 Outcome Variable • VLBW infants between 500g and 1500g • Neonatal mortality – White infants (n= 49,132) – Mortality in the first 28 days after birth – Black (n=24,918) – Mortality ascertained through transfers until discharge home • Inborn infants only Morales-11 03/18/04 Morales-12 03/18/04 2 Main Explanatory Variable: Hospital Minority-Serving Status Explanatory Variables: Case-Mix Variables • 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% • • • • • • • • • • • 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-13 03/18/04 Explanatory Variables: Hospital Variables Morales-14 03/18/04 Explanatory Variables: Process of Care Variables • Geography • Indicator variables: – 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 – Treatment with surfactants – Treatment with antenatal steroids Morales-15 03/18/04 Statistical Models Morales-16 03/18/04 Estimation Methods • 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-17 03/18/04 • Maximum-likelihood logistic regression models • Robust standard errors • Clustering of infants within hospitals • STATA 8.0 Morales-18 03/18/04 3 Infants by Case-Mix Variables III. RESULTS All Infants (n=74,050) Neonatal Infant Morality 11% Birth Weight (grams) 1048.5 Gestational Age (weeks) 1 28.5 M inute 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-19 03/18/04 Morales-20 03/18/04 Stratified Logistic Regressions: Neonatal Mortality on Case-Mix Variables Infants by Hospital Variables All Infants (n=74,050) Minority Serving Status <15% 15% to 35% >35% Urban Hospital Region Northeast Midwest South West Volume <40 Infants per Year Level C NICU Hospital Ownership Government Not For-Profit For-Profit Member, Council of Teaching Hospitals Percent Medicaid Admissions Expense per Hospital Admission ($1000s) White Infants Odds Ratio 31% 36% 33% 53% Black Infants 95% CI Odds Ratio 95% CI Minority-Serving Status 19% 27% 37% 16% 8% 28% <15% Black Infants 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)** 14.74 (11.48 – 18.9)** Congenital Malformation Vaginal Delivery 11% 83% 7% 48% 16% 11.7 1.00 16.50 1.36 (14.46 – 18.82)** (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)** Note. Models include year dummies. *p<0.05 **p<0.01 Morales-21 03/18/04 Morales-22 03/18/04 Pooled Regressions: Neonatal Mortality on Hospital and Case-Mix Variables Model 1 Odds Ratio Minority-serving Status 0-15% Black Infants 1.00 15-35 Black Infants 1.12 >35% Black Infants 1.28 Urban Setting Regional Location Northeast Midwest South West NICU Volume ?40 Infants <40 Infants Level of Care Levels A and B Level C Hospital Ownership Private Not-For-Profit Private For-Profit Government Teaching Hospital % Medicaid Admissions Expenses per Admission ($1000s) Average Education (Hospital Level) Average Income (Hospital Level) Treatment with Antenatal Steroids (Infant Level) Treatment with Surfactants (Infant Level) 95% CI Model 2 Odds Ratio 95% CI Model 3 Odds Ratio 95% CI (0.98 - 1.27) (1.04 - 1.51)* (0.98 - 1.23) 1.00 1.11 1.26 1.10 (0.98 - 1.27) (1.04 - 1.52)* (0.97 - 1.24) (0.90 - 1.31) (0.90 - 1.23) (1.00 - 1.45)* 1.00 1.09 1.05 1.19 (0.89 - 1.32) (0.90 - 1.24) (0.98 - 1.44) 1.00 1.29 (1.12 - 1.49)** 1.00 1.27 (1.10 - 1.46)** 1.00 0.86 (0.76 - 0.98)* 1.00 0.86 (0.76 - 0.98)* 1.00 0.83 1.07 1.02 1.04 1.01 0.91 (0.67 - 1.02) (0.91 - 1.26) (0.91 - 1.15) (0.55 - 1.97) (0.99 - 1.02) (0.80 – 1.03) 1.00 0.79 1.07 1.03 0.96 1.01 0.95 (0.65 - 0.95)* (0.90 - 1.26) (0.91 - 1.17) (0.51 - 1.84) (0.99 - 1.03) (0.83 - 1.09) 1.00 (0.99 – 1.02) 1.00 0.54 0.56 (0.98 - 1.01) (0.50 - 0.58)** (0.50 - 0.62)** 1.00 (0.98 - 1.27) 1.11 (1.10 - 1.50)** 1.25 1.10 1.00 1.09 1.05 1.21 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% 100% 80% 60% <15% 15%-35% >35% 40% 20% 0% White Infants Black Ifants Note. Models include case-mix model and year dummies. *p<0.05 **p<0.01 Morales-23 03/18/04 Morales-24 03/18/04 4 Thought Expereiment-2 IV. CONCLUSIONS • 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-25 03/18/04 Morales-26 03/18/04 Conclusions Implications • Minority-serving hospitals had higher neonatal • Minority-serving hospitals may provide worse mortality than other hospitals quality of care than other hospitals • The difference in neonatal mortality between • Hospital-level factors may be more important in 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 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 Morales-27 03/18/04 Morales-28 03/18/04 Potential Explanations for Results Contact Information: • Staffing patterns – Nurse-to-patient ratio morales@rand.org or 310-794-2296 – 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-30 03/18/04 5 Study Generalizability Hospital Characteristics All (n=74,050) • Compared with all US hospitals, VON hospitals are: Minority-Serving Status <15% Black Infants 15% to 35% Black Infants >35% Black Infants Urban Hospital Region Northeast Midwest South West Volume<40 Admits per Year Level C (Highest Level of Care) Hospital Ownership Government Not For-Profit For-Profit Member, Council of Teaching Hospitals Medicaid Admissions Expense per Admission ($1000s) – 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-31 03/18/04 White (n=49,132) Black (n=24,918) 31% 36% 33% 53% 39% 43% 18% 48% 8% 35% 57% 64% 19% 27% 37% 16% 8% 28% 19% 29% 32% 20% 8% 30% 20% 23% 48% 10% 8% 24% 11% 83% 7% 48% 16% 11.7 9% 84% 7% 46% 15% 11.8 14% 80% 6% 52% 19% 11.4 Morales-32 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 Morales-33 03/18/04 6