Appendix C: Additional Data Resources Related to Diabetes Care Quality

Diabetes Care Quality Improvement: A Resource Guide for State Action

This appendix contains additional information and detailed tables on the following:

  • National Healthcare Quality Report measures selection process
Table C.1: Diabetes measure set for the NHQR with endorsing organizations, data sources, and level of data
  • Descriptions of data sources for the process and outcome measures discussed in this Resource Guide (including notable differences between MEPS and BRFSS national rates) and data tables from the NHQR
Table C.2: Percent of non-institutionalized adults over age 18 saying they were diagnosed with diabetes who reported having important tests in the past year (or two years in one case), by national population subgroup, United States, 2000
Table C.3: Percent of adults age 18 and over with diagnosed diabetes who have specific HbA1c levels and who have specific blood pressure levels, United States, 1999-2000
Table C.4: Hospital admission for adults over age 18 for specific diabetes complications (excluding obstetric and neonatal admissions and transfers from other institutions) per 100,000 population age 18 and older, Healthcare Cost and Utilization Project (HCUP), United States, 2000
Table C.5: Lower extremity amputations in persons with diabetes per 1,000 population (all ages), National Hospital Discharge Survey, United States, 1997-2000
Table C.6: CDC Three-Year Baseline: Percent of adults age 18 and over with diabetes who had recommended diabetes tests in the past year, pooled 1997-1999
Table C.7: CDC Annual Trends: Percent who had a dilated-eye examination in the past year per 100 adults with diabetes, crude rates and age-adjusted rates, by State, 1995-2002
Table C.8: CDC Annual Trends: Percent who had a foot examination in the past year per 100 adults with diabetes, crude rates and age-adjusted rates, by State, 1995-2002
Table C.9: CDC Annual Trends: Percent who had an influenza vaccination in the past year per 100 adults with diabetes, crude rates and age-adjusted rates, by State, 1993-2001
  • Flow chart of steps for estimating State Medicaid spending on diabetes care
Figure C.1: Estimation steps for Medicaid spending on diabetes care

Return to Contents

 

National Healthcare Quality Report Measures Selection Process

Researchers have developed health care quality measures based on scientific evidence, practice guidelines, and consensus processes. Consensus building around measure sets has been used recently to narrow the list of measures and increase their acceptance.

Consensus building is a process by which stakeholders and experts in a field identify a connection between a measure and quality health care. The process generally includes expert judgment and evaluation, rigorous testing of the measure in the field to ensure that improved performance is linked to improved health, and review and agreement of the experts. Several organizations are involved in developing national quality measure sets, including:

  • The National Diabetes Quality Improvement Alliance (discussed in Module 4: Action).
  • The National Quality Forum.
  • The National Committee on Quality Assurance, the accrediting body for managed care health plans, with its HEDIS® performance measures.

In the first NHQR, AHRQ pursued a careful process to define the first set of NHQR measures. The underlying framework for selecting the NHQR measure set was developed by the Institute of Medicine in Envisioning the National Health Care Quality Report. Their matrix framework crosses components of health system quality (effectiveness, safety, timeliness, and patient centeredness) with consumers' health care needs (staying healthy, getting better, living with illness or disability, and end-of-life care). Measures were chosen to fill the cells of this matrix so that all areas of health care quality would be addressed. The measure selection process:

  1. Invited organizations with consensus-based measures (developed with experts and often rigorous testing) to submit them for review.
  2. Issued a public call for measures and data sources.
  3. Convened a Federal Interagency Workgroup to evaluate (based on importance, scientific soundness and feasibility) the 600 measures submitted and to select the final set of 147 measures.
  4. Invited public review of the final measure set.

 

Out of the set of 147 measures in the NHQR, 12 are diabetes measures. All diabetes measures were developed through consensus processes of the endorsing organizations. Table C.1 lists the NHQR diabetes measures, the organizations that endorse them through a consensus process, the data sources, and the analytic level (State or national) supported by the data.

Return to Contents

 

Data Source Description, Limitations, and Data Tables From the NHQR

Notable Differences Between MEPS and BRFSS National Rates

Some of the MEPS and BRFSS measures on diabetes are the same and both are used in the NHQR. However, only one MEPS measure is used in this Resource Guide because not only does MEPS not give State-level estimates, the methods used to derive the MEPS and BRFSS estimates for the same measures differ. As a result, the NHQR diabetes estimates from MEPS and BRFSS show notable differences for the HbA1c and immunization measures. MEPS reports that 90 percent of people with diabetes get one or more HbA1c test per year; BFRSS reports 79 percent. MEPS reports 55 percent of people with diabetes receive a flu vaccination; BRFSS reports 37 percent. MEPS and BRFSS are very close on the eye and foot examination rates, 67 versus 67 percent and 66 versus 65 percent, respectively.

