Module 3: Information-Interpreting State Estimates of Diabetes Quality (continued)
Diabetes Care Quality Improvement: A Resource Guide for State Action
Step 2: Interpreting the Data — What Does It Mean?
The data presented in Step 1 raise a number of questions for anyone involved in quality improvement. What does a State's position on the continuum of quality measures mean? What factors influence that position and the variability among the States? What factors can be controlled through decisionmaking and local efforts?
Factors That Affect the Quality of Diabetes Care
A number of factors affect the quality and outcomes of health care, as Figure 3.5 shows. Some factors may be difficult to change, such as biologically inherited traits; income, education, and social status; and general population characteristics. Others may be changeable in the medium or long term, but unchangeable in the short term, such as the supply of health care professionals, the makeup and mission of health care organizations, and the disease prevalence of the population (which represents ingrained patterns of personal behaviors and health system effectiveness or ineffectiveness).
Although State government and community leaders do not have control over many of these factors, there are some areas where implementing action at the State level can increase awareness and promote positive change. These include educating people with diabetes, targeting campaigns about the risks of obesity and sedentary lifestyles to the general public, raising awareness among professionals about health care processes that can improve outcomes for people with diabetes, and creating financial incentives to encourage providers to improve management of the disease.
For example, the CDC's Division of Nutrition and Physical Activity began funding 20 State programs on prevention of obesity in 2003. These programs focus on education of people at risk of diabetes and supportive environments for healthy eating and physical activity. (Information on specific State programs can be found at: http://www.cdc.gov/nccdphp/dnpa/obesity/state_programs/index.htm.)
Other States target minority populations that are disproportionately affected by diabetes in an effort to affect individual self-management and other external causes. Also, many States have passed legislation to secure and regulate insurance coverage for people with diabetes because absence of health care coverage can delay diagnosis, evaluation, education, and proper monitoring and management of the disease with disastrous consequences (go to information at http://www.ncsl.org/programs/health/diabetes.htm).
To better understand what influences a State's position and how it compares with other States, some factors that are presented in Figure 3.5 are discussed in more detail below.
Racial, Ethnic, and Socioeconomic Factors: As previously noted, the socioeconomic makeup of a State will likely play a role in how it compares to national norms on process and outcome measures. States with a higher proportion of individuals living in poverty, lower average education, and a more diverse racial and ethnic population, for instance, will likely find poorer outcomes for their population compared to the national population (IOM, 2003b).
The NHDR (AHRQ, 2003b) summarizes the racial, ethnic, and socioeconomic differences in diabetes across the entire Nation, where minority or lower socioeconomic status is associated with higher diabetes prevalence, higher diabetes death rates, higher rates of serious complications (including end stage renal disease and amputations). Nevertheless, process-of-care measures generally do not differ greatly among white and minority racial and ethnic groups at the national level (go to Table D.2, Appendix D). Absence of differences at the national level does not mean that such differences are nonexistent at the State and local level. Outcomes do differ among racial and income groups at the national level. For example, many more hospitalizations for long-term complications of the disease, including amputation related to diabetes, are seen for blacks compared to whites (Table D.2).
The socioeconomic makeup of a State should also play a role in the strategies that a State uses to improve diabetes care quality. For instance, States may be able to improve diabetes care quality through efforts targeted at population groups particularly at risk for diabetes complications. (The section on Dissemination: Minority and Rural Outreach in Module 4: Action describes approaches being used in some States.)
Biological and Behavioral Factors: The likelihood of developing the most common form of diabetes, type 2, is influenced by both biology and behavior (National Diabetes Education Program [NDEP], undated [a]). Risk factors for type 2 diabetes include:
- Family history of diabetes—Particularly, in the immediate family.
- Gestational diabetes—Women who develop gestational diabetes during pregnancy, children whose mother had gestational diabetes while carrying them (NDEP, undated [b]), and women who gave birth to at least one baby weighing nine pounds or more.
- Age—Risk of diabetes increases with age.
- Overweight/obesity—A known risk factor for diabetes. Overweight is defined as a body mass index > 25 (>23 if Asian American and > 26 if Pacific Islander) and obesity is a body mass index of > 30.
- Lack of exercise—Exercise less than three times a week is associated with developing diabetes and its future complications.
- Diet and nutrition—High calorie intake (proteins, carbohydrates, or fat) increases the risk of developing diabetes and its complications.
Some additional factors that contribute to developing complications in people who already have been diagnosed with diabetes include (NDEP, undated [a]):
- High blood pressure—Pressure greater than 140/90 mm/Hg is associated with increased risk of complications for people with diabetes.
- Abnormal lipid levels—HDL (high density lipoprotein, or "good") cholesterol less than 40 mg/dL for men and less than 50 mg/dL for women and triglyceride level greater than or equal to 250 mg/dL are danger signs of complications for people with diabetes.
