hi08 comparing payor performace

Upload: centerback

Post on 01-Mar-2016

219 views

Category:

Documents


0 download

TRANSCRIPT

  • Comparing payor performance to enhance health outcomes

    A new McKinsey tool enables payors to identify where

    their performance is weak, what they can do to

    improve it, and which peer organizations they can learn from.

  • 4949

    In many countries, health outcomes vary

    markedly across regions, often in ways

    that do not correlate with health care spending.

    The use of evidence-based interventions

    also varies markedly. These variations in out-

    comes and care delivery are coming under

    increasing scrutiny from a range of stakeholders,

    including payors, government officials, and

    the general public.1 These stakeholders share the

    concern that some patients are being given

    substandard care. Many are also troubled about

    the economic impact of substandard care,

    since high-quality care, particularly for chronic

    diseases, often lowers future health costs.

    For payors, substandard care is particularly

    worrisome, because their core mission

    is to fund care delivery and thereby improve

    public health.

    But reducing variations in outcomes and care

    delivery is no simple task. Payors have little

    or no direct control over some of the factors that

    strongly influence health outcomesespecially

    social determinants, such as education level

    and employment status. Payors can influence the

    extent to which providers use evidence-based

    interventions, which have been proven to enhance

    health outcomes. But few payors take a rigor-

    ous approach to identifying which results they

    most need to improve, which interventions

    they should promote, or how they can find exam-

    ples of best practices to emulate.

    We have developed a proprietary analytic

    framework to help payors take on this challenge.

    This framework, McKinseys Payor Health

    Index, enables payors to determine which health

    outcomes to focus on first, as well as which

    interventions and other actions would have the

    greatest impact on those outcomes. In addi-

    tion, the index allows payors to identify other

    organizations in similar circumstances so that

    they can ascertain which ones have lessons

    to teach them.

    The Payor Health Index was originally devel-

    oped in England but has since been adapted for

    use in several other countries, including the

    Netherlands, Saudi Arabia, and the United States.

    More recently, McKinsey has been developing

    a similar tool to help health systems determine

    what their priorities should be.

    In this article, we will describe how and why

    the Payor Health Index was developed,

    what results it provides, and how payors can

    use the results to improve health outcomes.

    Variations in care delivery

    As the focus on health care quality has increased

    around the world, even highly developed

    countries have discovered that they experience

    wide regional variations in health outcomes.

    In France, for example, infant mortality

    ranges from a low of 1.4 per 1,000 live births in

    Corse to a high of 4.5 per 1,000 in Alsace.2

    Similar differences are seen in many other

    countries (Exhibit 1).

    Regional variations have also been noted in

    other health outcomes, including cancer survivor

    rates, cardiovascular death rates, and years

    of life lost to chronic conditions. In England, for

    example, we identified an almost ninefold

    difference among the regions in the number of

    life years lost to diabetes (Exhibit 2). The dif-

    ference cannot be fully explained by how much

    the regions spend on health care (Exhibit 3).

    What does help explain the variations in out-

    comes are differences in underlying risk

    factors and the use of evidence-based inter-

    ventions. For example, both teenage pregnancy

    and smoking during pregnancy are known

    Paul Betts; Farhad

    Riahi, MD; and

    Siddharth Shahani

    1 For another look at how increased transparency into health outcomes and care delivery is affecting health systems, see How hospitals can respond to increased quality transparency, p. 58.

    2 National Institute for Statistics and Economic Studies, France, 2007.

  • 50 Health International 2009 Number 8

    risk factors for neonatal mortality. In Australia,

    the teenage pregnancy rate is six times higher

    in some regions than in others.3 In parts

    of England, only 4 percent of pregnant women

    smoke; in other areas, one-third of pregnant

    women do.4 Likewise, early detection with

    mammography can improve survival for women

    with breast cancer. In Italy, the percentage of

    eligible women who regularly undergo mammog-

    raphy is twice as high in some regions than in

    others.5 Similar variations in risk-factor

    prevalence and evidence-based intervention use

    can be found in countries around the globe.

