Anda di halaman 1dari 9

International Epidemiological Association 2001

Printed in Great Britain

International Journal of Epidemiology 2001;30:12601263


The role of medical care in contributing to

health improvements within societies
John P Bunker

In attempting to assess the contributions of medical care to

health improvements, the goals of care must first be addressed.
The saving of lives in acute life-threatening emergencies is an
important such goal, but it represents a very small component
of the total medical effort: while lives are saved, the sum of
such saving is too small to have a measurable impact on the
life expectancy of an entire population. A much larger effort is
devoted to preventive or curative measures, and these do have
a large and measurable effect on the life expectancy of the
population as a whole. An even greater component of medical
effort is devoted to improving the quality of life, or more accurately, to preventing or to minimizing the poor quality of life
associated with chronic disease: to the relief of pain, disfigurement, and disability.
There are no population-based data to allow a direct estimate
of the contributions of medical care to life extension or to the
quality of life. In the absence of such data, my colleagues and
I have created inventories of the outcome benefits of the
preventive or curative care for individual conditions. Based on
such an inventory of established life extending outcomes of
preventive and curative services for individual conditions, I
estimate that about half of the 71 2 years of increased life
expectancy since 1950 can be attributed to medical care. I credit
an even larger number of years of relief, or partial relief, of poor
quality of life to medical care. The data on which the estimates
are based are often incomplete, and the estimates are
approximations. They are more than speculative and less than

Thomas McKeowns Legacy

With a single major exception, the epidemiology of medical care
and its effect on health have received little attention over the
years. The exception is McKeowns Role of Medicine, based on
cause-specific mortality reports for the century ending in 1971.1
Life expectancy had increased by 23 years during the first half
century, but McKeown was able to attribute no more than a
year or two to advances in medical care. He presented no data
on the harm that medical care might incur, but his conclusion
that medical care had contributed little to health was interpreted by many as an attack on medicine, and it was linked
by many to Ivan Illichs claim that medicine does more harm
than good.2 Illichs Medical Nemesis: The Appropriation of Health,
Department of Epidemiology and Public Health, University College London
Medical School, 119 Torrington Place, London WC1E 6BT, UK. E-mail:

published in 1975, and McKeowns Role of Medicine, published

the following year raised questions that have remained largely
unanswered to this day.
The implications for public health of McKeowns and Illichs
books have been largely ignored or considered irrelevant by
clinicians, who are busy taking care of patients one at a time.
Basic scientists appear not even to have noticed their existence.
The Nobel Laureate and President of Rockefeller University,
Joshua Lederberg, wrote that by the 1960s we could celebrate
the conquest of polio and the transformation of formerly
lethal infections to easy targets for penicillin and other miracle
drugs greater life expectancyfrom 47 years in 1900 to 70
in 1960can be attributed almost entirely to this mastery of
infection .3 The Nobel Laureate and former research director
of Burroughs Wellcome, the pharmaceutical company, George
Hitchings, claimed that the increase in life expectancy over the
last 50 years has been attributed to new medicines.4 Philip
Abelson, a former editor of Science, wrote that pharmaceuticals
have been responsible for about half of the improvement in
health care during this century.5
The quarter century following publication of The Role of
Medicine has seen an explosion of new treatments, many of which
have been shown in clinical trials to result in marked improvements in health. Epidemiological evidence crediting medical
care with the extension of life began to appear. The American
economist Jack Hadley compared expenditures by the governments Medicare programme with regional death rates and
reported that for every 10% increase in expenditure there had
been a 12% fall in mortality.6 It was reported that while death
rates have continued to fall for all ages and almost all diseases,
death rates for diseases considered amenable to treatment
have fallen at much greater rates than have death rates for
diseases that do not respond favourably to treatment.7 From the
differences in their rates of fall it was estimated that medical
care could be credited with an extension in life expectancy of
between 5 and 18 years.8
Conflicting data also appeared. Age-adjusted death rates were
reported to be greater in countries with greater numbers of doctors, and presumably with more medical care.9 Equally difficult
to explain, death rates for diseases amenable to treatment were
reported to be greatest in areas with the most medical care
resources.10 Efforts to separate the effects of medical care for an
entire population from those of other determinants of health
have been fraught with similar contradictory evidence. It was in
an attempt to overcome these difficulties that my associates and
I created an inventory of the outcomes of individual medical
and surgical interventions, one at a time.11,12



