Editor’s note: David Langley is director of strategic research & analysis, and Lori Cook is a senior project manager, at Blue Cross Blue Shield of Maine, Portland. This article is the third in a three-part series designed to provide real-world business examples of the effective use and application of research and statistical tools for supporting resourcing and priority-setting decisions. These research application issues have arisen through the authors’ work with regional and national studies in health care and other industries. Each of the three articles in the series provides a summary review and example of how marketing research techniques, when approached as a credible discipline and with a clear view of specific decision support needs, can inform executive decision making. The first article in this three-part series, “Effective uses of ‘effect size’ statistics to demonstrate business value” appeared in the October 1999 issue of Quirk’s. The second article, “Consumers’ contradictions: ‘value’ and other brand attributes” appeared in the March 2000 issue of Quirk’s.

Many organizations have long-standing experience using customer segmentation for marketing purposes (e.g., decision-making and resource allocation around product design, sales, promotional strategies). In this context, “…segmentation is the process of partitioning markets into segments of potential customers with similar characteristics who are likely to exhibit similar purchase behavior.”1

This article outlines a recent health care case study of the use of segmentation tools and techniques for developing and refining service delivery through satisfaction-based quality improvement initiatives. In this case study, health plan customers (i.e., members2) are distinguished and segmented by their health status through a multi-method process for classification and assessment.

Health status and satisfaction

Numerous national and published health care studies have demonstrated a link between health status and satisfaction with health care and health plan. Consistently, these data point to differences in individuals’ ratings of their health care, providers, and health plan, based on health status. For example, individuals in poorer health report lower overall satisfaction with their health plans, more problems with health plan administrative services, lower satisfaction with access to health care, and lower satisfaction with quality and outcomes of care. Examples of research findings include the following:

  • Field tests with a standardized national health care survey - Consumer Assessment of Health Plans (CAHPS)3 - have demonstrated that “consumer ratings about health care were consistently higher for those in better health. Health status correlated with overall rating of health care, rating of personal doctor or nurse, and rating of the health plan, respectively, in a sample of chronically ill adults.”4 As a result of these finding, CAHPS data, when used for comparing health plans on the basis of the ratings by individuals covered by those plans, is case-mix adjusted to minimize systematic bias based on this patient characteristic.
  • Similarly, results of a national study (based on a longitudinal survey of 5,700 employees) indicated that employees in poorer health were less satisfied with their health plans than employees in better health. Relative to those in poorer health, healthier respondents were 5 percent more satisfied at baseline (1993), and 4 percent more satisfied at remeasure (1995).5
  • Hall, Milburn, Roter and Daltroy, in a 1998 study of patient satisfaction with medical care, cite more than 15 studies demonstrating the finding that “patients in poorer health, either emotionally or physically, tend to be less satisfied with their medical care.”6

Although there is no definitive understanding of why individuals in poorer health are less satisfied with their health plans and their health care than healthier individuals, the consistency and prevalence of these findings points to the importance of taking health status into consideration as a critical characteristic when assessing and addressing the needs and satisfaction of member/patient populations.

Health status as a segmentation variable for service quality improvement

In this health plan’s case study, health status is effectively used as a segmentation variable for health plan service quality improvement. The methods used in this case study for classifying health status include self-reported data from member surveys as well as the use of medical claims data. For purposes of establishing dichotomous “segments,” members who report being in poorer health or who are associated with medical claims for selected health conditions are distinguished from members who do not have these characteristics.

A multi-method approach to health-based service quality improvement includes the following assessment methods:

  • Statistical analysis of quantitative research findings to: a) define and assess experiences and perceptions of members by health status; and b) define and understand key satisfaction drivers relevant to the population of members in poorer health. Analytic tools include: crosstabulations, significance testing (t- and z-tests), factor analysis, regression analysis, and cluster analysis.
  • Use of claim-based operational data to segment and analyze customer inquiry rates/reasons by health status.
  • Qualitative validation through member and staff focus groups to further understand the needs and experiences of the population of members in poorer health.

