Module 2: Data - Understanding the Foundation of Quality Improvement (continued)

Understanding the Foundation of Quality Improvement (continued)

Filling Local Data Gaps

Finding data is a challenge for quality improvement programs. Two avenues can be used to locate relevant data: 1) developing an inventory of local data sources and 2) using published research to generate local estimates. The latter (generating local estimates) is acceptable for planning purposes, until better local sources are located and analyzed. One of the best sources for filling the data gaps will be the State DPCP staff.

Developing an Inventory of Local Data Sources

Local data (whether State, county, municipal, or individual health care provider data) are essential for quality improvement programs to have an impact locally. Local leaders and health care professionals must see their own data in comparison to other providers and to State, regional, and national benchmarks in order to appreciate the importance of their work.

Health care quality improvement programs should develop a complete inventory of data systems available at the State and local level. Doing so may reduce data-related costs and avoid duplicate data collection. Also, a review of local data in the context of the NHQR and NHDR should make clear where existing local surveys or data systems could be modified to add information comparable to the concepts used in those reports and, thus, to provide the raw materials for insights into health care and its quality at the local level.

Most States have data systems that can contribute to a review of health care quality at the State or local levels. Some of those data systems include:

• BRFSS data, available at the State level through the State health department.
• Statewide inpatient hospital discharge systems, for which HCUP and NHDS data can provide uniform national comparisons.
• State vital statistics, which include mortality rates by cause of death and for which the National Vital Statistics System can provide uniform national comparisons.
• Special disease registries, some of which are focused on diabetes.
• Other special data collection of State departments of health statistics and other State programs.

Specific data systems for populations that the State supports are also available in most States. These include:

• Medicaid information systems based on health care provider claims for reimbursement from Medicaid.
• State employee health benefit claims for reimbursement.
• Patient records from State- or county-run programs, such as mental health and substance abuse programs or school health programs.

Some examples of State-level data sources are available at the National Association of Health Data Organizations' Web site (http://www.nahdo.org/cs/). Other Federal or national systems compile data that describe State and local populations or health resources. These include:

• The CDC's Division of Diabetes Translation Web site, a valuable starting place to identify data and become familiar with the network of organizations and individuals associated with diabetes data collection at the State and national level (http://www.cdc.gov/diabetes/statistics/index.htm).
• Census population data by State, maintained by the U.S. Bureau of the Census (http://www.census.gov/popest/states/)
• The Area Resource File, a county- and State-level database of health care resources from several surveys and data sources, compiled by the Health Resources and Services Administration (HRSA). Quality of care in managed care organizations, provided through the National Committee for Quality Assurance (go to: http://www.ncqa.org/). (Local managed care organizations can be an important source of local data on health care quality).
• The Henry J. Kaiser Family Foundation's Web site (http://kff.org/statepolicy/index.cfm), a rich source of health and other information at the State-level compiled from many public databases and published studies.

Using Published Studies and Readily Available Data To Develop State or Local Estimates

Before resources are invested in data collection targeted to an improvement goal, some information can be assembled from existing sources and published research studies. Sometimes published studies on a topic can be used to derive estimates at the State or local level. These "ballpark estimates" should be replaced by more accurate local data when they are available.

To assess the impact of diabetes on the State, studies of diabetes nationally might be used. For example, if a national study shows how subgroups differ in diabetes prevalence or costs and provides estimates by those general subpopulations (e.g., age groups), then those general subpopulation characteristics in the State (or locale) can be applied to the national rates, thus resulting in State (or local) estimates for diabetes.

The more detailed and compatible the data are across sources, the better the estimate will be. However, existing data details are seldom sufficient, which limits the confidence of estimates that can be made from existing tables and published estimates. When this is the case, original analyses of the underlying data may be necessary. When actual data are available from State agencies for all or part of the information components, they are preferable to estimations from national data.

Two examples of deriving State estimates from national data and studies are presented here: 1) Medicaid spending on diabetes care, and 2) total cost burden of diabetes, by State.

Example One—Medicaid Spending on Diabetes Care: This example derives estimates of spending on diabetes care for State Medicaid agencies using the following components:

• National diabetes prevalence by age and by race/ethnicity separately.
• State Medicaid populations by age and by race/ethnicity separately.
• National expenditures related to diabetes for a younger and older adult population from a published study to derive the estimates.

 Table 2.2 shows the estimated expenditures. They are ballpark estimates of such spending likely occurring across State Medicaid agencies. (Figure C.1 in Appendix C charts the flow of data, assumptions, and calculations made to devise the Medicaid spending estimates for diabetes.)