    How the Health Index was built

    We began work on the Payor Health Index after

    noticing the differences in health outcomes

    in the various regions of England. We wanted

    to find a way to analyze and compare the

    performance of Englands primary care trusts

    (PCTs)the payors responsible for health

    care in the regionsso that we could help them

    improve their results.

    We focused on Englands payors because

    they have been tasked explicitly with

    improving health outcomes in their covered

    3 Australian Institute of Health and Welfare, 2007.

    4 Healthcare Commission: Acute Trusts 2007/08 National StandardsSmoking During Pregnancy.

    5 D. Giorgi et al., Mammography screening in Italy: 2005 data and 2006 preliminary data, Epidemiologia e Prevenzione, 2008, Volume 32, Number 2, Supplement 1, pp. 722.

    Exhibit 1

    Variations in infant mortality

    Within France, Sweden, and Italy, infant mortality rates vary by region.

    Infant mortality under age 1 (deaths per 1,000 live births by region)

    Average = 3.6

    Health International 2009PHIExhibit 1 of 6Glance: Within France, Sweden and Italy, levels of infant mortality vary by region. Exhibit title: Variations in infant mortality

    Alsace

    France

    Lorraine

    Champagne-Ardenne

    Haute-Normandie

    Ile-de-France

    Nord-Pas-de-Calais

    Source: National Institute for Statistics and Economic Studies, France, 2007; Central Bureau of Statistics, Sweden, 2007; Health for All Database, Italy, 2005

    4.5

    Poitou-Charentes 2.9

    4.5

    4.4

    Aquitaine

    Picardie

    Franche-Comt

    3.8

    3.8

    3.6

    4.2

    3.9

    3.8

    Languedoc-Roussillon

    Pays de la Loire

    Provence-Alpes-Cte dAzur

    Midi-Pyrnes

    Centre

    Rhne-Alpes

    3.5

    3.5

    3.4

    Basse-Normandie

    Bourgogne

    Auvergne

    3.2

    3.0

    3.0

    Corse 1.4

    Bretagne

    Limousin

    2.6

    2.5

    3.4

    3.4

    3.3

    Gotland

    Sweden

    Kalmar

    Norrbotten

    Vsternorrland

    Blekinge

    Dalarna

    4.8

    Kronoberg 2.3

    4.4

    3.7

    Jnkping

    Srmland

    Vstmanland

    3.3

    3.2

    3.2

    3.7

    3.3

    3.3

    rebro

    Halland

    Gvleborg

    Jmtland

    Vstra Gtaland

    stergtland

    2.9

    2.8

    2.8

    Skne

    Vrmland

    Stockholm

    2.6

    2.6

    2.5

    Uppsala

    Vsterbotten

    2.3

    2.2

    2.8

    2.7

    2.7

    Calabria

    Italy

    Sicilia

    Basilicata

    Puglia

    Campania

    Lazio

    5.4

    Molise 2.0

    5.1

    4.7

    Valle dAosta

    Friuli VG

    Emilia Romagna

    3.9

    3.7

    3.5

    4.6

    4.3

    4.2

    Abruzzo

    Lombardia

    Trentino AA

    Marche

    Piemonte

    Veneto

    3.4

    3.3

    3.1

    Liguria

    Toscana

    Sardegna

    2.6

    2.6

    2.6

    3.1

    2.9

    2.8

    Average = 2.8 Average = 3.7

    Above country average

  • 51

    Exhibit 2

    No standard outcomes

    In England, as in many other countries, differences in health outcomes are not easily explained.

    Primary-care-trust (PCT) outcomesOutcome metricDisease area

    Average Distribution across 152 PCTs

    Health International 2009PHIExhibit 2 of 6Glance: In England, as in many other countries, differences in health outcomes arenot easily explained. Exhibit title: No standard outcomes

    1 International Classification of Diseases (ICD) 10 codes I61-I64: intracerebral hemorrhage, other nontraumatic intracranial hemorrhage, cerebral infarction, stroke not specied as hemorrhagic or infarction.