A. Quantitative analysis

1. HEDIS7 3.0 Member Satisfaction Survey

A segmentation analysis of the relationship between health status and member satisfaction was conducted using 1998 HEDIS 3.0 Member Satisfaction Survey data. This annual mail survey of managed care members, administered through the National Committee for Quality Assurance (NCQA) standardized study design, provides comparable member satisfaction and health status data to health plans, purchasers, and regulators. The 1998 version of the survey included a battery of self-reported health status items, including the SF-128 and a chronic illness checklist.

Analyses of the data (i.e., statistical significance testing) were conducted to determine whether survey results of respondents in poorer health differed from those of healthier members. Physical health status and mental health status were considered separately. For the purposes of these analyses, “poorer physical health” was defined as 25th percentile on the SF-12 physical health scale and/or one or more chronic conditions based on the chronic conditions checklist; “poorer mental health” was defined as 25th percentile on the SF-12 mental health scale.

Findings of these analyses pointed to a clear and significant link between health status and plan satisfaction (see Fig. 1):

Figure 1

  • Respondents in poorer physical health were statistically significantly less satisfied overall with the plan, and less satisfied with:

    - quality of health care;

    - access to health care;

    - plan administration.

  • These respondents also reported significantly more contact with the plan, and were more likely to have contacted the plan specifically with a complaint or problem.

This analysis also identified unique differences in respondent satisfaction based on mental health status.

2. Key drivers analysis

A drivers analysis was conducted with the 1998 HEDIS 3.0 data to develop differential models of satisfaction drivers based on health status. This statistical modeling included factor and regression techniques to identify significant predictors of satisfaction.

Figure 2

Based on this analysis, the following key drivers sets were developed (see Fig. 2):

  • Satisfaction drivers for healthier members include the following:

    - benefits, costs, plan administration (including paperwork, availability of information about costs and benefits)(this was the most important driver);

    - plan coordination of care (delays/difficulties receiving care, referrals);

    - health care (quality, access, outcomes).

  • These items are also drivers for members in poorer health. However, they become less important as additional drivers emerge for this population; in particular, the role of “service” takes on added importance:

    - satisfaction with problem resolution when contacting the plan;

    - perceived change in plan performance re: service, access to care, and quality of care (e.g., service provided by customer service reps, problem resolution, claims problems, ease of choosing a personal physician, satisfaction with overall quality of care and services, availability of information about costs and benefits, and wait times for appointments for chronic conditions).

3. Other health plan survey data

Further statistical analysis of the relationship between health status and satisfaction was conducted with health plan data from another available satisfaction study to validate HEDIS 3.0 findings, as well as gain additional information regarding the relationship between health status and plan experience/satisfaction in areas not covered by the HEDIS survey.

Despite limitations in this survey’s number of health status questions, in comparison to the HEDIS 3.0 survey’s battery of health status items, analysis of this 1998 dataset identified marginally significant findings through statistical significance testing. These findings were directionally consistent with those of the HEDIS 3.0 analysis in terms of identifying a link between health status and satisfaction/experience with the plan:

  • Respondents in poorer health were generally less satisfied with:

    - access to care;

    - plan costs;

    - claims;

    - information and communications provided by the plan.

  • These respondents were also less likely to be knowledgeable about plan procedures and benefits (as indicated by several questions in the survey that “tested” member knowledge about specific plan procedures and benefits).

Additional analyses with this dataset included:

  • An aggregated analysis of two years of survey data (1997 and 1998) to assess the impact of health status on satisfaction. This analysis indicated that, despite the sparse measurement of health in this survey, the aggregate impact of health on satisfaction was significant, and that respondents in better health were roughly 10 percent more satisfied than respondents in poorer health.
  • A cluster analysis that clustered respondents based on reported problems with the plan and overall plan satisfaction. This analysis determined that the segment of respondents reporting the most problems and least satisfaction with the plan also had significantly lower health status ratings than other segments.