The difference for the HbA1c test rate is in part due to the structure of the survey questions and in part due to the treatment of respondents who "have not heard of HbA1c."; While BRFSS allows for this distinction to be made in the response options, MEPS does not. These respondents are counted as though they answered "no"; in BRFSS, and potentially not included in MEPS. The percentage of these responses in BRFSS is fairly low, at about 5% in 2001, but will still affect the final rate.

The difference for influenza immunizations is due to definitional differences between BRFSS and MEPS. The BRFSS rate is for adults age 18 to 64; the MEPS rate is for adults age 18 and over. Since flu shots are more often given to the elderly, the BRFSS rate is lower than the MEPS rate.

There are other differences between the two data sources that can contribute to the differences between estimates of the same measures. The surveys relate to different time periods, use different sampling approaches, and use different interview techniques, to name the obvious. Because of the differences in the estimates of the same measures and because only BRFSS permits State estimates, only BRFSS estimates are discussed in Module 3: Information of this Resource Guide.

Process Measures—MEPS data

The MEPS data provide national benchmarks by important segments of the population. Its breakdowns identify subgroups for whom diabetes care quality can be problematic and for whom solutions need to be targeted.

 

Table C.2 shows estimates for the five MEPS diabetes-related measures in the NHQR. The estimates are provided by national subgroups related to race, ethnicity, sex, age, education, employment status, income, health insurance status, respondent's location, and perceived health status. Table C.2 shows the rate per 100 respondents (or percent) and the standard error for each measure and subgroup. Only estimates that have a standard error that is less than 30 percent of the estimate (relative standard error < 30 percent) are shown on Table C.2. No statistical comparison tests were performed in Table C.2 but the estimates and standard errors can be used to make such comparisons (go to Appendix E for how to do this).

Outcome Measures—NHANES, HCUP, and NHDS Data

NHANES Data

The NHQR uses data from the National Health and Nutrition Examination Survey for two outcome measures related to diabetes—the HbA1c, a measure of blood glucose level over the prior two to three months, and blood pressure at examination. NHANES does not support State-level estimates but does provide clinical outcome estimates for the total national population that could be used as benchmarks. [Note: To be comparable to data from providers, the NHANES HbA1c and blood pressure values would have to be recalculated to exclude people who do not use the health care system during a year.]

The NHANES collects data from in-person interviews, physical examinations, and medical tests from a mobile vehicle which is set up as a medical office. With this survey method, NHANES is able to collect data that are detailed clinically, including laboratory results. Because of the expense of the NHANES (e.g., the cost of the mobile clinic and staff), the sample size on the NHANES is small, 9,965 participants, and does not support either State-level estimates or national subgroup estimates within the population of people with diabetes. Additional information on NHANES is available at: http://www.cdc.gov/nchs/about/major/nhanes/NHANES99_00.htm.

Table C.3 shows the percent of adults with diabetes by specific test values. The percent and standard error are provided.

 

Table C.3. Percent of adults age 18 and over with diagnosed diabetes who have specific HbA1c levels and who have specific blood pressure levels, United States, 1999-2000

Test ResultsPercentStandard error
HbA1c Levels:
   > 9.5 percent (poor control)13.52.6
   < 9.0 percent (needs improvement)79.12.7
   < 7.0 percent (optimal)37.03.8
Average blood pressure at exam
   <140/90 mm/Hg59.33.5

Source: Centers for Disease Control and Prevention, National Center for Health Statistics, National Health and Nutrition Examination Survey.

In addition to the HbA1c and blood pressure values, NHANES can provide LDL-C levels (< 130 mg/dL (needs improvement) and <100 (optimal)). Those LDL estimates were not available in time for publication of the first NHQR. The measure remains part of the official NHQR measure set and is to be included in the future.

HCUP Data

The NHQR uses inpatient discharge abstract data from the Healthcare Cost and Utilization Project for national estimates of three outcome measures that provide a window on the quality of ambulatory care— avoidable hospitalizations related to diabetes. While the national estimates are included in the NHQR, State-level data are not, except for one special analysis of admissions for uncomplicated uncontrolled diabetes (discussed in Module 3: Information). For year 2000 data used in the NHQR, 29 States contributed and more States had statewide discharge data systems maintained by State data agencies, State hospital associations, or statewide data consortia. These data systems can be used to generate these three outcome measures.

HCUP is a public-private partnership sponsored by AHRQ with 29 participating States in 2000, the time for which data are included in the first NHQR. The data are from statewide historical administrative databases going back to 1988. In 2000, HCUP included the:

  • Nationwide Inpatient Sample (NIS) — all hospitals and all of their inpatient discharges (6 to 7 million records per year) across the 29 States, weighted so that national estimates can be derived from it.
  • State Inpatient Databases (SID) — a census of inpatient discharge records for each participating State covering nearly 80 percent of the 36 million U.S. hospital discharges per year in 2000.
  • State Ambulatory Surgery Databases (SASD) — all discharge records for ambulatory surgery centers (hospital based and freestanding).
  • Kids' Inpatient Database (KID) — A sample of children's discharges from over 2,500 community hospitals.

In addition, AHRQ developed the Quality Indicators, which are measures of health care quality that make use of readily available hospital inpatient administrative data and available as public software to help analysts evaluate quality of care in hospitals and, indirectly, in ambulatory care settings. The AHRQ QIs are organized into three categories, the Prevention Quality Indicators, the Inpatient Quality indicators, and the Patient Safety Indicators. Additional information on HCUP data is available at: http://www.hcupus.ahrq.gov/databases.jsp. Additional information on the AHRQ QIs is available at: http://www.qualityindicators.ahrq.gov/.

 

The NHQR and NHDR used selected Prevention Quality Indicators to examine hospital admissions that evidence suggests could have been avoided, at least I part, through high-quality outpatient care. Table C.4 shows rates of three of these indicators related to diabetes—hospital admissions for three complications of diabetes that should be avoidable—1) uncontrolled uncomplicated diabetes, 2) serious short-term complications, and 3) serious long-term complications. The rates are defined relative to 100,000 people in the population of the State. Results are presented by patient characteristics (age, sex, income, and location of patient residence) and by hospital characteristics (region of the country). The rate and its standard error are shown.

The main limitation of HCUP data (or any other administrative data source) is that the data are collected for one purpose and used for another. Many State-level discharge data systems use data from hospital billing data and are thus collected for reimbursement purposes. However, these data are so valuable that they are used for many other purposes, such as cost tracking, quality monitoring, or health policy evaluations. Reimbursement incentives affect what data are collected and how they are collected. Thus, while mining these data for clues to quality, analysts should constantly be on the alert for data problems— incomplete or inaccurate entries or lack of adequate clinical detail.

NHDS Data

 

Table C.5 shows the rate of lower extremity amputations for people with diabetes per 1,000 population, for two time periods—1997 through 1999 and 1998 through 2000. NHDS pools data over several years, which is why the tables reflect an overlap in years. The estimates are provided by national subgroups when the size of the database supports the subgroup estimate (i.e., for age, sex, and black-white subgroups). The rate and standard error of the rate are provided. Each State with discharge data can generate estimates for all of the subgroups reported.

Return to Contents

 

Estimation Steps for Medicaid Spending on Diabetes Care by State

This section describes methods for estimating Medicaid spending on medical care for people with diabetes by State.

 

Estimates for Medicaid spending on diabetes care had to be constructed from multiple public data sources, as described in Module 2: Data. Because the estimation involved many assumptions, the method used is described in a flow chart (Figure C.1). The top level of the flow chart represents the original data sources; the middle levels show assumptions, adjustments, and calculations made with the original data; and the final level (at the bottom) of the flow sheet is the result. The adjustments were necessary to make different sources compatible with respect to population and time frame. This method was applied to data on Medicaid eligibles to get an estimate of the potential cost for Medicaid of medical care for diabetes patients.

These are only "ball-park"; estimates because of the assumptions that had to be made to work with available data. Obvious limitations in these estimates include omission of spending for children and the institutionalized population. Furthermore spending on medical care unrelated to diabetes is included when it should be excluded. Although spending for children and youth under age 20 is omitted, only 0.25 percent of this population has diabetes and the effect is likely to be small. The omission of the institutionalized population is a more serious downward bias on spending estimates because people with advanced stages of diabetes are more likely to be hospitalized or to reside in nursing homes, and their care is costly. The inclusion of spending for all medical care for people with diabetes 20 years of age and over is included in these estimates (rather than only the spending related to diabetes) because medical expenditures by type and age could not be identified readily. This overestimates expenditures related to diabetes only.

The resulting ball-park estimates are shown in Table 2.2 of the Resource Guide. Clearly, a better approach to deriving State Medicaid costs for diabetes care would be to use Medicaid claims, if they were readily available for all States.

Current as of August 2008
Internet Citation: Appendix C: Additional Data Resources Related to Diabetes Care Quality: Diabetes Care Quality Improvement: A Resource Guide for State Action. August 2008. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/professionals/quality-patient-safety/quality-resources/tools/diabguide/diabqguideapc.html