Life Expectancy
Estimation of months or years of increased life expectancy
attributable to the treatment of a particular condition involved
a three-step procedure: calculation of increases in life expectancy resulting from a decline in diagnosis-specific death rates,
estimation of increases in life expectancy when therapy is provided under optimal conditions, and estimation of how much of
the decline in disease-specific death rates could be attributed to
medical care when provided in routine practice.
To estimate gains in life expectancy attributable to a specific
disease, several approaches were used. Changes in life expectancy could be calculated directly from life-tables for those conditions for which mortality rates are available decade by decade.
This was possible for heart disease, cerebrovascular disease
and pneumonia using data published by the National Center
for Health Statistics in Maryland.13 For many other diseases
changes in mortality are reported by the National Center as a
single age-adjusted rate, precluding a full life-table estimate. For
example, the age-adjusted death rate for pneumonia and influenza fell from 26.2 per 100 000 in 1950 to 13.7 in 1989, a
reduction of 12.5 per 100 000. During the same period the ageadjusted death rate for all causes fell from 840.5 per 100 000 to
523.0 per 100 000, a difference of 317.5, and life expectancy
rose by 7.1 years. As a first approximation, the fall in death rate
for pneumonia and influenza was estimated to have contributed
12.5/317.5 7.1 = 0.28 years, or about 3 months. Such a back
of the envelope approximation when applied to age-adjusted
death rates for heart and cerebrovascular disease, as well as for
pneumonia and influenza, gave answers within 10% or 20% of
those based on a full standard life-table.
Estimation of increase in life expectancy under therapeutic
conditions was based, whenever possible, on clinical trials and
meta-analyses, or alternatively on observational or case-control
studies. The proportion of improvement attributable to medical
treatment was estimated from published treatment rates and
rates of risks and benefits of those receiving treatment. For
example, the death rate for cerebrovascular diseases, primarily
stroke, fell more than three-fold from 1950 to 1996, representing approximately 130 000 fewer deaths annually and
an increase in life expectancy of a little more than one year.
Medical control of hypertension in the US increased markedly during this period, from 10% of individuals with moderate or severe elevations of blood pressure to approximately
50%,14 and stroke mortality was reported to fall by 3540%
in randomized clinical trials of anti-hypertensive drugs.15
This would explain as much as 1520% of the reduction in
stroke mortality, with an increase in life expectancy of 21 23
Mortality from heart disease in the US fell by more than half
between 1950 and 1995, with a resultant increase in life expectancy of approximately 31 2 years, half to two-thirds of which
has been attributed to coronary care units, treatment of hypertension, and medical and surgical treatment of coronary artery
disease.16,17 The treatment of appendicitis, diabetes, and endstage kidney disease were estimated each to have contributed a
third of a year or more to life expectancy, with lesser contributions from the treatment of many other conditions. All told,
clinical services, composed of preventive services as well as
therapeutic intervention, we credited with 5 or 51 2 years of the


30 years increase in life expectancy since 1900, and half of the

7 or 71 2 years of increase since 1950.12

Relief of Pain, and of Physical and Mental

In the absence of population-based measures of the impact of
medical care on the quality of life, an inventory of estimated
years of pain and of physical or mental dysfunction prevented
or ameliorated as a result of treatment was similarly constructed.12
Estimates were based on the incidence of the condition, the
number of those treated, the average age of individuals suffering the condition, and the expected average years of survival.
The treatment of cataracts serves as an example of the application of this heuristic. Five to ten per cent of Americans develop
cataracts during a lifetime. There are over a million operations
to remove the affected lens or to replace it with a prosthesis in
the US each year. For 7590% of individuals there is restoration
or marked improvement in visual acuity.18 Assuming that the
life expectancy at the time of surgery is, on average, 5 years, the
operation can be credited with 2040 years of improved quality
of life for each 100 members of the population, or 21 25 months
on average for individuals.
In addition to large numbers of years of symptomatic relief,
there are many years of disability that are prevented by therapeutic intervention. For example, the effective treatment or
prevention of hypertension contributes to a fall in the subsequent incidence of non-fatal as well as fatal heart attacks and
strokes, and to a resultant fall in months and years of poorer
physical and mental function. Similarly, if it is possible to
prevent dyspepsia of peptic ulcer, or to prevent recurrent biliary
colic by removing the gall bladder, the months and years of
relief or amelioration can be estimated. The months and years
can then be added to create an inventory roughly comparable
to the so-called DALY, the number of disability-adjusted life
years. The years of improved physical or mental function, or
of prevention or amelioration of pain and suffering, add up to
an estimated 500 per 100 individuals, that is, about 5 years on
average per individual.12

