Printer Friendly

2008 State Snapshot Methods

Last Updated: February 28, 2008

All-State Snapshot Measures

State Snapshot Summary Measures

1. Defining the Content
2. Classifying State Performance
3. Scoring State Performance (Meter Score)

Best Performing States Table

State Snapshot Strongest (Weakest) Measures

State Snapshot Focus on Diabetes

1. Prevalence
2. Quality-of-Care Performance Measures
3. Diabetes Costs
4. Disparities in Treatment

State Snapshot Focus on Asthma

1. Prevalence
2. Quality-of-Care Performance Measures
3. Quality Improvement

State Snapshot Focus on Healthy People 2010

State Snapshot Focus on Disparities

State Snapshot Ranking Table

State Snapshot Contextual Factors

Appendix I: 2008 NHQR Measures, by 2008 State Snapshot Summary Measure Assignment

Appendix II: U.S. Census Region and Division Definitions Used in the 2008 State Snapshots

Acknowledgments

Endnotes

Internet Citation

All-State Snapshot Measures

The State Snapshot primarily include State-level estimates selected from the 2008 National Healthcare Quality Report (NHQR). A few Snapshot sections include supplemental State-level analyses. Additional data sources are noted in the appropriate methodology section.

State Snapshot Summary Measures

The following methods were used to develop summary measures from the 2008 NHQR for various focus areas. Each summary measure combines multiple NHQR measures in a way that accounts for how a State performed on each measure within a year: better than average, average, or worse than average. The method encompasses three sequential decisions:

1. Defining the Content of the summary measure, that is, deciding which NHQR measures to include.
2. Classifying State performance into better than average, average, or worse than average on each NHQR measure in the summary measure.
3. Scoring State performance (meter score) on each NHQR measure and on multiple NHQR measures into a summary measure. Data were available for 2 years in the NHQR: baseline and most recent data year. Both years were used to create performance scores.

Each of these decision points is discussed separately below.

1.     Defining the Content

All NHQR measures that had State-level estimates available within the 2008 NHQR Data Tables Appendix were grouped into summary measures. These included an overall health care quality measure and 12 other summary measures within three broad areas: type of care (three summary measures), setting of care (four summary measures), and care by clinical area (five summary measures). In addition, one summary measure was defined to track clinical preventive services. NHQR measures were assigned to each of these areas on the following basis:

• Type-of-care summary measures track consumer aims (staying healthy, getting better, living with illness) with provider roles (preventing sickness, treating acute disease, and managing chronic illness) in maintaining health. The three summary measures are:
  • Preventive care: Measures that assess whether health care providers deliver specific services that prevent disease and detect it early.
  • Acute care: Measures that assess how well health care providers deliver specific services known to cure disease or speed recovery.
  • Chronic care: Measures that assess how well health care providers monitor and manage patients with incurable conditions so that the patients can live better lives.

• Setting-of-care summary measures track the quality of care delivered in different care settings. They are:
  • Hospital care: Measures that assess the quality of care provided to patients with specific health problems when they are treated in the hospital.
  • Ambulatory care: Measures that assess the quality of care provided to patients with specific conditions when they are treated in doctors' offices, clinics, and other sites of walk-in care.
  • Nursing home care: Measures that assess the quality of care provided to residents of nursing homes.
  • Home health care: Measures that assess the quality of care that is given by home health agencies to clients who receive care at home from a health care professional.

• Care-by-clinical-area summary measures track the quality of care delivered for specific types of conditions. These measures include prevention, process, and outcome measures covered under care types and settings referenced above but reorganized by clinical area. They are:
  • Cancer care: Measures that assess the quality of care provided to patients with cancer. These measures address cancer screening rates (seven measures) and cancer mortality rates (seven measures).
  • Diabetes care: Measures that assess the quality of care provided to patients with diabetes. These measures address prevention (one measure), processes of care (four measures), and outcomes of care (three avoidable hospitalizations).
  • Heart disease care: Measures that assess the quality of care provided to patients with heart disease, including heart attack (also called acute myocardial infarction, or AMI) and heart failure. These measures address prevention (three measures), processes of hospital inpatient care (nine measures), and outcomes of ambulatory care (one avoidable hospitalization).
  • Maternal and child health care: Measures that assess the quality of care provided to pregnant women and to children. These measures address prevention (two measures) and outcomes of care (seven measures).
  • Respiratory disease care: Measures that assess the quality of care provided to patients with asthma or pneumonia and to those at risk of influenza. These measures address prevention (six measures), processes of care (four measures), and outcomes of care (four measures).

• The Clinical Preventive Services summary measure represents compliance with selected recommendations of the U.S. Preventive Services Task Force and the CDC's Advisory Committee on Immunization Practice. These two expert bodies use the best research evidence available to make recommendations on preventive services for people without symptoms of disease. Such services include immunizations, tests to screen for the presence of diseases, and behavioral counseling (such as programs that encourage smokers to quit). Most preventive services are provided in primary care ambulatory clinical settings.