In 1999, NCQA replaced the HEDIS 3.0 survey with the HEDIS/CAHPS 2.0H instrument: this new instrument no longer contained the battery of health status items included in the HEDIS 3.0 instrument. Therefore, in order to be able to continue to identify and monitor the impact of health on satisfaction, the 1999 version of this additional health plan study was enhanced to include five of the 12 SF-12 items, as well as a chronic illness/risk checklist that focused on identifying high-cost/high prevalence medical conditions (as noted above, the 1998 version of this survey contained only a limited set of health status measures). Analysis of these data provided a more robust view of the differences in satisfaction by health status (see Fig. 3).

Figure 3

B. Operational data

In addition to analyses of quantitative research findings, an analysis of plan operational data was conducted to: a) identify the population of members in poorer health; and b) determine specific needs and issues of this population through analysis of inquiries.

As a starting point, claim-based data was used to identify members with selected high-cost/high-prevalence illnesses or health risk. With this claim-based data then linked to the health plan’s on-line customer inquiry system that tracks and categorizes all member contacts, further analyses were conducted by health status using inquiry rates and reasons for these inquiries.

Figure 4

Results indicate that members in poorer health call the health plan at approximately twice the incidence rate of healthier members (see Fig. 4). Discussion and feedback from service staff regarding these data further suggests that this differential contact rate is not simply attributable to this population using more services (and therefore requiring more contact with the plan), but that this population has unique needs and experiences with the plan that contribute to this higher contact rate.

C. Qualitative validation

Focus groups of customers and service staff were used in 1999 and 2000 to gain a more in-depth understanding regarding the needs and experiences of the population of members in poorer health by structuring the discussions to: validate survey findings, validate and operationalize the key drivers sets identified through statistical modeling, drill down on experiences and issues of this population, and explore potential solutions for effectively addressing dissatisfiers.

Primary learnings from these focus groups included the following:

  • Members in poorer health appear to have varying experiences with the system: While some members report few problems (e.g., they appear to be highly educated about their health plan and its procedures and have good working relationships with their health care providers), others appear to have a more difficult time understanding how their health plan works and how to negotiate the health care system.
  • There are a number of barriers to smooth care for this population. Some administrative procedures that work smoothly for members in good health and for those who are informed and pro-active regarding their health care can create “hoops” for the very sick.
  • The time of needing health services is often a time of crisis. Effective education and communication is critical for this population; this segment appears to need access to information at the point of needing services.
  • This population for whom administrative procedures are problematic has a need for a more coordinated approach to care and health plan services.

Implications/next steps

Learnings from this multi-method assessment process have clearly pointed to the differential experiences and satisfaction levels of the segment of members in poorer health. Relevant findings and implications for targeting service improvement include:

  • Members in poorer health contact the health plan service center at approximately twice the incidence rate of healthier members; they also report lower satisfaction and more problems with their health care. Although service is important for all members, the role of service takes on added significance when members experience problems with their health.

The experiences of members in poorer health appear to vary based on a number of factors (e.g., education, working relationships with providers). In general, however, the time of needing services is often a time of crisis. These members need:

- effective, timely access to information about plan benefits and procedures;

- access to smooth delivery of care and services;

- these members need to experience a simplified, easy-to-understand, and coordinated process when interacting with their health care providers and health plan.

In order to address these issues, health plan initiatives have been implemented to further integrate and coordinate the delivery of care and services. As part of the efforts and to effectively target improvement activities to specifically meet the needs of the target population, further analytic work is being implemented:

  • What is the specific and actionable causal relationship between health status and satisfaction? Although numerous data point to the association between health status and satisfaction, to date there is no clear evidence in the literature regarding why individuals in poorer health are less satisfied. Examples of present hypotheses include the following:9

- These individuals give poorer ratings in general.

- Some individuals are likely to give negative ratings about anything, including their health and medical care they receive (correlated error).

- Individuals in poorer health get worse care.

- Patient complexity makes it harder to deliver care to those in poorer health.