    Source: Payor Health Index

    Coronary heart disease

    Standardized mortality ratio

    147Highest11980th percentile8920th percentile58Lowest

    104

    Cervical cancer

    5-year relative survival %

    74Highest6880th percentile

    6020th percentile56Lowest

    64

    Infant mortality

    Deaths in first year per 1,000 live births

    10.8Highest

    6.180th percentile4.020th percentile2.3Lowest

    5.1

    Diabetes Years of life lost per 10,000 people

    14Highest

    6.280th percentile3.120th percentile1.6Lowest

    4.8

    Stroke1 % deaths within 30 days of admission

    42Highest

    3080th percentile2320th percentile

    8Lowest

    26

    Comparing payor performance to enhance health outcomes

    populations. Furthermore, theylike payors in

    other publicly funded health systemsmust

    bear the long-term costs of health care provision.

    For such payors, it can often be less expensive

    to provide high-quality care than to address

    the complications that eventually arise when such

    care is not offered.

    In developing the Payor Health Index, our first

    step was to answer two questions: what were

    the most important disease areas to study, and

    what publicly available data sources could

    we use to investigate those disease areas? We

    defined important disease areas as those that

    are highly prevalent and have a strong impact on

    overall public health. (Colds, for example,

    are very common but have little impact. AIDS

    significantly affects patients but has a fairly

    low prevalence in most economically advanced

    countries.) We used publicly available data

    because we wanted to study all of Englands

    payors using the same source information.

    In total, we used almost a dozen different demo-

    graphic and clinical databases.

  • 52 Health International 2009 Number 8

    The result was a list of 11 disease areas:

    cardiovascular disease, diabetes, cancer, asthma,

    stroke, chronic obstructive pulmonary disease,

    pediatric health, sexual health, geriatric

    health, mental health, and alcohol/drug abuse.

    For each of these areas, we then defined the

    most important health outcomes to study, using

    published clinical studies as a guide. Some

    of these outcomes reflected patients current

    health status (for example, hypertension

    and asthma prevalence). Others indicated how

    often preventive services were being deliv-

    ered (the percentages of women whose breast

    cancers were detected at an early stage, for

    instance, or of elderly patients who were given

    pneumococcal vaccination). In still other

    cases, the outcomes reflected the relative success

    of treatment (for example, years of life lost,

    survival without permanent disability, percentage

    of low-birth-weight deliveries).

    Our next step was to identify the factors that

    could explain the regional variations in

    these outcomes, such as the use of evidence-based

    interventions, the available resources (for

    example, the number of physicians), and under-

    lying population risk factors (such as smoking

    rates and eating habits). Because we were

    using publicly available data, we could not always

    analyze the factors likely to have the strongest

    impact on outcome variations (see sidebar,

    Understanding current data limitations, p. 55).

    Nevertheless, we were able to quantify a dozen or

    more metrics for each disease area. We then

    used this information to create, for each PCT, an

    individual index for each disease area.

    What the Health Index tells a payor

    The index for each disease area is presented in

    a simple visual format designed to convey

    a great deal of information quickly (Exhibit 4).

    Exhibit 3

    Health versus spending

    The number of deaths from diabetes does not correlate with per-patient spending.

    Relationship between primary-care-trust (PCT) spending on diabetes and deaths from diabetes in each PCT

    6

    5

    Deat

    hs fr

    om d

    iabe

    tes

    mel

    litus

    as

    a %

    of

    all d

    iabe

    tic p

    atie

    nts

    in a

    PCT

    r2 = 0.0024

    Health International 2009PHIExhibit 3 of 6Glance: The number of deaths from diabetes does not correlate with spending.Exhibit title: Health versus spending

    r2 is the proportion of variance explained by a regression.

    Source: Payor Health Index

    4

    3

    200

    2

    1

    00 400 600 800 1,000 1,200

    PCT spending on diabetes per diabetic patient,

  • 53Comparing payor performance to enhance health outcomes

    Exhibit 4

    Cancer health index

    Metrics on cancer health can be grouped into four major categories.