Socioeconomic factors may be related to underlying biological factors or behavioral factors. The accumulated stress of poverty, low levels of control in jobs and relationships, low job and life satisfaction, and societal discrimination against minority groups can influence health status (Williams, 1999).
External Environment: In addition to individual characteristics (some of which are amenable to change with personal motivation), each State has different infrastructure and other environmental factors over which policy-makers may or may not have control. These factors include the collective health status of the population, the distribution of health care services within locales, distribution of wealth and tax resources among communities, and government programs and leadership.
State leaders will face different health care system challenges, including:
- Health system infrastructure—Availability of health professionals, emergency rooms, and hospitals beds.
- Uninsured populations—The presence of vulnerable and uninsured populations and the need for special State programs to cover the cost of health care for them.
- Safety net infrastructure—The availability of a safety net of health care providers as a last resort for those who cannot afford health insurance and private health care.
- Provider knowledge—Providers who are not up to date with state-of-the-art knowledge to manage diabetes effectively and of patient education programs to help patients learn to manage their diabetes.
- Public education—The need for public education programs that raise patient awareness of the warning signs of the disease, its potential complications, the importance of diet and exercise, and the effectiveness of personal self-management, including knowing when to consult a doctor.
- Government resources—The funds, in a time of tight State budgets, to stimulate quality improvement activities related to diabetes care.
- Leaders to champion quality improvement— Those leaders who can draw attention to the problems associated with diabetes and harness the commitment of health professionals to change practices and monitor results.
- Knowledge of what to do—The identification of effective quality improvement programs that are based on scientific evidence.
- Adequate data systems to assess progress— Availability of data systems that can provide comparable comparisons across providers, communities, and even with other States.
The inter-relationship between all of the factors in Figure 3.5, then, affects how a State compares with other States on measures of diabetes care quality. It is difficult to measure all of these factors at the State or local level and to analyze and show their effect with data.1 One analysis of the NHQR compares hospital admissions for uncomplicated, uncontrolled diabetes to State environmental factors that are readily available — measures of poverty, obesity, and diabetes prevalence.2 This analysis was possible because 14 States in the Healthcare Cost and Utilization Project3 provided their State discharge data for inclusion in this analysis for the NHQR.
Figure 3.6 shows the resulting associations among admissions for uncontrolled, uncomplicated diabetes and rates of obesity, poverty, and diabetes prevalence. Diabetes prevalence does not vary much across the States, but obesity and poverty rates do. Admission rates also vary greatly across these States; most of these State admission rates are significantly different from the national average, and the low-to-high rates differ fourfold in magnitude. Furthermore, States that have very high admission rates have higher obesity and poverty rates than the States with lower admission rates.
Yet, as noted earlier in this module, poverty and obesity alone do not account for all the differences between States in rates of avoidable hospitalizations for diabetes. Other factors certainly play a role. The health system infrastructure, rate of the uninsured, provider knowledge and incentives, public education, funding and leadership, knowledge of what to do, and information systems—all will affect the challenges that State leaders face in leading communities to improve health care for people with diabetes.
Interpreting Process and Outcome Measures Together
The four States presented earlier in the State-level comparison are included in Figure 3.6. Examining these States in terms of process measures, this one outcome measure, and underlying population characteristics is instructive.
- Georgia, which had better HbA1c testing rates for two or more times per year than the other four States, also has very high rates of avoidable hospitalizations for uncomplicated, uncontrolled diabetes. This suggests the need to examine the adequate of ambulatory care; perhaps HbA1c testing is not translating into improved glycemic control for patients. Georgia has one of the highest rates of poverty (usually correlated with lower education) among the States; perhaps additional targeted patient education would be beneficial. Furthermore, Georgia ranks third among States in medically underserved or health personnel shortage areas (Hawkins and Proser, 2004). This suggests that less access to ambulatory care in some areas may lead to more hospitalizations for early stage diabetes. Whenever process and outcomes measures do not agree, they should be examined critically in the context of the State environment.
- Massachusetts, which had process rates that were not distinguishable statistically from the national average but that were notably higher on HbA1c testing rates for two or more times per year and influenza immunization, has one of the lowest rates of uncontrolled diabetes hospitalizations among the 14 States. Massachusetts' population also has lower rates of the underlying problems of obesity and poverty compared to other States.
- Michigan, which had process-of-care rates indistinguishable from the national average and on the lower end of diabetes care quality, had a moderately low rate of these avoidable hospitalizations. This is despite the fact that Michigan has a population with high obesity rates (but not high poverty rates).
- Washington, which had process measures that were fairly similar to the national average with the exception of its high immunization rate, had one of the lowest rates of these avoidable hospitalizations. Washington's population has one of the lowest poverty and obesity rates.
These combined views of diabetes care in the States suggest that the underlying populations and personal risk behaviors and perhaps self-management of the disease have more of an effect on the outcomes of care than whether or not a particular test is given. The test itself is not sufficient for improving diabetes outcomes. Complicated interactions of many factors influence diabetes outcomes. Furthermore, often the results on one measure are not consistent with findings on another measure, even when the measures are related. This indicates the importance of improving information systems that can track problems and enhance understanding of the effectiveness of quality improvement programs.
