Milbank Quarterly Vol. 85 No. 2

Volume 85, Number 2, 2007

To print this article, click here for the pdf version
Abstracts of other articles in this issue

Subscriptions, sample copies, and back issues

Using Population Segmentation to Provide Better Health Care for All:
The “Bridges to Health” Model

JOANNE LYNN, BARRY M. STRAUBE, KAREN M. BELL, STEPHEN F. JENCKS, and ROBERT T. KAMBIC

Centers for Medicare and Medicaid Services, U.S. Department of Health and Human Services

The model discussed in this article divides the population into eight groups: people in good health, in maternal/infant situations, with an acute illness, with stable chronic conditions, with a serious but stable disability, with failing health near death, with advanced organ system failure, and with long-term frailty. Each group has its own definitions of optimal health and its own priorities among services. Interpreting these population-focused priorities in the context of the Institute of Medicine’s six goals for quality yields a framework that could shape planning for resources, care arrangements, and service delivery, thus ensuring that each person’s health needs can be met effectively and efficiently. Since this framework would guide each population segment across the institute’s “Quality Chasm,” it is called the “Bridges to Health” model.

Keywords: Health care reform, community health planning, health services needs and demand, person-focused health.
rossing the Quality Chasm (IOM 2001a) envisioned an approach to health that focuses on the individual person or patient and met six specific aims for care: it must be safe, effective, efficient, patient centered (i.e., meets the patient’s desires and preferences within the care delivery environment), timely, and equitable.
The same IOM report proposed that this person-focused system require three elements: an information-rich environment supported by health information technologies, the patient’s (or advocate’s) engagement in all aspects of care, and coordination among teams of caregivers. Shifting to this model from our fragmented, provider-focused health care system (in which patients must adjust to their providers’ time and practice patterns) will require changes in how we assess and monitor the quality of health care, pay for health and reimburse for health care, monitor health and health needs, define optimal health, and prioritize health needs. Tailoring services in an ad hoc way to match each citizen’s situation, however, would be difficult and costly.
A practical alternative, widely used in other industries, is to stratify the customer population into groups that are sufficiently homogeneous to enable arranging a set of commonly needed supports and services to meet their expected needs. Our current system essentially segments the population by the provider whose services the patients are using at the moment—for example, a nursing home population, a hospitalized population, a home health care population, or an office-based care population. The results are dehumanizing and produce discontinuous, wasteful, and unreliable care. In this article, we suggest stratifying the population based on health prospects and priorities, rather than on the health care provider of the moment. No one can expect a trauma center to meet the long-term behavior management needs of a dementia patient, just as no one would expect an SUV to meet the needs of a thrifty driver when gas is scarce, or an airport hotel to have a ski lift.
This article illustrates how the concept of segmenting patient populations can lead to more creative and effective strategies for safe, efficient, effective, timely, patient centered, and equitable health care and thus a better understanding of how to achieve better health for both the individual and all people. Since our model tries to bridge the quality chasm (IOM 2001a) for each population segment, we call it the “Bridges to Health” model. In a health care system designed around the predictable needs of various populations, clinicians find it easier to respond to individual patients’ needs and preferences. Although this approach may have many applications, this article describes its uses only for the federal initiatives for quality and health information technologies. This is the first publication of this approach, but comments from scores of our colleagues over several years have helped shape the ideas. We present this approach here in order to invite comment and correction and to enable others to use it and report on its merits.
The Populations and Matching Services
Table 1 proposes segmenting the entire population into eight groups and illustrates each group using a representative person. (Later tables deliberately vary the labels slightly to help the reader better understand the eight populations.) Three considerations shape this proposal:

The set of population segments must be limited if the health care system is to offer a sensible array of integrated services for each segment and to make those services available almost everywhere.

The set of population segments should include everyone; that is, at every point in his or her life, every person should fit into one of these categories.

The people in each population segment must have sufficiently similar health care needs, rhythms of needs, and priorities to make the segment useful for planning, but each segment must be different enough to justify separate consideration. Planners must be able to structure the supports, service arrays, and care delivery arrangements so that they will meet the needs of anyone in that segment reasonably well, even though they may be mismatched to other segments.