- Members in poorer health have more contact with plan/providers.

- Plan utilization review discriminates against those in poorer health.

- Good health outcomes lead to increased satisfaction, while poor health outcomes lead to decreased satisfaction even when medical treatment is caring and competent.

- Good patient-doctor relationships product good outcomes, which lead to increased satisfaction.

  • What is the appropriate definition of the member population in poorer health? Although the classifications and the resulting segmentation used in this case study offer a clear direction for improving service delivery and for shaping further analytic work, they are operationally problematic, and do not identify a specific population that can be actionably targeted for improvement activities.

  • What particular set of plan experiences and contact contributes most to dissatisfaction? Although drivers analysis and qualitative data have provided insights into the types of issues that cause dissatisfaction, more definition is needed around the specific processes or events that contribute to an overall pattern of difficult interactions between member and plan.

Be attentive

This work, both analytic and applied business decision making, is instructive for understanding how marketing research tools (customarily used to support marketing initiatives) can be effectively used to guide understanding of customers’ service needs. Delivering appropriate, competitive, and best-practice levels of customer satisfaction is clearly dependent on service functions that meet each customer’s unique set of needs. For this work to be useful and relevant, the researcher must be attentive to the careful use of correct analytic methods and the current literature regarding these tools. Likewise, the business decision maker must be thinking analytically about the direction being offered by study findings and integrating these with their practical, day-to-day understandings of customers’ needs and expectations.

Suggested readings on market segmentation tools and techniques

Weinstein, A. Market Segmentation: Using Niche Marketing to Exploit New Markets. Chicago, Ill.: Probus Publishing Company, 1987.

Blankenship, A.B. and Breen, G.E. State of the Art Marketing Research. Chicago, Ill, : American Marketing Association; NTC Business Books, 1993.

Etzel, M.J., Walker, B.J., and Stanton, W.J. Marketing, 11th Edition . McGraw Hill Higher Education, 1997. (See Chapter 7.)

DSS Research Inc. Web site: www.dssresearch.com/ library/Segment/understanding.asp.

References

1 Weinstein, A. Market Segmentation: Using Niche Marketing to Exploit New Markets. Chicago, Ill.: Probus Publishing Company, 1987, p. 4.

2 The term “member” is a convention in widespread use by managed care organizations and other health plans. A “member” is an insured person, whether policyholder, dependent, or other.

3 The CAHPS survey was developed by a consortium of Harvard Medical School, RAND, and the Research Triangle Institute and sponsored by the Agency for Health Care Policy and Research (AHCPR).

4 Agency for Healthcare Research and Quality Web site: www.ahcpr.gov/qual/cahps/faqdata.htm.

5 Allen, H. and Rogers, W. (1997). “The Consumer Health Plan Value Survey: Round Two.” Health Affairs, 16(4), 156 – 166.

6 Hall, J.A., Milburn, M.A., Roter, D.L., and Daltroy, L. H. (1998). “Why are sicker patients less satisfied with their medical care? Tests of two explanatory models.” Health Psychology, 17(1), 70-75.

7 HEDIS is a registered trademark of the National Committee for Quality Assurance (NCQA).

8 The SF-12 is a standardized 12-item measure of health status that assesses concepts related to physical and psychological functioning such as role limitations due to physical or emotional problems, bodily pain, general health, vitality, social functioning, and psychological distress/well-being. Physical and mental health subscales are generated from the SF-12. (Ware, J.E., Kosinski, M., Keller, S.D. SF-12: How to Score the SF-12 Physical and Mental Health Summary Scales. Boston: The Health Institute, New England Medical Center, Second Edition, 1995).

9 Hall, J.A., Milburn, M.A., Roter, D.L., and Daltroy, L. H. (1998). “Why are sicker patients less satisfied with their medical care? Tests of two explanatory models.” Health Psychology, 17(1), 70-75; Agency for Healthcare Research and Quality Web site: www.ahcpr.gov/qual/cahps/faqdata.htm.