Iatrogenic Disease, Mortality and Morbidity

Treatment can shorten as well as lengthen life expectancy. This
is dramatically evident in surgery and anaesthesia which,
however well performed, entail some risk of death and has been
used to explain, at least in part, differences in life expectancy
between countries. The two-fold greater rates of discretionary
surgery in America, relative to England and Wales, that I reported
in 197019 could thus be expected to reflect this risk and was
estimated to account for a third to a half of the greater life
expectancy in England and Wales at that time.20 It was assumed
that the rates of emergency and lifesaving or life-extending
surgery were essentially the same in each country and that the
discrepancy in surgical rates applied primarily to discretionary
surgery for non-life-threatening conditions.
Iatrogenic mortality may similarly help to explain the
observations by Cochrane and by Carr-Hill and their associates
that greater numbers of doctors and medical resources, and
presumably more discretionary medical and surgical care, are



associated with higher death rates.9,10 Iatrogenic mortality is

also reflected in the observation of brief but dramatic decreases
in population death rate when doctors strike and surgery for
elective (but not emergency) operations are suspended.21
That the magnitude of iatrogenic disease is substantial, in
medicine as a whole as well as in surgery, has now been
documented in the US. A year 2000 report by the Institute of
Medicine in Washington estimates that adverse events
occurring in the course of treatment, one-third to two-thirds
of which could be attributed to medical error, are responsible
for between 44 000 and 98 000 deaths annually in America.
Approximately half of adverse events, non-fatal as well as fatal,
occurred in surgical patients and half were attributed to
medication.22 This number of deaths would represent between
2% and 4% of all deaths in America, equivalent to a loss of
life expectancy of between 4 months and 8 months. Later in
the same year it was claimed that there are 225 000 or more
iatrogenic deaths in the US annually.23 This estimate was
based on data which may be methodologically flawed.2426 The
estimate also fails to take into account evidence that the
incidence of iatrogenic death may have fallen during the years
subsequent to those on which the Institute of Medicine report
is based.27
In the absence of current and reliable data, I estimate that
there may be between 75 000 and 150 000 iatrogenic deaths in
the US annually, including those that result from medical error.
A substantial number of these deaths, perhaps between a third
and a half, will have been imbedded in the observational data
on which the analyses reported in the foregoing were based.11,12
Adjusting for this confounding and for the shortened remaining
years of life of the elderly patients in whom the preponerance
of adverse events occur, I estimate that there is a loss of life
expectancy of 6 months or possibly as many as 12 months as a
result of iatrogenic deaths occurring during medical and surgical

Gains from Medical Care in Context

The gains in life expectancy and quality life that I credit to
medical care should be seen in the context of other determinants
of health. Compared to the very large gains in life expectancy in
the first half century that resulted from improvements in public
health, the contribution of medical care is relatively small. With
improvements in public health largely complete, medical care
is now the major determinant of life expectancy, its impact
substantially greater than that of the social environment or
In England and Wales, during the years 19871991, there was
an estimated 91 2-year difference in life expectancy between
men in professional social class I and unskilled labourers in
social class V; for women the differential was 61 2 years. These
were the extremes and represented a relatively small proportion
of the population, approximately 7% professionals and 4%
unskilled labourers. For the very much larger numbers of men
classified as as non-manual and skilled or partly skilled workers
the difference in life expectancy was very much smaller: less
than one year difference between skilled manual (class IIIM)
and skilled non-manual (class IIIN), and a little more than
11 2 years difference between social class II, managerial and
technical/intermediate, and social class IV, partly skilled,

