Maryland's Approach to Racial and Ethnic Minority Health Data Analysis

Slide presentation from the AHRQ 2011 conference.

On September 21, 2011, David Mann made this presentation at the 2011 Annual Conference. Select to access the PowerPoint® presentation (940 KB). Plugin Software Help.


Slide 1

Maryland's Approach to Racial and Ethnic Minority Health Data Analysis and Reporting

2011 AHRQ Annual Conference

Maryland's Approach to Racial and Ethnic Minority Health Data Analysis and Reporting

Dr. David A. Mann
September 21, 2011

Office of Minority Health and Health Disparities
Maryland Department of Health and Mental Hygiene (DHMH)

Slide 2

Uses of Data for Disparities Elimination

Uses of Data for Disparities Elimination

  • Identify, Locate and Quantify Disparities.
  • Understand Causes of Disparities and Plan Interventions.
  • Track Progress Towards Elimination.

Slide 3

Causal Chain for Health Outcomes

Causal Chain for Health Outcomes

Image: Flowchart depicts the causal chain for health outcomes, as described below:

Health outcomes, and thus disparities in health outcomes, occur along a causal chain of measurable steps. Level 1 factors are social determinants of health such as education, employment, housing, food and exercise opportunities, etc. These lead to the level two risk factors, such as poor diet, physical inactivity, smoking, obesity, high blood pressure, and high cholesterol. These level two risk factors drive level three, the frequency of disease in the population, expressed as either incidence or prevalence. Examples of are incidence and prevalence of diabetes, coronary heart disease, stroke, etc. Finally, the frequency of disease is one driver of the morbidity (disability and suffering) and mortality (death) due to diseases like diabetes, coronary heart disease, and stroke. This disease-specific morbidity and mortality is level 4 in the causal chain of health outcomes.

Case-specific event rates, which is the rate at which persons move from risk factor to disease and from disease to morbidity and mortality, also influence the frequency of disease, disability and death in the population. The health care system and the broader public health system work to reduce the frequency at each level primarily by reducing the transition rates from one level to the next, as well as reducing the frequency of the initial level factors.

At each level, individual or group genetic patterns will influence the susceptibility to acquiring a risk factor or disease, and thus be an additional influence on the primary factor frequency and the case-specific transition rates along the causal chain.

Genetics: At each step, individual or group genetic patterns can influence the susceptibility to moving from one level to the next.

An example of this chain is: Food desert plus no safe place for exercise (level 1) leads to Obesity (level 2) which leads to Diabetes (level 3) which leads to Diabetes-related: blindness, End-Stage Renal Disease, amputations, and death (level 4).

Slide 4

Data Sources for Health Outcomes

Data Sources for Health Outcomes

Image: Flowchart showing 4 levels of illness.

Each of these four levels of health outcome in the causal chain have useful data sources.

Level 1: Social Determinants of Health: Non health data sources: Poverty rate, unemployment rate, HS graduation rate,crime rate, etc.

Level 2: Risk Factor Prevalence: Behavioral Risk Factor Surveillance System (BRFSS) data, other local surveys, "claims-coded prevalence"* .

Level 3: Disease Frequency: BRFSS data, other local surveys, "claims-coded prevalence"* .

Level 4:  Morbidity and Mortality: Vital Statistics data, Centers for Disease Control and Prevention's (CDC) Wonder, BRFSS, registries, "claims-coded prevalence"*.

*"Claims-coded prevalence": prevalence estimate using the count with relevant codes from administrative data as numerator; and one of three denominators: Utilizers, enrollees, or an entire population.

Slide 5

Health Outcomes >> Utilization

Health Outcomes >> Utilization

(4 levels of illness)

Level 1: Social Determinants of Health
Level 2: Risk Factor Prevalence
Level 3: Disease Frequency
Level 4: Morbidity and Mortality

Levels 2, 3, and 4 lead to: 

Case-Specific Event Rates:

Health Care Utilization Data*:

  • Disparities in Utilization Rates:
    • More may be better: Joint replacement, cardiac revascularization, etc.
    • More is worse: diabetic amputations.
  • Disparities in Costs:
    • Frequency Disparity in Cost.
    • Severity Disparity in Cost.

* Utilization data may be provider-based (hospital discharge or ER data), or may be payer-based (insurance data). In the future it may be medical record based (EMR + HIE). Data accuracy and unique ID may vary by source.

Slide 6

What Maryland Has Done

What Maryland Has Done

  • (L4) Mortality: Vital Statistics Reports and CDC Wonder.
  • (L3) Disease Frequency:
    • Incidence: Cancer Registry, HIV/AIDS registry, U.S. Renal Data System (end stage renal disease [ESRD] incidence).
    • Prevalence: BRFSS (prevalence of doctor diagnosis only).
  • (L2) Risk Factor Prevalence:
    • Behavioral factors from BRFSS: smoking, obesity, physical activity. Smoking also from state tobacco survey.
    • Screening factors from BRFSS: mammography, colonoscopy.
  • (L1) Social Determinants of Health:
    • County level social risk profiles.