Although the Medicaid population is primarily women and children, the diabetes population is disproportionately elderly. Data from each source were reconfigured to reflect the same underlying population and adjusted to reflect the same year of reference to make data compatible across sources. Because prevalence and cost are so different by age, the estimates were first generated separately for the adult nonelderly population and the elderly population and then were reassembled. Children and youth under 20 were excluded because certain pieces of information were unavailable for them and because prevalence of diabetes (type 1 and type 2) among them is small (0.25 percent).

Another consideration for diabetes is its higher prevalence among certain racial and ethnic groups. Prevalence rates by race/ethnicity were applied to those respective subgroups of Medicaid. Also, Medicaid enrollees of unknown age or race/ethnicity were distributed in proportion to the known age or known race/ethnicity subgroups. Finally, data from different years were adjusted to be compatible.

The estimates in Table 2.2 have limitations. The obvious limitations in these estimates include omission of spending for children and the institutionalized population. First, 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. Second, 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. Third, however, for people dually eligible for Medicaid and Medicare (which is most of this Medicaid population over 60 years of age), some of the expenditures for diabetes will be paid for by Medicare and not by Medicaid, which results in higher estimates here than should be the case. The net effect of these latter two offsetting biases cannot be determined from these data. Fourth, 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 care. The net effect of all of these limitations is unclear. What is clear is that a State's Medicaid data will be a more accurate source for calculating expenses for Medicaid related to diabetes.

One should note that the estimates presented in Table 2.2 are approximations to State Medicaid spending on diabetes. Estimates calculated from State Medicaid information systems for diabetes prevalence and actual Medicaid payments would be more accurate.

The estimates here can be useful for understanding the implications of diabetes for health care costs and the possible returns from investment in diabetes care quality. States governments (e.g., State Medicaid Directors) may have actual costs of diabetes for their population. If so, then these actual costs would be preferable to estimates based on national averages from various data sources. Corroboration from external sources can increase the confidence in State and local estimates based on difference methods.

Example Two—Estimates of the cost burden of diabetes for each State: This example estimates the total cost of diabetes care for each State's total population. The total cost of diabetes care includes its direct and indirect costs. Direct costs are directly associated with treatment of the disease, including medical expenditures for routine services, treatment of complications, and the increase in general medical conditions attributable to diabetes. Indirect costs are dollar estimates associated with decreased productivity, disability, and premature death. At the end of this section is an exercise for calculating a State's costs with different assumptions that might be generated from State data.

 Table 2.3 shows estimates of the cost of diabetes for each State's total population using readily available data and following the methods of Hogan, Dall, and Nikolov (2003). This is a more direct calculation than the Medicaid calculation because a State's total population is more likely to have characteristics similar to the total U.S. population than is the Medicaid population.

Table 2.4 is a step-by-step exercise that shows how the estimates were generated; it provides a guide to States who want to use different assumptions. The data needed include: the size of the State population, the prevalence of diabetes in the State, and estimates of the cost burden. For the estimates in Table 2.3, the State populations are from the U.S. Bureau of the Census (go to: http://www.census.gov/popest/states/NST-ann-est.html). State-level diabetes prevalence is available through the CDC at: http://www.cdc.gov/diabetes/statistics/prev/state/table15.htm.

The direct and indirect costs of medical care for individuals with and without diabetes were estimated for the Nation by Hogan, Dall, and Nikolov (2003). Although they used diabetes prevalence estimates from the National Health Interview Survey (NHIS), the estimates in Table 2.3 use the CDC's BRFSS prevalence data because they were available by State. Thus, the estimates of State-level direct and indirect costs when summed across all States differ slightly from the Hogan and colleagues' national estimate of cost burden.

For direct cost per person with diabetes, estimates from Hogan et al. are used. Their total direct cost burden per person with diabetes in 2002 is $13,243. The age-adjusted estimate of the direct cost of care per person without diabetes is $5,642. The $7,601 difference is used in Table 2.3 to net out the regular medical care costs for patients with diabetes (that is, cost unrelated to diabetes and its sequelae). The $7,601 cost is then multiplied by the State diabetes prevalence to derive the State estimate for the direct cost of care for diabetes.

For indirect cost per person with diabetes, the Hogan et al. estimate ($3,289 annually) is multiplied by the State diabetes prevalence to derive the State indirect cost estimate. The total cost burden is the sum of the direct and the indirect diabetes costs for each State.