    Wait for treatment

    Highest performer

    Average, all PCTs

    PCT1 figureUnitsMetricsScores, 15A

    C

    BD

    73.5102102.7SMR3 1. Cancer mortality2A. Outcomes 2.62

    5-year survival

    Detected early

    8.06.36.4 2. Lung

    55.047.846.6 3. Colon

    82.378.676.8% 4. Breast

    73.764.363.8 5. Cervical

    75.667.158.0 6. Prostate

    74.762.169.0% 7. Breast

    88.680.081.0% 8. Cervical

    Screening coverage

    B. Interventions 4.59

    81.169.676.1% 9. Cervical

    84.173.382.3%10. Breast

    C. Resources 1.98

    100.099.098.811. 31-day

    100.099.799.612. 2-week

    100.091.688.0%13. 62-day

    1.219.935.614. MRI long waiters2

    9.16.46.515. Proportion spent on cancer

    5.31.00.216. Medical oncologists

    D. Risk factors 2.79

    % 17.626.627.117. Estimated smoking prevalence

    Health International 2009PHIExhibit 4 of 6Glance: Metrics on cancer health can be grouped into four major categories. Exhibit title: Cancer health index

    1 Primary care trust.2Low gure = good performance.3Standard mortality ratio.4Full-time equivalent.

    Source: Payor Health Index

    FTE4 per 100,000 population

    First, it tells a PCT how well it is doing on each

    metric studied. In addition, it provides national

    averages for each metric, as well as the results

    achieved by the highest-performing payor on each

    metric. This format enables a PCT to ascertain

    how its cancer mortality rate, for example, com-

    pares with the national average, as well as

    by how much it could lower that rate if it were

    to achieve results comparable to those of

    the highest-performing PCT.

    The metrics are grouped into four major

    categories: outcomes, interventions, resources,

    and risk factors. In addition, the individual

    scores for each metric in a category are combined

    into a global score for that category, which

    enables a PCT to compare its performance

    at a glance with that of other payors. The

    global scores range from 1 (low performance)

    to 5 (high performance).

  • 54 Health International 2009 Number 8

    For example, a PCT that has comparatively low

    mortality from cardiovascular disease and

    coronary heart disease, as well as a low prevalence

    of diabetes, hypertension, and obesity, would

    be given a high global score for cardiovascular

    outcomes. To receive a high global score for

    cardiovascular interventions, the PCT would have

    to show, among other things, that it is doing

    a good job controlling blood pressure and choles-

    terol levels in its population. High scores for

    risk factors and resources would indicate that the

    PCT has minimized the risk factors in its

    population (by reducing smoking rates, for

    example) and has sufficient resources to provide

    high-quality care.

    Finally, the global scores are graphed into

    a diamond. A PCT performing ideally

    would have a diamond that is almost entirely

    blue. In our experience, however, no payors

    perform ideally, and thus some white

    always appears. The proportion of the diamond

    that is white, and the areas where the white

    appears, indicate how much of a problem a payor

    may have. For example, if the only large area

    of white appears at the bottom of the diamond,

    the payor may not have a problem at all; the

    white simply indicates that its resource allocation

    in that disease area is small. This could be

    appropriate if the prevalence of risk factors in

    its population is low, if it is using evidence-

    based interventions appropriately, and if it is

    achieving good results (the payor could devote its

    resources to disease areas with greater needs).

    Conversely, a payor with a lot of white only at the

    top of the diamond has a major problem,

    because it is achieving poor outcomes while

    expending significant resources.

    Exhibit 5

    Mixed performance

    Payor performance often varies by disease area.

    PCT B PCT CPCT1 A

    Outcome score (1 = poor, 5 = good)

    Cardiovascular

    Cancer

    Mental health

    Sexual health

    Drugs/alcohol

    Child health

    Diabetes

    Stroke

    Elderly health

    Asthma

    COPD2

    Health International 2009PHIExhibit 5 of 6Glance: Payor performance typically varies by disease area. Exhibit title: Mixed performance

    1Primary care trust.2Chronic obstructive pulmonary disease.

    Source: Payor Health Index

    1 2 3 4 5 51 2 3 4 1 2 3 4 5

  • 55Comparing payor performance to enhance health outcomes

    We have found that the performance of most

    payors is mixed; they tend to do well

    in some disease areas and poorly in others

    (Exhibit 5). Very few payors score well

    across the board. This is an important point to

    remember, given the paucity of outcomes

    data available in most countries. The fact that

    a payor achieves good results in one area

    (for example, cardiovascular disease manage-

    ment) does not guarantee that it will

    achieve similarly good results in other areas.

    How payors can use the results

    The wealth of data the Payor Health Index

    provides is often invaluable to payors. The index

    enables them to determine which disease areas

    and which outcomes to focus on, as well as

    which underlying factors they should attempt to

    change first. In addition, it helps payors identify

    peers that are achieving better outcomes,

    which then allows them to learn from their peers

    and improve their own performance.

    Pinpointing what to focus on

    The experience of three PCTs, all of which had

    scored poorly on breast cancer outcomes,

    illustrates the kind of help the Payor Health Index

    can provide. The first PCT discovered that its

    screening rate (the percentage of women

    who regularly receive mammograms) was only

    half the national average; it was therefore not

    surprising that the PCTs early detection and

    five-year survival rates were low. For this payor,

    the challenge was to determine how it could

    better reach out to the women in its community

    to persuade them to undergo mammography.

    Understanding current data limitations

    At present, both the health data gathered directly by payors

    and the information that is publicly available have two

    important limitations that must be kept in mind when payor

    performance is being compared: availability and reliability.

    These limitations do not prevent us from comparing

    performance, but the comparisons would be stronger if the

    available data were more robust.

    Availability. Throughout the world, efforts to monitor the quality of care are in their infancy, and thus the data

    that providers are required to report to payors are incomplete

    and somewhat arbitrary. In some cases, the metrics

    were selected simply because they are easy to measure,

    rather than because they are the best predictors

    of outcome.

    An even greater problem with data availability is that many

    types of information are not being systematically collected.

    Breast cancer provides a good illustration of this problem.

    Currently, few, if any, payors or government agencies are

    collecting basic demographic information from women about

    age at first period, total number of pregnancies, or age at

    menopause. They are also failing to collect more sophisticated

    information, such as family history of breast cancer

    or the presence of genetic markers. In most communities,

    information about smoking rates in women is available,

    but smoking is a much, much weaker risk factor for breast

    cancer than family history or genetic markers.

    Reliability. Publicly available information is not always as reliable as we would like. For example, it may have

    been derived from relatively small surveys conducted over

    a brief period of time, and errors may have crept in as

    the numbers were extrapolated upward. Alternatively, the data

    may have come from a large-scale (for example, national)

    survey, but the local population may not be a representative

    sample of the national population.

    Because of these limitations, the risk factors and interventions

    included in the Payor Health Index are not always the ones

    that would have been best to use; some were simply the best

    available. We strongly believe that payors should demand to

    be given better information. This may be easier to achieve with

    providers than with local governments, but it is something

    payors should strive for if they want to improve health

    outcomes.

  • 56 Health International 2009 Number 8

    The second PCT had a higher-than-average

    screening rate but a lower-than-average

    early detection rate. The cause of this anomaly

    was not immediately evident. It is unlikely

    that the wrong women were being screened,

    because the guidelines for patient selection

    are very clear. The PCT is currently investigating

    whether its providers equipment may be

    malfunctioning and whether radiologists may be

    reading the mammograms inaccurately.