Slide 7

What Maryland Has Done (2)

What Maryland Has Done (2)

  • Cost of disparities analysis in discharge data:
    • Hospital discharge data analysis of Black-White hospitalization disparities.
  • Cost of disparities analysis in Medicare data:
    • Analysis of ambulatory care sensitive conditions (ACSC) admissions in Medicare recipients age 65+.
    • Removes problem of out of state admissions.
  • Examples of this work, which illustrate various themes and lessons, follow:
    • Issues of age-adjustment are central to most analyses.
    • Pros and cons of rate ratios vs. rate differences are important.

Slide 8

Mortality Data by Race and County (L4)

Mortality Data by Race and County (L4)

Age-Adjusted All-Cause Mortality (rate per 100,000) by Black or White Race and by Jurisdiction, Maryland 2004-2006 Pooled

Image: Line graph displays the following data:

CountyBlack or
African American
White
Baltimore City1211988
Kent1117800
Wicomico1023875
Caroline999898
Dorchester998866
Talbot997690
Anne Arundel992833
All of Maryland980768
Harford971826
Prince George's968800
Baltimore County962804
Calvert961865
Somerset923916
Worcester910718
Queen Anne's907778
St. Mary's903853
Carroll900816
Charles891880
Washington878831
Cecil834864
Frederick755775
Allegany713890
Montgomery709560
Howard661648

Somerset has a smaller disparity than Montgomery...

But Somerset has much worse Black mortality than Montgomery, and the 2nd worst White mortality.

Lesson: The disparity metric displayed alone can be misleading!!!

Slide 9

Cause-Specific Mortality Data by Race and County (L4)

Cause-Specific Mortality Data by Race and County (L4)

Age-Adjusted Mortality Rates (per 100,000), Selected Causes of Death for Blacks or African Americans and Whites, Somerset County, Maryland 2002-2006

Image: Bar chart displays the following data:

  • Black 966.4 and White 965.2 for all causes.
  • Black 342.4 and White 342.5 for heart disease.
  • Black 230.8 and White 258.0 for cancer.
  • Black 65.9 and White 25.1 for diabetes.

Lesson: For small counties (like Somerset) or small racial or ethnic groups, pooling multiple years of data can allow metric estimation even for less common outcomes (like diabetes compared to heart and cancer).

Source: CDC Wonder online Database, Compressed Mortality Files 2002-2006.

Slide 10

Rate Ratio vs. Rate Difference

Rate Ratio vs. Rate Difference

Image: A chart displays Black versus White Mortality for the 14 leading causes of death in Maryland, 2008. The largest disparity by rate difference: Heart, Cancer; the largest disparity by rate Ratio: HIV/AIDS, Homicide.

Lesson: "Worst" Disparity Depends on Which Metric is Used is shown.

Slide 11

Ratio vs. Difference: Implications for Trends and Evaluation

Ratio vs. Difference: Implications for Trends and Evaluation

Chart is titled: Hypothetical Results of a Minority Health Program: Success or Not?

Image: A chart shows all cause Black versus White mortality rates and ratios for the hypothetical results of a minority health program.

Age-adjusted Rate per 100,000

RaceAll Cause
Mortality 2020
All Cause
Mortality 2030
Change% Change
Black20090-110-55%
White10030-70-70%
Difference10060-40-40%
Ratio2.03.01.050%

Lesson: Rate ratio disparity metrics, considered in isolation, can underestimate the success of minority health programs.
This is crucial to understand if trends in such metrics are used for funding decisions.

Slide 12

U.S. Renal Data System Data for ESRD Incidence (L3)

U.S. Renal Data System Data for ESRD Incidence (L3)

Graph titled: Incidence of All-Cause ESRD by Age and Race, Maryland 1991-2001 Pooled (DHMH Analysis of US Rental Data System data).

Image: Line graph displays the following data:

Race0-2425-3435-4445-5455-6465-7475+
White1.255.928.7618.5440.4790.86101.61
Black4.7329.4459.26108.95228.95345.69327.97
Asian2.546.2412.7415.0939.54109.14137.63
American Indian4.2221.7624.1155.88142.03258.61453.17

Lesson: Fine age stratification for age-adjustment, plus long multi-year pool can make the data robust for estimation in smaller groups.

Slide 13

BRFSS Data for Risk Factor Prevalence (L2)

BRFSS Data for Risk Factor Prevalence (L2)

Percent of Persons (45-64 yrs) Classified as Obese (BMI >29.99)
Maryland BRFSS 2004-2008

Image: Bar chart displays the following data:

Race%
Non Hispanic, White27.6%
Non Hispanic, Black37.3%*
Non Hispanic, Other17.8%*
Hispanic30.0%

Source: Maryland BRFSS Data 2004 to 2008.