 Table 2.4: Estimating the cost burden of diabetes for a State in 2002

Return to Contents

Summary and Synthesis

This module orients users to the importance of data as the foundation of the quality improvement cycle. Data are essential for assessing the situation and measuring the impact of a quality improvement project. Without it, State leaders could spend effort and resources without accomplishing the most important goal-improving the health outcomes of their residents. Data, used effectively, should guide the quality improvement process and enhance a team's effectiveness in focusing on the right goal and making the right decisions.

Module 2 describes two components of data collection for quality improvement: 1) measurement and 2) data sources. The National Healthcare Quality Report and the National Healthcare Disparities Report now provide easy access to the health care quality measures and related data sources that are national (and sometimes State-level) in scope. This module highlights the diabetes-related measures and data in those reports.

Important considerations when using data include data limitations and making certain that comparison data are truly comparable to the State-level data. Taking an inventory of existing State and local data sources and using existing national data and studies can help to fill in gaps in local data, at least in the planning stages of a quality improvement program.

Once data have been identified or collected, the next step is analyzing and translating those data into information that can be used to make policy-level decisions. Module 3: Information interprets the data from a State perspective and begins to explore its meaning.

Return to Contents

Resources for Further Reading

Data and Data Tools on the Internet

Many data resources are available on the Internet, including many sources used in the NHQR and NHDR. Some Web sites allow users to manipulate the data to produce tables and other useful outputs. Such resources include:

• HCUPnet http://hcupnet.ahrq.gov

  HCUPnet allows users to select national statistics, or detailed statistics for certain States, for various conditions and procedures. The interactive program also allows users to compare types of patients and types of hospitals.

• HCUP User Support (HCUP-US) http://www.hcup-us.ahrq.gov/home.jsp

  This Web site is designed to answer HCUP-related questions; provide detailed information on HCUP databases, tools, and products; and offer assistance to HCUP users.

• MEPSnet http://meps.ahrq.gov/mepsweb/

  This Web site offers users statistics and trends about health care expenditures, utilization, and health insurance, including national and regional health insurance estimates.

• BRFSS Annual Survey data http://www.cdc.gov/brfss/technical_infodata/index.htm

  This Web site has detailed technical information about the survey in addition to downloadable data sets in ASCII and SAS formats.

• BRFSS http://www.cdc.gov/brfss/

  This Web site provides useful background information about the BRFSS implementation, technical information, and documentation.

• DATA2010 http://wonder.cdc.gov/data2010/

  This Web site includes data from a number of different State and national data sources and can be used to monitor the objectives for Healthy People 2010.

Diabetes Registries

Some additional Web sites offer links to useful tools and information to facilitate data collection at the local level. Two Web sites that offer instruction for implementing disease registries to track the treatments received by people with diabetes and other chronic conditions are:

• http://www.healthdisparities.net/

  This Web site, associated with the HRSA Health Disparities Collaboratives, offers a number of useful tools, including helpful information for creating and assessing computer registries.

• http://www.chcf.org/documents/chronicdisease/ComputerizedRegistriesInChronicDisease.pdf

  This Web site offers a primer on the use of disease registries for a variety of chronic conditions, including diabetes.

Other Useful Web Sites

• Agency for Healthcare Research and Quality: https://www.ahrq.gov/
• AHRQ Quality Indicators http://www.qualityindicators.ahrq.gov
• National Committee on Quality Assurance: http://www.ncqa.org/
• National Diabetes Quality Improvement Alliance: http://www.nationaldiabetesalliance.org/ (No longer accessible).
• National Quality Forum: http://www.qualityforum.org/

Return to Contents

Associated Appendixes for Use With This Module

Appendix A: Acronyms Used in This Resource Guide

The acronyms employed to describe the organizations endorsing the NHQR quality measures are described in Appendix A, along with all other acronyms used throughout this Resource Guide.

Appendix B: List of All NHQR Data Sources, Including Those Supporting State Estimates

Appendix B lists the 25 data sources used in the NHQR and highlights the 10 data sources that provided State-level data in the NHQR.

Appendix C: Additional Data Resources Related to Diabetes Care Quality

Appendix C lists additional data resources that may be helpful in studying diabetes care in a State. It includes separate sections, with accompanying tables, on the NHQR measures selection process (go to Table C.1), details on data source description and limitations (Tables C.2-C.10), and steps for estimating Medicaid spending on diabetes care by State (Figure C.1). Details on notable differences between MEPS and BRFSS national rates are included, as well as further information on the data sources for the process and outcome measures discussed in this module.