    The third PCT had a different problem: both

    its screening rate and its early detection

    rate were markedly above the national average,

    but its five-year breast cancer survival rate

    was below average. When the payors chief exec-

    utive looked into the problem, she discovered

    that it most likely resulted from the regions

    fragmented network of oncology providers. Best-

    practice payors use a high-volume provider

    typically a regional cancer centerthat offers

    high-quality services, including breast surgery,

    radiotherapy, and chemotherapy.

    We have seen similar results in other disease

    areas. Occasionally, the Payor Health

    Index highlights underlying factors that a payor

    has little or no control over (the number of

    single-parent families in a community, for exam-

    ple). Even in these cases, however, it provides

    helpful information because it enables the payor

    to make better decisions about how to combat

    certain conditions. A payor that wanted to

    improve its asthma outcomes, for instance, could

    realize that it needed to enlist the help of other

    organizations, such as housing authorities

    and environmental agencies, in order to reduce

    some of the risk factors for that disease.

    Identifying peers to learn from

    Many payors claim that their circumstances are

    too unusual to permit easy comparisons:

    their population is too rural (or too urban), its

    average age is too young (or too old), and so

    forth. In actuality, this is rarely the case.

    There are many meaningful ways to categorize

    payors into peer groups.

    For example, we segmented all of Englands

    PCTs into nine sets, based on each communitys

    deprivation level and its risk factors for diabetes.

    We then looked closely at the three sets

    of PCTs that had a high number of diabetes

    risk factors to see how their performance

    compared (Exhibit 6). As expected, performance

    within each set varied widely. However, the

    relative level of deprivation (low, moderate, or

    high) had very little impact on performance; in

    fact, the PCT with the highest performance was in

    a highly deprived community. These findings

    support our contention that payors can learn

    from their peers even their peers in more

    deprived communities.

    Using the Health Index in other settings

    Since we first developed the Payor Health Index,

    we have adapted it for use in other countries.

    In each case, the purpose and approach remained

    the same, but we tailored the index either to

    better reflect locally available data or to provide

    greater focus on, and a more extensive set

    of metrics for, disease areas of particular concern

    (for instance, cancer and diabetes).

    More recently, we have begun to develop a similar

    tool, which we have dubbed Health Insights,

    for use in health systems. The aim of the Health

    Insights is to provide health system executives

    with the intelligence required to make the most

    effective use of their health care budgets. Health

    Insights is designed to permit comparisons

    both within and across health systems; it allows

    executives to measure how well the regions

    within their own systems are doing, as well as

  • 57Comparing payor performance to enhance health outcomes

    how well their systems compare with the health

    systems in other countries. This tool assesses

    the resources spent, outcomes achieved, and key

    drivers of performance, and it identifies the

    areas most in need of improvement. Our hope is

    that the executives who opt to use Health

    Insights will develop into a peer community

    of regional health leaders who share experiences

    with and learn from one another.

    Around the world, increasing attention is being

    paid to both the cost of health care and the

    quality of care delivered. As a result, it is doubtful

    that many countries will continue to tolerate

    the wide regional variations in health outcomes

    that are seen today. The Payor Health Index

    and its offshoot, Health Insights, can help

    payors and health systems identify where their

    performance is weak, what they can do

    to improve it, and which peer organizations they

    can learn from.

    Paul Betts, a senior research analyst with McKinseys

    health care practice in London, works extensively with

    Englands National Health System. Farhad Riahi, MD,

    a principal in the London office, leads McKinseys work

    on clinical health economics. Siddharth Shahani, a

    research analyst at McKinseys Health Systems Institute,

    currently focuses on making health information compa-

    rable across different health systems.

    Exhibit 6

    Performance varies

    Even within peer groups, payor performance varies.

    Diab

    etes

    mor

    talit

    y, y

    ears

    of

    life

    lost

    per

    10,

    000

    peop

    le

    16

    14

    12

    10

    8

    6

    4

    2

    Health International 2009PHIExhibit 6 of 6Glance: Exhibit title: Payor performance varies even within peer groups

    Source: Payor Health Index

    High risk factorsHigh deprivation

    High risk factorsSome deprivation

    High risk factorsLow deprivation

    Each dot represents a different payor