e.g. warehousemen, security guards, machine tool operators.29

After adjusting for the size of each of the social or occupational
groups, the net effect of social class on life expectancy of the
whole population is 3 years, of which about a third can be
charged against the use of tobacco, and possibly as much as a
third against poorer access to and poorer quality of medical
The impact of an unhealthy lifestyle is large for the individual
but relatively small for the life expectancy of the entire population. The individual who smokes and is perhaps overweight
and inactive may lose 7, 8, or more years of life expectancy.
With about a quarter of the population smoking, the population
as a whole would gain about 11 2 years if every smoker quit.
Overweight individuals and those classified as obese would,
by returning to normal weight, gain as much as a half and 11 2
years of life expectancy, respectively. With more than half of
Britons and Americans overweight and over a fifth obese, their
return to normal weight could gain about a half-year of life
expectancy spread across the population. The inactive but
otherwise healthy individual can, by exercising, increase life
expectancy by as much as 12 months. Roughly adjusting for
confounding by social class, I estimate that the population would
gain 2 or 21 2 years of life expectancy if everyone assumed the
lifestyle of the fittest.28

Prospects for Better Health

How then should governments, committed to improving the
health of the populations they serve, invest for the future? The
provision of medical care, the development of healthier personal habits, and the creation of a more just social environment
each hold the potential to improve health. The benefits of
medical care that have already been achieved can be estimated
with considerable precision and they are substantial. The benefits of a healthier lifestyle have been achieved only to a very
limited degree, while inequalities in health have increased and
are continuing to do so.
If national policy is to be directed to improvement of the
health of the population as a whole and reducing inequalities in
health, such policy must take into account the potential and the
limitations of each. Increased investment in medical care would
make the greatest and most predictable contribution to the
reduction of death and to the relief or amelioration of suffering
and disability. The gains from increased investment in medical
care would begin to be seen almost at once, the benefits of
health promotion only as rapidly as the public responds with a
healthier lifestyle. Redistribution of wealth and resources for
the sole purpose of reducing inequalities in health would be a
long-term strategy of uncertain success.
Each strategy can potentially lessen inequalities in health. All
deserve urgent consideration, and it is a question of how much
to invest in each. The British government, if it hopes to improve
the health of everyone, and of the worst off in particular must
not place its hopes too heavily on lifestyle changes, nor should
it expect that social reform, however desirable in the name of
social justice, will resolve the problem of inequalities in health.
The governments stated commitment to the reduction of the
death rate from cancer, coronary heart disease, accidents, and
from suicide and trauma is, by contrast, a reasonable and attainable target for medical care.


pressure: overview of randomised drug trials in their epidemiological

context. Lancet 1990;335:82738.

1 McKeown T. The Role of Medicine: Dream, Mirage, or Nemesis? London:

Nuffield Provincial Hospitals Trust, 1976.

2 Illich I. Medical Nemesis: The Appropriation of Health. London: Calder and

16 Goldman L, Cook EF. The decline in ischemic heart disease mortality

rates: an analysis of the comparative effects of medical interventions

and changes in lifestyle. Ann Intern Med 1984;101:82536.
17 Hunink MGM, Goldman L, Tosteson ANA et al. The recent decline

Boyars, 1975.
3 Lederberg J. Medicines old battle against the bugs isnt over at all.

International Herald Tribune, 26 January, 1996.

4 Hitchings GH. Health care and life expectancy. Science 1993;262:1632.
5 Ableson PH. Improvements in health care. Science 1993;260:11.
6 Hadley J. More Medical Care, Better Health? Washington DC: Urban

Institute Press, 1982.

7 Charlton J, Velez R. Some international comparisons of mortality

amenable to medical intervention. Br Med J 1986;292:295301.

8 Mackenbach JP. The contribution of medical care to mortality decline:

McKeown revisited. J Clin Epidemiology 1996;49:120713.

9 Cochrane AL, St Leger AS, Moore F. Health service input and

mortality output in developed countries. J Epidemiol Community

Health 1978;32:20005.
10 Carr-Hill RA, Hardman GF, Russell IT. Variations in avoidable

mortality and variations in health care resources. Lancet 1987;i:

11 Bunker JP, Frazier HS, Mosteller F. Improving health: measuring the

effects of medical care. Milbank Q 1994;72:22558.

12 Bunker JP. Medicine matters after all. J Roy Coll Physicians 1995;

13 National Center for Health Statistics. Health United States, 199697 and

Injury Chartbook. Hyattsville, Maryland, 1997.