Table 2 lays out the health-related concerns, major components of health services and supports, and life goals typical of each segment of the population. The last column of table 2 matches each population segment with the goals of health care developed by the Foundation for Accountability (FACCT) (Lansky and Bethell 2000) and adopted by the Agency for Healthcare Research and Quality (AHRQ 2004; IOM2001b) for the congressionally mandated annual review of health care quality and by the Institute of Medicine in its “Priority Areas for National Action” report (IOM 2003).

Other researchers have used paradigm cases from a small array of population segments to guide reform. The “Esther Project” in Sweden, for example, uses paradigm cases to test whether disabled elderly persons with certain characteristic clinical profiles can count on good care (Institute for Healthcare Improvement 2006). The results of that inquiry guide the priorities for health care improvement.
At any one time, nearly every citizen’s situation best matches the characteristics of one particular segment; but over time, most citizens move from one segment to others. Most of us spend most of our lives as healthy people (population segment 1), with occasional forays into and out of maternal and infant care (population segment 2) and acute illness (population segment 3). Eventually, most people accumulate one or more chronic conditions that require ongoing upkeep and then enter population segment 4. A small number of people live a long time with serious disabilities that are not particularly progressive, so they require information, tools, supports, and services to enable them to live full lives with disabilities (population segment 5). In general, persons with established chronic conditions (population segment 4) and serious disabilities (population segment 5) will not return to being healthy (population segment 1) or to merely needing acute care (population segment 3) or maternal/infant care (population segment 2), although they may have pregnancy or acute illness superimposed on their long-term condition. In general, substantial long-term conditions endure throughout such episodes, so the health and care system arrangements for population segments 4 and 5 should include access to acute and maternal services when needed.
Eventually, almost everyone experiences one of the end-of-life courses. The transition from being chronically ill or disabled to the degree of disability envisioned in population segments 6, 7, and 8 often is gradual. The U.S. Department of Veterans Affairs’ health care system makes veterans eligible for its Home-Based Primary Care program when “they face so many challenges that they are just too sick to come to clinic” (personal communication, Thomas Edes, MD, chief of Home and Community-Based Care for the U.S. Department of Veterans Affairs, January 20, 2007). The Gold Standards Framework in Britain (Gold Standards Framework 2006) and various palliative care efforts in this country (Lynn 2004) use the criterion of “being sick enough that death in the next six months would not be surprising.” This transition may require arbitrary and replicable definitions when eligibility for costly services such as home care and hospice is determined. Return from the last phase of life trajectories (population segments 6, 7, and 8) to other population segments is so unlikely as to justify only an occasional exception to the routine. Transitions from a course with organ system failure (population segment 7) to frailty (population segment 8) are more common, as a person living with a dominant organ system failure ages and accumulates multiple comorbidities and the syndrome of frailty. The care arrangements for these population segments should plan for the more common transitions.
Characterizing Quality
Crossing the Quality Chasm (IOM 2001a) envisioned a system that is safe, effective, efficient, patient centered, timely, and equitable. These aims have become the common framework for assessing providers, although they also frame a more comprehensive person-focused approach to health maintenance and improvement. Matching the priorities of each population segment to the six IOM aims is the central characterization of our Bridges to Health model. Table 3 provides a working understanding of the definitions of quality health care, which includes ensuring that information and supports for patients are available for self-management, measuring progress toward health-related goals, and prioritizing areas needing improvement in each cell of the framework. For example, whereas improvements in safety (a column) could require the prevention of falls and pressure ulcers for the frail elderly (a row), ensuring correct-site surgery is more important to those with acute problems (another row).
As each population segment (a row in table 3) intersects with each IOM aim (a column in table 3), the resulting cell offers a way to define interventions that will lead to optimal health in each population for each characteristic. The definitions in the cells in this matrix are examples and are not comprehensive. Managers and policymakers who use this approach should fill the grid with examples pertinent to their population and programs, and academicians may consider the precise allocation of issues and programs across the grid. All will see, however, that the matrix provides a ready check on the inclusiveness and scope of a quality health system’s definition. A robust row or column is likely to reflect substantial attention; conversely, a weak cell is likely to stand out as needing attention. Accordingly, the Bridges to Health framework enables well-targeted efforts to eliminate the Quality Chasm.