A complete list of the NHQR measures considered, and the summary measures to which they were assigned, is included in Appendix I.

2.     Classifying State Performance

Each NHQR measure in a State for which data were available in a year was classified twice: once reflecting its regional performance and once reflecting its national performance. The States assigned to each of the nine regions are listed in Appendix II; they are based on the nine U.S. Census Divisions.

The same approach was used below to classify each State's NHQR measures across all States (national performance) and across just those States within its region (regional performance).

Calculating the all-State and regional averages. For the all-State (national) and regional averages, we used estimates from all States that had available data for the measure. A State was excluded from the all-State or regional average if any of the following three conditions existed:

• The State estimate was unavailable.
• The standard error of the State estimate was unavailable.
• The relative standard error (RSE) of the estimate was greater than or equal to 30 percent (RSE ≥ 30%).

The RSE is calculated by dividing the standard error by the estimate. Thus, to be included in the all-State average, the standard error of a State estimate had to be less than 30 percent of the State estimate.

Instead of a typical State average from estimates weighted by the number of observations available for a State, the all-State and regional averages are from estimates weighted by the inverse of their variances, which approximates the count of observations. The differences between averages using these two methods are very small. We use the average weighted by the inverse of the variance (or a precision-weighted average) because the NHQR data tables do not include the number of observations for many of the NHQR measures.

Assigning categories. For each NHQR measure within a State, three categories were created. These categories distinguished better-than-average, average, and worse-than-average results for each NHQR measure for each State compared to the Nation and the State's region, by data year. All measures were translated into a worst-to-best metric so that measures for which "higher" represents a better result could be combined accurately with measures for which "lower" represents a better result.

To determine where each State estimate fits within the better-than-average, average, and worse-than-average categories, we applied statistical tests to each State's NHQR measures. To ensure that statistical tests gave reasonable results, we carried out the test for a State estimate only when the estimate for an NHQR measure had an RSE below 30 percent. This criterion was not applied in the tables of the NHQR. We applied it here because we were explicitly comparing States and needed more stringent criteria for statistical reliability across the items of comparison (States).

The statistical criteria used are noted in the table below.

3.     Scoring State Performance (Meter Score)

Within each of the 14 particular summary measures, each State received two sets of performance meter scores per data year—one set for national performance (n) and one set for regional performance(r), as follows:

• 1 point for each NHQR measure that was better than average.
• 0.5 point for each NHQR measure that was average.
• 0 points for each NHQR measure that was worse than average.

Depending on the comparison (national or regional), the meter score was calculated using NHQR measures taken either from all States (for comparisons to the entire Nation) or from States within the region (for comparisons to the State's region). The total number of points assigned in either comparison was divided by the total number of NHQR measures available within the respective State (A + B + C). Thus, the two equations were:

National meter score = ((A_n*1) + (B_n*0.5) + (C_n*0)) * 100
                           A + B + C

Regional meter score = ((A_r*1) + (B_r*0.5) + (C_r*0)) * 100
                           A + B + C

where A_n, B_n, and C_n indicate the comparisons to the Nation and A_r, B_r, and C_r indicate the comparisons to the region.

The result of these equations will always be: 0 < meter score < 100, equal to 0 if all NHQR measures are worse than average and equal to 100 if all NHQR measures are better than average. Scores between 0 and 100 will represent the mix of measures that are worse than average, average, and better than average. Higher scores represent better performance because the score increases with the number of measures that are average and increases more rapidly with the number of measures that are better than average. These scores are the basis for the performance meter "needles," which represent the score from 0 to 100 on a 180-degree semicircle for visual presentation. The two needles represent two different years—the most recent year of data available (a solid needle) and a baseline year (a dashed needle).

After the meter score is calculated for a summary measure, the score is assigned to one of five categories as follows for visual discrimination on the 180-degree semicircle:

• Very Weak: 0 ≤ score < 20
• Weak: 20 ≤ score < 40
• Average: 40 ≤ score < 60
• Strong: 60 ≤ score < 80
• Very Strong: 80 ≤ score ≤ 100

All meters show a solid needle for the most recent year of available data if there are a minimum number of measures reported for the composite. The minimum is set to five measures for national comparisons and set to three for the regional comparisons. The baseline year is represented as a dashed needle when the baseline has more than two-thirds of the measures available in the most recent year. This formula is applied to ensure similar comparisons between the baseline and most recent year. The text below the meter will indicate when there is insufficient data.

Best Performing States Table

Because comparison to the average does not represent the best that a State can achieve, a table has been added to the 2007 NHQR State Snapshots for each summary measure to allow States to compare their score on each summary measure to the top five States in the Nation, plus ties.