14 Drizd T, Dannenberg AL, Engel A. Blood pressure levels in persons

1874 years of age in 197680, and trends in blood pressure from

196080 in the United States. Vital and Health Statistics Series 11,
No. 234. DHHS 861684. Washington DC: US Government Printing
Office, 1986.
15 Collins R, Peto R, MacMahon S et al. Blood pressure, stroke,

and coronary heart disease Part 2, short-term reductions in blood

International Epidemiological Association 2001


Printed in Great Britain

in mortality from coronary heart disease, 19801990: the effect

of secular trends in risk factors and treatment. JAMA 1997;
18 Brenner MH, Curbow B, Javitt JC et al. Vision change and quality of

life in the elderly: response to cataract surgery and treatment of other

chronic ocular conditions. Arch Ophthalmol 1993;111:68085.
19 Bunker JP. Surgical manpower: a comparison of operations and

surgeons in the United States and in England and Wales. N Engl J Med
20 Bunker JP, Wennberg JE. Operation rates, mortality statistics and the

quality of life. N Engl J Med 1973;289:124950.

21 Roemer MI. More data on post-surgical deaths related to the 1976 Los

Angeles doctor slowdown. Soc Sci Med 1981;15C:16163.

22 Kohn LT, Corrigan JM, Donaldson MS (eds). To Err is Human:

Building a Safer Health System. Washington DC: National Academy

Press, 2000.
23 Starfield B. Is US health really the best in the world? JAMA 2000;284:

24 Freemont-Smith K. Adverse drug reactions in hospital patients (letter

to editor). JAMA 1998;280:1741.

25 Kravitz GR. Adverse drug reactions in hospital patients (letter to

editor). JAMA 1998;2801:741.

26 Rooney C. Increase in US medication-error deaths (letter to editor).

Lancet 1998;351:165657.
27 Brennan TA. The Institute of Medicine report on human errors

could it do harm? N Engl J Med 2000;342:112325.

28 Bunker JP. Medicine Matters After All. London: Nuffield Trust, 2001.
29 Hattersley L. Trends in life expectancy by social classan update.

Health Stat Q 1999;Summer:1624.

International Journal of Epidemiology 2001;30:12631267

Commentary: Can health outputs of routine

practice approach those of clinical trials?
Julian Tudor Hart

John Bunker entered anaesthetics in the late 1940s, just when

its leaders were starting its promotion from amateurism to its
present status as the most rigorous and audited physiological
practice in medicine. He thus became a participant yet but
objective observer of surgical advance over the next 50 years, of

Welsh Institute for Health & Social Care, Glyntaff Campus, University of
Glamorgan, Pontypridd CF37 1DL, UK. E-mail

its gains and of its collateral losses. His multi-authored book

Costs, Risks and Benefits of Surgery1 revived interest in clinical health
economics, pioneered by Codman in 1914.2
His important paper brings his work up to date, concluding
that possibly half our added years of life gained over the past 50
years may be attributable to medical intervention.3 As McKee
says in an excellent recent review of the efficacy of care,4 many
of his assumptions are optimistic. McKee points to the yawning
gap between what was possible in research trials, and what has



been achieved in practice, and doubts whether this gap will

close simply with passage of time.
Most research trials take place in a contrived world whose
population and whose professional carers both differ profoundly
from those in the world of real practice. Trials generally exclude
people with complex problems. For manual workers and their
families in areas of industrial decline, who present the greatest
challenge to medical services in developed economies, single
problems are exceptional. Using European self-reported data,
morbidity (any chronic disease or disability) lies around 73%
in the age group 5579, and comorbidity (two or more chronic
diseases or disabilities) lies at 60%.5 Most studies of general
practice detect far less comorbidity. For example five practices in
Holland with 25 534 patients were searched for casually detected
comorbidity, and showed isolated disorder for 89% of chronic
lung disease, 88% of hypertension, 74% of ischaemic heart
disease, 69% of diabetes, and 69% of degenerative arthritis.6 But
where problems are sought more systematically and proactively
(as they must be to define exclusions from trials) the higher
subjective rates are confirmed. After more than 20 years of
systematic proactive case-finding for major chronic health problems in Glyncorrwg, we accumulated 154 men and women with
treated hypertension. