One particularly promising strategy emerging from NHPC efforts to address disparities is the development of interactive mapping and analysis tools. These tools help plans quickly identify geographic areas with characteristics that signify good opportunities for interventions. For example, software tools and algorithms, such as those developed by RAND for the NHPC, enable plans to highlight census tracts that have a high volume of members with diabetes from a given race/ethnic group who have not received one or more recommended elements of care. Figure 1A shows such a map of a health plan's market area, and Figure 1B provides more detail on this area. These maps serve, in some fashion, as a "geographic Pareto chart," and a plan can use this information to focus more efficiently on a variety of interventions, ranging from targeted mailings to providers in this area to communitywide education.
The effectiveness of displaying complex data through maps, such as those shown in Figures 1A and 1B, rests on a number of general principles, including making large data sets coherent and encouraging the viewer to make comparisons by region and race/ethnicity1. For example, Figure 1A shows that the Hispanic diabetic members in one plan who are not receiving LDL tests tend to be clustered in a relatively small number of areas. The Pareto chart in Figure 1B shows that four of the clusters account for a significant proportion of disparities observed in that market; clusters (C, E, F and A) account for 80 percent of the Hispanic diabetics in that market not receiving LDL tests. This type of information has helped plans focus more efficiently on where they may want to implement an intervention.
Several plans now use these tools to target and develop interventions and have noted that the tools provide a way to focus resources, enabling interventions that would have otherwise been cost prohibitive. As plans have grown more sophisticated about the possibilities for assessing and acting on disparities, they also have noted the need for measures of other types of information that may contribute to disparities.
GIS tools make it easy for plans to begin to assess these potential contributing factors. For instance, Figure 2 highlights how GIS tools can help different stakeholders and decision-makers test their working hypotheses about factors that may explain the observed pattern. Figure 2 shows an enlarged view of a portion of Cluster C, a predominantly Hispanic Area (>90 percent of residents are Hispanic) with relatively high rates of health plan diabetic members who had not received LDL tests. At the local level shown, it is apparent that the patterns of care are not as homogenous across census tracts as one might assume based on the overall low performance rate for the cluster.
When plan decision-makers are initially presented with such maps and asked to speculate why they think Hispanic members living in some census tracts tend to receive worse care (as indicated by the darker shading) than Hispanic members living in nearby census tracts, the most common answer is that Hispanic members in the areas receiving worse care tend to be poorer than those receiving better care. However, upon closer examination of Figures 2B and 2C, decision-makers could easily see that poverty levels do not appear to play a large role in the observed care patterns shown in Figure 2A. Rather, levels of linguistic isolation appear to have a much larger role. Indeed, showing the data on maps such as these helped convince decision-makers that linguistic isolation is an important consideration in any intervention they might devise and helped plans identify specific census tracts or neighborhoods where they might want to target their efforts. Though not shown in this example, the map data generally is presented in combination with other types of information (e.g., summary tables of subgroup characteristics and statistical associations) to further clarify the observed patterns.
Prior to seeing this type of information mapped, decision-makers with quality improvement backgrounds tended to assume that when aggregate data (e.g., within a member service area or region) showed a consistent disparity between the receipt of indicated care by African-American or Hispanic members versus White members, their strategy should simply be to target all members belonging to the disadvantaged racial/ethnic group. Based on that logic, for example, one health plan’s initial strategy to address an observed disparity in LDL test rates between diabetic Hispanics (60 percent) and Whites (70 percent) was to develop and mail new member education materials about diabetes to all members residing in predominantly Hispanic census tracts. However, maps highlighted local variation in quality of care between neighboring census tracts (all of which were predominantly Hispanic), leading the plan to modify its approach, resulting in a better-targeted, less costly intervention.
1. Lurie N, Fremont AM, et al. “The National Health Plan Collaborative to Reduce Disparities and Improve Quality.” The Joint Commission Journal on Quality and Patient Safety, 34 (5): 256-265, 2008.
- 1. What Categories of Race/Ethnicity to Use?
- 2. Direct REL Data Collection Methods
- 3. Section 5: Case Studies
- 3.1. Harvard Pilgrim Health Care: Pilot Test of IVR Outreach Calls as a Mechanism for Collecting REL Data
- 3.2. Molina Healthcare's TeleSalud Program: Providing Direct Access to Language Services
- 3.3. Kaiser Permanente: Qualified Bilingual Staff Model
- 3.4. Kaiser Permanente: Health Care Interpreter Certificate Program
- 3.5. The National Health Plan Collaborative to Reduce Disparities and Improve Quality
- 4. Chapter 5: Promising Practices in Interpreter Training and Competency Assessments