* = significantly different from NH White rate.

18-44 and 65+ show a similar pattern to 45-64.

Lesson: Coarse age stratification for age-adjustment, plus multi-year pooling can make the data robust for estimation in smaller groups.

Slide 14

Utilization Analysis for Cost of Disparities

Utilization Analysis for Cost of Disparities

Image: Bar charts shows Black vs. White Disparity Ratios for Adults with Asthma, Maryland 2006:

Prevalence: 1.3
ED Visits: 4.3
Hospitalization: 2.4
Mortality: 2.4

Source: This figure is Figure 8-5 from the DHMH report, Asthma in Maryland 2007.

330% more ED visits and 140% more hospital admissions with only 30% more asthma indicates a disparity in disease management success.

Formula for attributable fraction in the exposed: (RR-1)/RR (2.4-1)/2.4 = 1.4/2.4 = 58.3% of Black Asthma hospitalizations are excess.

Slide 15

Discharge Data Analysis of Cost of Disparities

Discharge Data Analysis of Cost of Disparities

Chart titled: Cost of Disparities, Maryland 2004. Cost of Excess Black or African American Admissions Hospital Component of Hospital Admissions. MHHD Analysis of HSCRC Hospital Discharge Data.

  • For all admissions, Black disparity excess costs in 2004 were 59 million dollars for Medicaid and 481 million dollars for all payers.
  • For heart disease, 5 million for Medicaid and 38 million for all payers.
  • For cancer, 1 million for Medicaid and 7 million for all payers.
  • For diabetes, 3 million for Medicaid and 26 million for all payers.
  • For asthma, 2 million for Medicaid and 18 million for all payers.
  • For neonatal intensive care unit admissions, 3 million for Medicaid and 20 Million for all payers.

How might out of state admissions be affecting these estimates?

  1. Check consistency with estimates in Baltimore City, an "internal" jurisdiction where admissions out of state are less likely.
  2. Check consistency with estimates from Medicare data, where the out of state issue does not exist.

Slide 16

Medicare Data Analysis of Cost of Disparities for Maryland

Medicare Data Analysis of Cost of Disparities for Maryland

Chart titled: Cost of Disparities, Maryland 2006. Cost of Excess Black or African American Admissions Hospital Component of Hospital Admissions. MHCC Analysis of Maryland Medicare Data.

  • 13 million dollars for congestive heart failure.
  • 2 million dollars for urinary tract infection.
  • 2 million dollars for dehydration.
  • 5 million dollars for diabetes.
  • 1 million dollars for asthma.
  • 1 million dollars for hypertension.

Does not include Physician component of Hospital Admission. Does not include Emergency room costs. Does not include Outpatient Care costs.

Analysis of Medicare data in persons age 65+ is consistent with the statewide discharge data analysis.

Analysis of payer-based claims data (vs. provider-based data) where available avoids the missing out-of-state utilization issues.

Frequency disparity vs. Severity disparity.

Source: Differences in Hospitalizations for Ambulatory Care Sensitive Conditions Among Maryland Medicare Beneficiaries—2006. Maryland Health Care Commission.

Slide 17

Discharge Data Analysis of Cost of Disparities

Discharge Data Analysis of Cost of Disparities

Why was analysis restricted to Black vs. White in 2004?

Count of admissions missing race data: 30,087
Count of admissions missing Hispanic ethnicity data: 51,483

Count of admissions recorded as American Indian or Alaska Native: 1,537
Missing race as percent of known AIAN = 1957%

Count of admissions recorded as Asian or Pacific Islander: 12,011
Missing race as percent of known API = 250%

Count of admissions recorded as Hispanic: 19,449
Missing Hispanic ethnicity as percent of known Hispanic = 265%

Count of admissions recorded as Black or African American: 207,495
Missing race as percent of known Black or African American = 15%

Slide 18

Contact Information

Contact Information

Office of Minority Health and Health Disparities
Maryland Department of Health and Mental Hygiene
201 West Preston Street, Room 500
Baltimore, Maryland 21201

Web site: http://www.dhmh.maryland.gov/hd/ 
Chartbook: http://www.dhmh.state.md.us/hd/pdf/2010/Chartbook_2nd_Ed_Final_2010_04_28.pdf  (Plugin Software Help)

Phone: 410-767-7117
Fax: 410-333-5100
E-mail: healthdisparities@dhmh.state.md.us

Current as of March 2012
Internet Citation: Maryland's Approach to Racial and Ethnic Minority Health Data Analysis. March 2012. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/news/events/conference/2011/mann/index.html