This score is the same as the meter score described above. Each table simply lists the meter score for the State of interest and the meter scores individually for the five States with the highest scores. No statistical test is applied to this comparison. It simply shows how far from the best performers the State of interest is in the context of the meter scoring.

In addition, these tables include selected percentiles (75^th, 50^th, 25^th) for the meter scores for all States available. Each represents the meter score cutoff for that percentage of States. For example, the score for the 75^th percentile is the score for which 25 percent of the States were higher and 75 percent of the States were lower. This gives a view of the spread of scores for each summary measure.

State Snapshot Strongest (Weakest) Measures

The strongest (weakest) measures for a State are based on two criteria related to all States:

• First, the measures were selected that were better (worse) than average compared to all States with data.
• Second, among those better-than-average (worse-than-average) measures for each State, the measure that ranked the highest (lowest) among the States with data were selected in turn until at least five measures were identified.

The ranking for this purpose was the ordinal rank from 1 to 51 across the States and the District of Columbia, when all jurisdictions collected the measure. When fewer jurisdictions collected a measure, the ordinal rank was inflated to a relative position as if all 51 had collected the data. For example, if 25 States collected percent of women receiving mammograms, then their ordinal rank would range from 1 to 25. To get a rank comparable to all 51 jurisdictions, the ordinal rank would be multiplied by 2.04 (2.04 = 51/25) to obtain an adjusted relative rank from 2.04 to 51.

When the fifth strongest (weakest) measure was tied in rank with additional measures beyond it, all of those measures were included in the strongest (weakest) list. For example, if the fifth measure was ranked 2 and the next three measures on the list were also ranked 2, then eight strong measures would be listed for the State.

State Snapshot Focus on Diabetes

The Focus on Diabetes section of the NHQR State Snapshots provides information on prevalence, quality, disparities, costs, and potential savings from quality improvement for diabetes care. Diabetes increasingly affects residents of every State, and State health policymakers should understand these issues more completely. The measures and methods used to develop the Focus on Diabetes estimates for the 2008 NHQR State Snapshots are described below.

1.     Prevalence

The maps visually summarize the prevalence of diabetes for each State in 2007 categorized into four quartiles: greater than 9.1 percent, from 8.0 to 9.1 percent, from 7.1 to 7.9 percent, and less than 7.1 percent. Prevalence is highest in the first quartile (greater than 9.1 percent) and lowest in last quartile (less than 7.1 percent). The data for the maps are self-reported by adults and come from the Behavioral Risk Factor Surveillance System (BRFSS).

2.     Quality-of-Care Performance Measures

The summary measure for the quality of diabetes processes of care is created and scored in the same manner as the summary measures described in Scoring State Performance (Meter Score) above. The summary measure for diabetes outcomes of care is created differently to show the actual number of hospitalizations for diabetes. This was done because it is useful for States to be able to ascertain the number of hospitalizations related to diabetes to determine the potential for cost savings.

The four diabetes process measures are from the Behavioral Risk Factor Surveillance System (BRFSS), which collects data on health behaviors in most States. These include measures of appropriate care for people with diabetes: hemoglobin A1c (HbA1c) testing, eye exams, foot exams, and flu shots. When data are not available for all of these measures for a State, that State is not reported. When data are available for only 1 year, that year of data is reported for that State.

The four diabetes outcome measures are from AHRQ's Healthcare Cost and Utilization Project (HCUP). These are measures of avoidable hospital admissions for long-term diabetes complications, short-term diabetes complications, uncontrolled diabetes without complications, and amputations related to diabetes.¹ More information on HCUP, the participating statewide data sources, and the use of HCUP data in the NHQR can be found in the HCUP Methods Series Report #2007-06 (available at http://www.hcup-us.ahrq.gov/reports/2007_06.pdf).

These four diabetes outcome measures report the number of hospital admissions for different levels of diabetes severity, with each measure defined per 100,000 people in the State. Because the denominators of these outcomes are the same, the numerators can be added to determine the total number of diabetes admissions per 100,000 people in the State. The diabetes outcomes bar chart shows the total number of diabetes admissions per 100,000 people in the State, in the region, and in the Nation. The national estimate is labeled "Nation" rather than "All States" because it is a weighted national estimate that accounts for missing States. ("All-State" estimates are estimates that include States with available data.) The regional estimate is based on the four U.S. Census Regions instead of the nine U.S. Census Divisions due to the lack of sufficient State estimates within each U.S. Census Division. States included within each U.S. Census Region are listed in Appendix II.

3.     Diabetes Costs

The Focus on Diabetes section provides information about the potential impact on State government health care costs of implementing diabetes interventions. This section presents estimates of the burden to State governments of diabetes among State employees and their dependents. It also presents the excess costs incurred by State governments if their employees are not in a disease management program or intensive intervention for improving their diabetes care. States are significant purchasers of health care, providing health care not only to State government employees but also to poor people and people with disabilities. The focus on State government employees is possible because AHRQ has sponsored work to synthesize and translate research findings into information that can aid the decisions of employers, including State government employers.

These estimates were developed with the Employers' Diabetes Costs Calculator, a tool developed from research on diabetes care, its costs, and the effectiveness of disease management. The tool was developed by The Lewin Group for AHRQ to aid employers' decisions on quality improvement related to diabetes care for their employees. The calculator provides public and private employers a rough estimate of their health care costs associated with diabetes and of the excess costs associated with poor control of blood glucose. The result reflects the potential savings that might be realized from a carefully designed disease management program or other type of quality improvement program for diabetes care. AHRQ staff and external experts have reviewed the calculator, but additional reviews and further refinements may occur. The estimates in the 2007 NHQR State Snapshots are revisions of estimates that appeared in the 2006 and 2005 State Snapshots. The revised estimates use more recent data and better methods and should be used in place of the previous State-level estimates.

Three steps are needed to estimate diabetes costs:

1. Determine the number of covered lives of State government employees and their dependents and the number of covered lives with diabetes.
2. Estimate health care expenditures associated with diabetes care.
3. Estimate excess costs associated with poor control of blood glucose.

Calculations for each of these steps are detailed in the following paragraphs.

Step 1: Determine Number of Covered Lives With Diabetes

This step involves calculating the following:

1. Number of covered lives of State government employees and their dependents by age, gender, and race/ethnicity estimated by State based on multiple data sources.
2. Diabetes prevalence by age, gender, and race/ethnicity based on national diabetes prevalence rates for these subgroups.

State government employees and their dependents. Several data sources were used to estimate the number of State government employees by race/ethnicity, gender, and age because this information is not readily available from one source. First, the number of State government employees was taken from the Bureau of Labor Statistics, 2004 Quarterly Census of Employment and Wages (QCEW).² To determine the number of State government employees by age, the age distribution of the employed population in the State was estimated from the Bureau of Labor Statistics Current Population Survey (CPS) averaged over 3 years, 2003-2005, and then applied to the QCEW data.³ Then, in order to determine race/ethnicity and gender distribution, two main sources were used: the U.S. Census Equal Employment Opportunity (EEO) Data Tool and U.S. Census State population estimates.

The EEO database provided race/ethnicity data for State government employees in cities with a minimum population of 100,000.⁴ The distribution of these employees by race/ethnicity was applied to all State government workers to obtain statewide counts of employees by race/ethnicity. When EEO data were missing for the State, the race/ethnicity distribution was taken from the Census data for the State's entire population.⁵ This was done for Alabama, Alaska, Florida, Illinois, Indiana, Kansas, Kentucky, Maryland, Michigan, Missouri, Nevada, New Jersey, New Mexico, New York, Tennessee, and Washington. For Hawaii, approximately 20 percent of the State's population was missing when Census race categories in the EEO data tool were used because the tool did not include the mixed race category. To account for people of mixed race, Claritas race data were used.⁶ The race/ethnicity distribution was assumed to be the same for males and females.

The race/ethnicity and gender distributions were then applied to the estimated number of State government employees by age to produce the number of State government employees by age, gender, and race/ethnicity for each State.

To estimate the number of State government employees who have dependents covered by their health insurance, the model estimates employees who select family coverage and have children. Estimates of the percentage of employees who select family coverage were based on AHRQ's Medical Expenditure Panel Survey (MEPS) data.⁷ The number of children per employee who selects family coverage was based on State averages from the U.S. Census Bureau.⁸

Number of covered lives with diabetes. To estimate covered lives with diabetes, the national diabetes prevalence rate was applied to the number of covered lives by State, described above. These prevalence rates were calculated using the 2005 files of the National Health Interview Survey (NHIS), with prevalence rates stratified by age, gender, and race/ethnicity.⁹ Because NHIS data are based on self-reported diabetes prevalence, another step is needed to account for the number of people with undiagnosed diabetes. The total prevalence estimate is multiplied by 1.42, the factor suggested by the 2005 CDC statistic that for every 100 people diagnosed with diabetes, approximately 42 people with diabetes have not yet been diagnosed (see the National Diabetes Fact Sheet)

Step 2: Estimate Health Care Expenditures Associated With Diabetes Care

Estimates of average, per capita health care expenditures for privately insured people with and without diabetes were calculated by combining information from the following:

• The Lewin Group's Health Benefits Simulation Model (HBSM).¹⁰
• Medical Expenditure Panel Survey data and information from a major insurance company to produce diabetes-attributed costs by age group.
• Estimates of the prevalence of diabetes for different age groups based on an analysis of the 2005 National Health Interview Survey.
• Medical Care Component of the Consumer Price Index to update the estimates to current year dollars.
• The Council for Community and Economic Research's¹¹ Cost of Living Index that provides a cross-State comparison of health care costs.
• Wage data for 2005 from the Bureau of Labor Statistics to estimate indirect costs of lost productivity.

A review of the literature identified no current estimates of health care costs for people with and without diabetes for the privately insured population. Consequently, cost estimates in the Employers' Diabetes Cost Calculator were calculated using the following steps:

1. Annual medical expenditures (excluding nursing home expenses) in the 2004 MEPS were calculated for each privately insured person.
2. For each person, diagnosis codes were used to identify whether the person had diabetes during the year and whether he or she had a health encounter for each of the major comorbidities associated with diabetes (see list of diagnosis codes in Hogan et al., 2003).
3. For each age group, two regression models were estimated, with annual medical expenditures as the dependent variable. These models attempt to define a boundary around the cost of diabetes observed in individuals who have other health problems that may or may not be associated with diabetes. In the first model, the only explanatory variable was the indicator of diabetes. The result, which does not control for other conditions, is an upper bound on the annual additional cost of diabetes compared to people without diabetes. In the second model, the explanatory variables also included the indicator variables for each comorbidity group to isolate the cost of diabetes while holding constant the cost associated with other serious comorbidities. This result, which overcontrols for the comorbidities associated with diabetes, was considered a lower bound on the annual additional cost of diabetes.
4. The annual costs estimated from the two models were averaged to produce an estimate of diabetes-attributed annual cost.

Estimates of the share of employee health care expenditures that was spent on diabetes care were made by comparing diabetes cost estimates for State employees and dependents to the State budget spent on all health care for State employees and dependents. The budget figures were obtained from the National Association of State Budget Officers.¹² For three jurisdictions (Alaska, District of Columbia, and New Mexico), the share estimates looked unreasonable and were omitted. The budget estimates for those three jurisdictions may not be as complete as for other States.

Step 3: Estimate Excess Costs Associated With Poor Control of Blood Glucose

The distribution of HbA1c levels for diabetic employees and their dependents was estimated by fitting the employer population to the distribution of HbA1c levels in the CDC National Health and Nutrition Examination Survey data for 2001-2002.¹³ The distribution is based on the reported HbA1c levels of respondents who either (1) have been told by their physician that they have diabetes and had HbA1c levels greater than 6, or (2) have not been told by their physician that they have diabetes or have been told that they are borderline diabetic and had HbA1c levels greater than 7.

Costs were estimated by assessing the impact of two hypothetical interventions. One assumes that a population's HbA1c levels can be reduced by 0.48 percentage point, on average. Another assumes that the reduction can reach 1.09 percentage points, on average. Evidence suggests that carefully designed diabetes care quality improvement programs can achieve a 0.48-point average reduction. Intensive disease management programs can achieve a 1.09-point average reduction. Both reductions imply improved glycemic control of the population.¹⁴ Improved glycemic control results in fewer complications for people with diabetes over time.^{15  16}

Differences in cost associated with an assumed improvement in HbA1c levels were based on the findings of another study. That study observed inpatient and outpatient health care charges of patients with diabetes in a large commercial health plan for 3 years. The researchers analyzed the difference in health care costs for patients who started the study period at different HbA1c levels.¹⁷ That study did not reevaluate the HbA1c levels of the study subjects at the end of the 3 years. Thus, the estimates of lower costs associated with better glycemic control assume that changes in HbA1c levels lead to fewer complications, which result in lower costs.

These assumptions were applied to State employee and dependent populations so that estimates of the cost impact of reducing their HbA1c levels by either 0.48 percentage point or 1.09 percentage points represent the difference in health care costs for States' employee populations. Their distribution of HbA1c levels was based on national HbA1c distributions for people with diabetes and a distribution of HbA1c levels in which everyone has shifted down either by 0.48 percentage point or 1.09 percentage points. The State Snapshots Web site rounds the percentage point reductions to 0.5 and 1.0, respectively, and rounds all dollar estimates to the nearest $100,000 to denote the precision that the estimates are likely to provide.

In addition to estimates of health care cost savings, estimates were made of the cost impact of gains in productivity resulting from a population's reduced HbA1c levels. The findings of a study examining the impact of HbA1c levels on rates of absenteeism and productive capacity¹⁸ were used to estimate the change in these rates. The estimates were based on downward shifts of either 0.48 percentage point or 1.09 percentage points in the distribution of State employee and dependent populations' HbA1c levels. The changes were then applied to median hourly wage data from the BLS to produce estimates of cost savings from gains in productivity under both conditions. The State Snapshots Web site rounds the percentage point reductions (to 0.5 and 1.0, respectively) and rounds all dollar estimates to the nearest $100,000 to denote the precision that the estimates are likely to provide.

For more detail on these calculations, go to Employers' Diabetes Costs Calculator.

4.     Disparities in Treatment

Data. The Disparities in Treatment section presents the percentage of adults with diabetes who had an HbA1c measurement in the past year based on data from the Diabetes Supplement to the Behavioral Risk Factor Surveillance System (BRFSS).¹⁹ The BRFSS is a household survey that, as noted above, collects data on health behaviors, including diabetes care, in most States. BRFSS data are limited in several ways:

• They are self-reported and reflect the perceptions of respondents. For example, respondents may not know about HbA1c testing or may have difficulty recalling whether they had an HbA1c test.
• A few States (Illinois, Mississippi, and Oregon) did not collect the Diabetes Supplement to BRFSS; thus, disparities data for them cannot be reported.
• Some jurisdictions did not report data for one or more data years.
• Small samples, which are typical in BRFSS, result in higher variance and poorer reliability of estimates. To improve the estimates, BRFSS data were pooled together for 3 years for this analysis.
• Some States do not have sufficient sample sizes for comparisons of subpopulations, such as by race/ethnicity. Estimates based on a cell size of less than 30 or with relative standard errors greater than 30 percent of the estimate were not used.

Racial/ethnic comparisons. In the Disparities in Treatment section, three racial/ethnic categories from BRFSS are presented:

• Non-Hispanic Black
• Hispanic
• Non-Hispanic White

Other racial/ethnic categories are not included due to small sample sizes. That is, some States either do not have many people with specific racial/ethnic heritage or did not collect large enough samples of minority groups to support analyses.

Maps of gaps in HbA1c testing. The maps visually summarize the comparison of two subpopulations across two geographic areas in terms of relative rates of HbA1c testing. That is, non-Hispanic Blacks are compared to non-Hispanic Whites, Hispanics are compared to non-Hispanic Whites, and low-income groups are compared to high-income groups within a State. Then those relative rates are compared to the same relative rates across all States with data. The result is a ratio of a ratio that represents the gap within the State in treatment between two subpopulations relative to the gap for all States with data.

For the maps, the State's gap is presented in terms of three groups and an unknown category. Assignment to the groups depends on the relative size of the gap and the relative direction of the gap between the group of interest (e.g., non-Hispanic Blacks versus non-Hispanic Whites in the State) and the comparison group (e.g., non-Hispanic Blacks versus non-Hispanic Whites in all States). To capture size and directional effects, the ratio of ratios can be assessed for whether it is substantially below, near, or substantially above 1.0. The cutoff used was below and above 5 percent. Thus, the three categories are defined as follows:

• Worse than the all-State gap: a State's relative rate is more than 5 percentage points lower than the all-State relative rate.
• Similar to the all-State gap: a State's relative rate is equal to or within 5 percentage points of the all-State relative rate (that is, within 5 points above 1.0 or 5 points below 1.0).
• Better than the all-State gap: a State's relative rate is more than 5 percentage points higher than the all-State relative rate.

Bar chart of racial/ethnic HbA1c testing rates. The bar chart shows the percentage of adults with diabetes who had a hemoglobin A1c measurement in the past year for three geographic areas. The bars represent the percentages for the State, for the region (i.e., one of nine Census Divisions) in which the State is located, and for all States with data. The data are from BRFSS. (Go to Data for a description of the BRFSS and the jurisdictions missing this information.)

Region (Census Division) or State peer-group average. The calculation of the average for each State's peer group uses the individual responses for all people in the reporting States in the relevant Census Division. (Select Appendix II for list of Census Divisions.)

All-State average. The calculation of the all-State average uses the individual responses for all the people in all reporting States.

State Snapshot Focus on Asthma

The Focus on Asthma section of the NHQR State Snapshots provides information on prevalence, quality, and estimates of the potential financial returns from quality improvement programs for populations diagnosed with asthma. Such information is important to State health policymakers, health plan administrators, and others considering or planning these programs.

1.     Prevalence

The maps visually summarize the prevalence of asthma for each State in 2007 categorized into four quartiles: greater than 9.0 percent, from 8.4 to 9.0 percent, from 7.8 to 8.3 percent, and less than 7.8 percent. Prevalence is highest in the first quartile (greater than 9.0 percent) and lowest in last quartile (less than 7.8 percent). The data for the maps are self-reported by adults and come from the Behavioral Risk Factor Surveillance System (BRFSS).

2.     Quality-of-Care Performance Measures

The three asthma outcome measures are from AHRQ's Healthcare Cost and Utilization Project (HCUP). These are measures of avoidable hospital admissions for asthma for children ages 2-17, adults ages 18-64, and adults ages 65 and over. More information on HCUP, the participating statewide data sources, and the use of HCUP data in the NHQR can be found in the HCUP Methods Series Report #2007-06 (available at http://www.hcup-us.ahrq.gov/reports/2007_06.pdf).

These three asthma outcome measures report the number of hospital admissions for different age groups, with each measure defined per 100,000 people of that age group in the State. The quality of asthma care bar chart shows the total number of asthma admissions per 100,000 people per age group in the State, in the region, and in the Nation. The national estimate is labeled "Nation" rather than "All States" because it is a weighted national estimate that accounts for missing States. ("All-State" estimates are estimates that include States with available data.) The regional estimate is based on the four U.S. Census Regions instead of the nine U.S. Census Divisions due to the lack of sufficient State estimates within each U.S. Census Division. States included within each U.S. Census Region are listed in Appendix II.

3.     Quality Improvement

Asthma care improvement programs typically follow the guidelines of the National Asthma Education and Prevention Program (NAEPP). The NAEPP calls for education of patients and providers to better manage the disease. The NAEPP activities for patients focus on self-management to avoid triggers, anticipate problems, and use medications appropriately. The activities for providers focus on accurate diagnosis, appropriate medication prescribing, patient monitoring, and patient education on how to maintain control and avoid attacks. Often education leads to reductions in the need for hospitalization, emergency department visits, urgent office visits, and missed work or school days due to asthma attacks, although it may result in greater pharmaceutical use. Ultimately, net reductions save health care dollars and improve productivity.

The estimates of the health care savings and productivity gains provided in the Focus on Asthma section were developed using the Asthma Return-on-Investment Calculator. The calculator, developed for AHRQ by Thomson Reuters with AHRQ funding, is a tool which uses findings from the literature to estimate potential financial returns of quality improvement programs.

The calculator combines specific characteristics of the targeted population and of the planned asthma program with current research findings (as of March 2007), to generate results which include:

1. Health care savings per participant
2. Productivity gains per participant
3. Return on investment (the number of dollars saved for each dollar invested in the program)

The estimates presented in the focus on asthma section were developed based on a hypothetical asthma care quality improvement intervention program which 1) targets patients with persistent asthma with at least one acute care visit in a one year period, 2) assumes 25 percent of eligible patients will participate, 3) costs $395 per patient per year, and 4) runs for four years. This hypothetical intervention targets the sickest asthma patients, and, thus, shows the best estimated returns. Not all hypothetical intervention programs will result in a favorable ROI. For the purpose of this modeling, State employees are the same as privately insured. For a detailed description of the calculator and the default settings used to generate the estimates, refer to the model description section of the Asthma Return-on-Investment Calculator (available soon).

State Snapshot Focus on Healthy People 2010

The Focus on Healthy People 2010 section compiles a table of 24 measures that reflect U.S. health goals and that are reported by States. These health goals are intended to increase life expectancy, improve quality of life, and eliminate health disparities throughout the Nation. Launched by the Department of Health and Human Services in 2000, the goals provide Federal, State, and local government agencies and nongovernmental organizations with a framework for assessing progress in a comprehensive set of eight focus areas:

• Access to Quality Health Services
• Cancer
• Chronic Kidney Disease
• Heart Disease and Stroke
• HIV
• Immunization and Infectious Diseases
• Maternal, Infant, and Child Health
• Mental Health and Mental Illness

The table, sorted by focus area, displays the Healthy People 2010 target rate, the most recent State rate and data year, and the baseline State rate and data year. Measure definitions are also provided.

State Snapshot Focus on Disparities

The Focus on Disparities section presents data on disparities in the quality of care among racial, ethnic and socioeconomic groups. State-level quality of care measures for non-Hispanic Blacks, non-Hispanic Asian/Pacific Islanders, Hispanics, and low-income individuals are compared against the same measures for non-Hispanic whites. A performance meter displays how the disparities in quality of care for the State relates to the disparity in the Nation—whether the State level is about average, below average or above average when compared to the National level.

The quality of care measures used to rate a State’s performance on the Performance Meter focus on potentially preventable hospitalizations in the following clinical areas:

1. Respiratory Care
2. Heart Disease
3. Diabetes

Disparities are presented by these clinical areas in a table which lists the measures related to each clinical area, and which indicates, for each measure, how the quality of care or outcomes of the minority group compare to the quality of care or outcomes for non-Hispanic Whites in that State and how the State performance compares to U.S. performance. These comparisons are determined by calculating the ratio of the minority group measure to the non-Hispanic white reference group measure at the State and National levels.

Data for this focus area are from AHRQ's Healthcare Cost and Utilization Project (HCUP) and use version 3.1 of the AHRQ Quality Indicator (QI) software. More information on HCUP can be found on the HCUP User Support Web site under Method Series Reports (http://www.hcup-us.ahrq.gov/reports/methods.jsp).

State Snapshot Ranking Table

To enable simple direct comparisons of States on some health care quality measures underlying the summary measures, States were ranked from 1 to 51 on a select set of 15 measures from the NHQR for which all States reported. These measures include core measures for the most common diseases reported in the NHQR. Core measures represent the most important and scientifically credible measures of health care quality for the Nation. They were selected by the Department of Health and Human Services Interagency Workgroup for the NHQR. Many of the core measures selected by that workgroup did not have State-level data. The other measures in the core areas were selected to round out the group of 15 reported in the State Snapshots.

State Snapshot Contextual Factors

The context of the State's environment is shown in a series of dials. Seven dials relate to State demographics, three relate to health status, and three relate to health care resources. These factors provide a backdrop to the State's health care quality and may aid in interpreting the State's performance meters. These contextual factors might have a cause, effect, or other indirect association with the results shown in the performance meter. For example, if a high percentage of the State's population is without health insurance, a high percentage of the State's population might not use preventive services.

The dials show the State's rate for the factor and the range of rates across all reporting States. An orange wedge on each dial shows the spread of values for all reporting States (or reporting States in the region), ranging from the State with the lowest to the State with the highest value. The arrow (on top of the orange portion) represents the State's percent or rate of the factor.

Data sources for the contextual factor dials follow:

• The Kaiser Family Foundation (KFF) (see http://www.statehealthfacts.org). The KFF compiles data from:
  • Healthleaders, Inc., 2008
  • Interstudy Competitive Edge, Part II: Managed Care Industry Report, 2007
  • Current Population Survey, U.S. Census, 2006, 2007 and 2008
  • Behavioral Risk Factor Surveillance System, Centers for Disease Control and Prevention, 2006, 2007 and 2008
  • Annual Survey, American Hospital Association, 2006
• Morbidity and Mortality Weekly Report, Centers for Disease Control and Prevention, February 2005
• State Profiles, Reforming the Health Care System, 2005, by AARP Public Policy Institute

Appendix I: 2008 NHQR Measures, by 2008 State Snapshot Summary Measure Assignment

This appendix lists the NHQR measures included in the 13 summary measures, excluding the overall summary measure. The overall summary includes all measures in the tables below (except for those in the excluded table) reported by a State. Individual measures may appear in multiple groupings. The list of measures is organized by:

• Preventive care
• Acute care
• Chronic care

• Hospital care
• Ambulatory care
• Nursing home care
• Home health care

• Cancer
• Diabetes
• Heart Disease
• Maternal and Child Health
• Respiratory Diseases

Types of Care: Preventive Care Measures

Types of Care: Acute Care Measures

Types of Care: Chronic Care Measures

Settings of Care: Hospital Care Measures

Settings of Care: Ambulatory Care Measures

Settings of Care: Nursing Home Care Measures

Settings of Care: Home Health Care Measures

Care by Clinical Area: Cancer Care Measures

Care by Clinical Area: Diabetes Care Measures

Care by Clinical Area: Heart Disease Care Measures

Care by Clinical Area: Maternal and Child Health Care Measures

Care by Clinical Area: Respiratory Diseases Care Measures

Clinical Preventive Services

Appendix II: U.S. Census Region and Division Definitions Used in the 2008 State Snapshots

Acknowledgments

The 2008 State Snapshots were developed from the 2008 National Healthcare Quality Report through a team effort including the Agency for Healthcare Research and Quality (Ernest Moy, Jeffrey Brady, Karen Ho, Karen Migdail, Bruce Seeman, Biff LeVee, Margaret Rutherford, Doreen Bonnett, Roxanne Andrews), Thomson Reuters (Healthcare) Inc. (Rosanna Coffey, Jillian Dudek, Elizabeth Stranges, Minya Sheng, Susan Raetzman), ML Barrett, Inc. (Marguerite Barrett), Social & Scientific Systems (Paul Gorrell, Laurie MacCallum, Mikki Hall, Anil Koninty, Debbie Machen).

These State Snapshots have built on work of earlier years and contributions of the above individuals and others: Agency for Healthcare Research and Quality (Edward Kelley, Dwight McNeill, Marybeth Farquhar, DonnaRae Castillo, David Atkins, Christine Williams, Sandi Isaacson, Mary Nix, Kathy Crosson, Gerri Michael-Dyer, Marjorie Shofer), Thomson Reuters (Healthcare) Inc. (Craig Hunter, Julia Nisbet, Jim Blakley, Mirjana Milenkovic, Kathy Hickey, Angela Fulmer), Social & Scientific Systems (Dale Byington, Nathalie Fike), ECRI (Vivian Coates, Steve Rhoads, Evan LeGault, and Pamela Nash),Kenney IS Consulting (Tim Kenney), the National Governors Association, the National Conference of State Legislators, the Council of State Governments, the Association of State and Territorial Health Officials, and the Federal Interagency Workgroup for the National Healthcare Quality Report. Additional support was provided by AcademyHealth (Enrique Martinez-Vidal, Amanda Brodt), the Lewin Group (Tim Dall, Sarah Stout), and the Madison Design Group (Russ Surles, Anne Kerns, Darin Ruchirek).

Endnotes

Internet Citation

2008 State Snapshots: Methods. Derived from 2008 National Healthcare Quality Report. March 2008. Rockville, MD: Agency for Healthcare Research and Quality. http://statesnapshots.ahrq.gov/.