Filling the Bridges to Health matrix with improvement activities related to the three key reforms of the Quality Chasm report helps prioritize the development of products and policies needed to serve all populations. Table 4 applies the Bridges to Health model to opportunities for health information technologies.

Table 5 provides a similar overview of some of the Centers for Medicare and Medicaid Services’ current and proposed initiatives to address quality. The Bridges to Health model shows the strength and breadth of the current strategies. For example, only a few existing interventions address timeliness and equity, and strategic planners can consider whether these areas offer important opportunities. The structure of table 5 also serves as a template for reporting the progress of widespread improvement activities.

Population Size and Costs

Table 6 estimates the number of people in each segment and the cost of their care. Because expenditures are made over a period of time, during which people may move from one segment to another, the spending estimates in table 6 are based on our best estimates of the costliest segment for each person for a substantial part of the year. These estimates include the cost of drugs and paid long-term care, but not unpaid caregiving, loss of income, or disability income. Since data are not generally organized around the proposed categories, evidence of population sizes and costs is drawn from triangulating the relevant data—that is, using data from two or more sources to produce a single estimate (see the notes to table 6). Such estimates are first approximations, both because independent errors affect each source and because the method requires combining sources that use different definitions and time periods.

In addition to generating more reliable estimates generally, the future development of the model will require estimating the frequency and importance of transitions among population segments, examining the nature of outliers, and sharpening definitions. Table 6 gives the likely relationships of population size and cost at this stage. Some segments contain most of the population, and others consume most of the money. Persons living with a serious disability and those passing through a period of frailty at the end of life are quite costly, despite being small in number at any particular time.
Payer Differences
The U.S. health care system has a variety of payers, each with different scope, coverage, and delivery models. Table 7 estimates the comparative distribution of population segments among commercial insurers, Medicaid, Medicare, and the Veterans Health Administration. Clearly, the priorities of Medicare, Medicaid, and the Veterans Health Administration should be somewhat different from one another and quite different from those of commercial insurers. An overwhelming proportion of the stable disabled and the three segments comprising the last phase of life are in the public systems. Medicaid plans now cover about 40 percent of all pregnancies and deliveries (Kaiser Family Foundation 2002; Martin et al. 2003) and have more responsibility in this arena than do other public agencies and nearly as much as all commercial health care plans together. Since public payers also bear substantial responsibility for the overall health of their beneficiaries, policymakers and program managers may find our framework helpful in establishing payment incentives and quality standards that address the populations for which they are responsible.

Health Needs of the Population Segments
Our model rests on pursuing the health of each population segment. Achieving this goal for some population segments, such as the healthy, might require mainly ancillary services to supplement current care arrangements, such as creating electronic health records, having information available and controlled by the patient through a personal health record, and providing reminder systems for patients and clinicians. Reliably improving the health of some population segments, however, probably requires quite fundamental changes in service delivery arrangements and the availability of important options. The last four populations are so disabled or sick, for example, that substantial reengineering to ensure continuity of clinicians and to involve patients or their advocates in planning their care across multiple settings could prove to be among the highest priorities.
Development Needs for the Bridges to Health Concept
We have discussed the Bridges to Health model with scores of clinicians, managers, and policymakers, who usually find it easy to grasp and quickly move on to productive discussions using the model. Although the population segmentation approach in the Bridges to Health model rests on well-established business principles, it is in its infancy in health care. With the insight from applying this model and with improved data, researchers and managers can test alternative ways of segmenting the populations and establishing effective, efficient, and coordinated care that supports both patients’ engagement in observed differences in priorities and optimal solutions for discernible populations.
Some existing subpopulations challenge our proposed segmentation. Medicare law and financing, for example, make end-stage renal disease (ESRD) patients a distinct group, with its own quality improvement program. These patients fit in the chronic condition segment (population segment 4) early on and may worsen to have serious chronic conditions with exacerbations (population segment 7) unless another condition dominates in shaping the last part of the patient’s life. ESRD patients traditionally have received dialysis mostly from specialized providers. Whether it is wise to continue that separation is a matter of current debate, since end-stage renal disease patients accumulate many more health problems and need the services of the nonrenal care delivery system, and since the distinct ESRD delivery system cannot also address the need for comprehensive and reliable services for patients with milder renal failure who do not yet qualify as “end stage.” As we suggest here, the Bridges to Health model does not continue to treat ESRD patients as a separate segment.
Another classification challenge arises from mental illness. Mild mood and adjustment disorders readily fit into the first four populations: healthy, maternal/infant health, acute illness, and chronic conditions. Care delivery systems that have been optimized for these populations will certainly need to include reliable services for the ordinary run of mental distress, since the people in these populations so often have these needs. Serious, persistent mental illness is disabling but only modestly life threatening and thus is part of our segment 5. The optimal services use many of the same resources as those for persons living with congenital disabilities or spinal cord or brain injury, such as home assistance, environmental modification, and transportation services. The mental health services for persons with conditions like schizophrenia or substance abuse have often been quite different from those for the rest of health care. Population segment 5, people with relatively stable but substantial disability, includes diverse etiologies, living situations, and resources. Just as the acute care delivery arrangements include substantial subdivisions (e.g., among traditional medical and surgical specialties), the disabled population might be served best by keeping all in a population segment that optimizes life opportunities, autonomy, and direct support in the community but that expects programs to often subdivide according to particularly widespread or distressing clinical situations. More experience and analysis will clarify how best to match service delivery to the variety of patients in the relatively stable but seriously disabled population, segment 4.
For purposes like quality measurement and eligibility for services, the population assignments need to last for an administratively feasible period. People enter the last three populations only in their last phase of life and then do not often move into another group. A person in the stable disabled population remains there until he or she enters an eventually fatal course and becomes part of one of the last three populations. Transitions among the first four population segments and between those and the last four will always be common and may require establishing administrative procedures for payment and quality measurement purposes.
Conclusion
Each person needs somewhat different services for optimal health. Clinicians, therefore, always need to customize their service. Working within systems that reflect the likely priorities and needs of large segments of the population leads to efficient and reliable health care and supports the improvement of health across the entire U.S. population. Our Bridges to Health concept enables a rational customization of health care around important and coherent segments of the population and thus is more useful than the usual focus on diagnoses or provider types. The examples and estimates in this article represent first approximations and “proof of concept.” Further research and debate can generate better estimates and a more complete concept. One particularly important perspective in our work is that of the affected patients. Even though providers and the public have some sense of what serves people well in certain circumstances, it is the voices of patients and family members who face those circumstances that should determine their health and clinical priorities.
In sum, the proposed Bridges to Health model enables a pragmatic transformation of the arrangements for care and services so that each citizen can count on maintaining optimal health throughout life.
References
Agency for Healthcare Research and Quality (AHRQ). 2003. Medical Expenditure Panel Survey, Household Component data (using 2003 data). Available at http://www.meps.ahrq.gov/mepsweb/survey_comp/household.jsp (accessed May 18, 2006).
Agency for Healthcare Research and Quality (AHRQ). 2004. 2004 National Healthcare Quality Report. Rockville, Md. Available at http://www.ahrq.gov/qual/nhqr04/nhqr2004.pdf (accessed August 8, 2006).
Alzheimer’s Association. 2006. Statistics about Alzheimer’s Disease. Available at http://www.alz.org. (accessed April 4, 2006).
American Academy of Pediatrics. 2005. Pediatric Preventive Care Cost, Estimated US Average, 2005, by Patient Age, According to Recommendations for Preventive Pediatric Health Care and Recommended Childhood and Adolescent Immunization Schedule, US, 2005. Available at http://www.aap.org/research/ImmandPrevCareCost.pdf (accessed May 18, 2006).
American Cancer Society. 2006. Cancer Facts and Figures 2006. Available at http://www.cancer.org/downloads/STT/CAFF2006PWSecured .pdf (accessed May 18, 2006).
Anderson, G., and J. Horvath. 2004. The Growing Burden of Chronic Disease in America. Public Health Reports 119(3):263–70.
Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds. 2006. 2006 Annual Report. Available at http://www.cms.hhs.gov/reportstrustfunds/downloads/ tr2006.pdf (accessed May 18, 2006).
Centers for Disease Control and Prevention (CDC). 2006. Chronic Disease Overview. Available at http://www.cdc.gov/nccdphp/overview.htm (accessed May 18, 2006).
Community Tracking Survey. 2006. Online Query System. Available at http://www.hschange.com/index.cgi?file=cts1 (accessed May 18, 2006).
Family Caregiver Alliance. 2006. Selected Long-Term Care Statistics. Available at http://www.caregiver.org/caregiver/jsp/content_node.jsp?nodeid=440 (accessed August 8, 2006).
Francis, W. 2006. The Price of Everything and the Value of Nothing. Testimony before the Joint Economic Committee, May 10. Available at http://www.jec.senate.gov/Documents/Hearings/francistestimony10may2006.pdf (accessed May 18, 2006).
Gold Standards Framework. 2006. Prognostic Indicator Guidance. Available at http://www.goldstandardsframework.nhs.uk (accessed December 31, 2006).
Guttmacher Institute. 2006. Facts on Induced Abortion in the United States, May 4, 2006. Available at http://www.guttmacher.org/pubs/fb_induced_abortion.html (accessed May 18, 2006).
Hogan, C., J. Lynn, J. Gabel, J. Lunney, A. O’Mara, and A. Wilkinson. 2000. Medicare Beneficiaries’ Costs and Use of Care in the Last Year of Life. Washington, D.C.: Medicare Payment Advisory Commission.
Institute for Healthcare Improvement. 2006. Improving Patient Flow: The Esther Project in Sweden. Available at http://www.ihi.org/IHI/Topics/Flow/PatientFlow/ImprovementStories/ImprovingPatientFlowTheEstherProjectinSweden.htm (accessed December 31, 2006).
Institute of Medicine (IOM). 2001a. Crossing the Quality Chasm. Washington, D.C.: National Academy Press.
Institute of Medicine (IOM). 2001b. Envisioning the National Health Care Quality Report. Washington, D.C.: National Academy Press.
Institute of Medicine (IOM). 2003. Priority Areas for National Action: Transforming Health Care Quality. Washington, D.C.: National Academy Press.
Kaiser Family Foundation. 2002. State Health Facts, Number of Births Financed by Medicaid (with 1,661,320 births in Medicaid in 2002). Available at http://www.statehealthfacts.org/ cgi-bin/healthfacts.cgi?action=compare&category=Medicaid+% 26+SCHIP&subcategory=Births+Financed+by+Medicaid&topic =Total+Medicaid+Births (accessed August 8, 2006).
Kaiser Family Foundation. 2005. Projected 7.4% Increase in 2005: >2 trillion. (For 2004, see www.kff.org, Trends and Indicators in the Changing Health Care Marketplace, citing Centers for Medicare and Medicaid Services, Office of the Actuary, National Health Statistics Group.) Available at http://www.cms.hhs.gov/ NationalHealthExpendData/ (see Historical; NHE summary including share of GDP, CY 1960–2004; file nhegdp04.zip) (accessed May 18, 2006).
Kirschstein, R. 2000. Table 1: Cost of Illness and NIH Support for Selected Diseases and Conditions. In Disease Specific Estimates of Direct and Indirect Costs of Illness and NIH Support: Fiscal Year 2000 Update. Available at http://ospp.od.nih.gov/ecostudies/COIreportweb.htm (accessed May 18, 2006).
Lansky, D., and C. Bethell. 2000. Empowering Consumers to Make Informed Choices. In Connecting with the New Healthcare Consumer, edited by D.B. Nash, M.P. Manfredi, B. Bozarth, and S. Howell. Boston: Jones and Bartlett.
Lynn, J. 2004. Sick to Death and Not Going to Take It Anymore! Reforming Health Care for the Last Years of Life. Berkeley, Calif.: University of California Press; New York: Milbank Memorial Fund. Also in chap 2, “No Surprise” Question, online at http://www.medicaring.org/sicktodeath/ index.html (accessed December 31, 2006).
Martin, J.A., B.E. Hamilton, P.D. Sutton, S.J. Ventura, F. Menacker, and M.L. Munson. 2003. Births: Final data for 2002. National Vital Statistics Reports 52, no. 10. Available at http://www.cdc.gov/nchs/data/nvsr/nvsr52/nvsr52_10.pdf (accessed August 8, 2006).
Mature Market Institute. 2004. 2004 MetLife Market Survey of Nursing Home and Home Care Costs, September. Available at http://www.metlife.com/WPSAssets/16651817681106065148V1FNursing% 20Home%20Home%20Care%20Costs.pdf (accessed February 26, 2007).
National Center for Health Statistics, Centers for Disease Control and Prevention. 2002. Table 12: Live Births by Hispanic Origin of Mother and by Race for Mothers of Non-Hispanic Origin. United States, Each State and Territory, 2002. National Vital Statistics Reports, 52, no. 10. Available at http://www.cdc.gov/nchs/data/nvsr/nvsr52/nvsr52_10.pdf (accessed May 18, 2006).
National Center for Health Statistics, Centers for Disease Control and Prevention. 2006. Table 116: Nursing Homes, Beds, Occupancy, and Residents, According to Geographic Division and State: United States, 1995–2003. In Health, United States, 2005. Available at http://www.cdc.gov/nchs/data/hus/hus05.pdf (accessed May 18, 2006).
U.S. Census Bureau. 2006. U.S. Population as of April 1, 2006. Available at http://www.census.gov/popest/national/tables/NA -EST2006-01.xls (accessed February 26, 2007).
U.S. Environmental Protection Agency (EPA). 2006. Cost of Illness Handbook. Available at http://www.epa.gov/oppt/coi/pubs/toc.html (accessed May 18, 2006).
Waldrop J., and S. Stern. 2000. Disability Status 2000, Table 1. Census 2000 brief no. C2KBR-17. Washington, D.C.: U.S. Census Bureau. Available at http://www.census.gov/prod/2003pubs/c2kbr-17.pdf (accessed February 26, 2007).

Acknowledgments: This article has been revised many times, guided by the generous input of scores of colleagues. We especially acknowledge the helpful critiques of leaders at the Centers for Medicare and Medicaid Services, the Institute for Healthcare Improvement, the Veterans Healthcare System, Kaiser Permanente, the Johns Hopkins Bloomberg School of Public Health, the National Institute of Mental Health, the Commonwealth Foundation, the United Hospital Fund of New York City, and RAND. We also acknowledge the assistance of Amanda Pomeroy at RAND in documenting the estimates in table 6. This article did not require research on human subjects, and we have no financial conflict of interest with regard to the subject matter.
Address correspondence to: Joanne Lynn, Office of Clinical Standards and Quality, CMS, 7500 Security Blvd., Baltimore, MD 21244-1850 (email: Joanne.lynn@cms.hhs.gov).

(c) 2007 Milbank Memorial Fund
The Milbank Memorial Fund is an endowed operating foundation that engages in nonpartisan analysis, study, research, and communication on significant issues in health policy. In the Fund's own publications, in reports or books it publishes with other organizations, and in articles it commissions for publication by other organizations, the Fund endeavors to maintain the highest standards for accuracy and fairness. Statements by individual authors, however, do not necessarily reflect opinions or factual determinations of the Fund.
This file may be redistributed electronically as long as it remains wholly intact, including this notice and copyright. This file must not be redistributed in hard-copy form.
Online producer: Stephanie Moe-Quiggle