None of the above analysis tracks State results with their diabetes care quality improvement programs. No full-scale evaluations have yet been published of State interventions in health care quality improvement. Through interviews with State officials, this Resource Guide identifies a few programs that likely have influenced the quality measures discussed here. They are described in detail in Module 4: Action.
Summary and Synthesis
This module shows how data from the NHQR can be analyzed and interpreted to answer the question of how a State compares to other States and national benchmarks on health care quality for one disease — diabetes. Maps and charts can be used to help State leaders and quality improvement teams, whether or not they are trained in statistics and analysis, understand where their State stands in terms of diabetes quality.
A key question for all States is: What goals should the State set as targets for specific diabetes care quality measures? The NHQR can be used to identify consensus-based measures, as shown in Module 2: Data. States may identify and define other measures as well. The advantage of the NHQR measures is that the best-in-class State estimate, which can be derived easily from the NHQR, shows what has already been achieved by some States. It is a reasonable target for most measures. However, some measures might be so crucial to good diabetes outcomes that the target should not be limited by what other communities have achieved to date. Improvements above and beyond the best-in-class States may be warranted. Experts in diabetes care and local community leaders can help make these types of judgments.
Another key question is: Are all States able to meet the challenge of the best-in-class States? The answer depends on the measure, the factors that relate to that measure (health system versus consumer actions), and the current health and socioeconomic status of the State population. The analyses in this module reveal that many factors influence diabetes care, making the assessment between diabetes outcomes and processes of care difficult to affirm. Nevertheless, State-level baseline estimates of diabetes care enable States to assess their starting point and to evaluate their progress over time.
Some States may be able to assemble better data than are available nationally to understand the quality of care in their State. This has been done in some States (go to for example Michigan in Module 4: Action. State leaders can assess the quality of diabetes care using the NHQR data to obtain an idea of where their State stands in comparison to other States and the Nation. One thing is clear from the NHQR data and the information this module derives from it—no State measures up to all the guidelines for diabetes care completely. The next module provides insights on what actions some States have taken to improve diabetes quality.
Finally, diabetes is only one of many conditions that warrant improvements in health care quality. While this Resource Guide focuses only on diabetes, State leaders will ask: What other conditions are ripe for care measures for all diseases examined in the numerous tables of the NHQR by State. Thus, assembled in one place, State leaders can scan the list of measures to see how their own State compares to the national average across all NHQR measures. Once diabetes quality improvement is on track, State leaders may want to start with Appendix F to inspire their next campaign to improve health care quality.
Associated Appendixes for Use With This Module
Appendix D. Benchmarks From the NHQR
Appendix D provides additional detail on benchmarks and how they were developed and defined for this Resource Guide. It also explains the best benchmarks for stimulating quality improvement. This appendix notes that methods used to generate the benchmarks must be understood to ensure they are compatible with a State's estimates.
Appendix E. Information on Statistical Significance
Appendix E shows how to compare State estimates to benchmarks using statistical significance and p-values that take into account the expected random variation in estimates. This appendix also shows how to calculate p-values when estimates and standard errors are provided and when estimates, and thus standard errors, must be derived from the data provided.
Appendix F: NHQR Quality Measures for All Conditions by State
Appendix F lists quality measures for all conditions and topics in the NHQR. It includes the national estimate and then an indicator for whether or not the State estimate (not shown due to space limitation) is statistically greater, lower, or no different from the national average. The measures for which State-level data are not reported in the NHQR are excluded from the table. This resource can help State leaders identify which diseases, in addition to diabetes, are in need of quality improvement. Many of the same data issues related to diabetes are applicable to other disease topics, although different data sources and limitations may apply to them.
1 This State-level analysis is feasible because of information collected at the State level. Similar analyses may be possible for smaller geographic areas within States. For example, the HCUP data, described below, permit analyses at the county or finer market areas. Data related to health care resource are generally available at the county level, although data on health risk behaviors of the population generally are not. State analysts could use their county level databases to compare diabetes quality measures based on HCUP data with other characteristics of counties.
2 Diabetes prevalence, poverty and obesity rates were selected because they were most closely related to admissions for these avoidable hospitalizations among a set of other factors studied (including age of the population, insurance coverage, and health resources).
3 HCUP Partners providing their data for this analysis were: Arizona Department of Health Services, Colorado Health & Hospital Association, Georgia Hospital Association, Hawaii Health Information Corporation, Iowa Hospital Association, Kentucky Department for Public Health, Maine Health Data Organization, Massachusetts Division of Health Care Finance and Policy, Michigan Health and Hospital Association, Missouri Hospital Association, Texas Health Care Information Council, Washington State Department of Health, West Virginia Health Care Authority, Wisconsin Department of Health and Family Services.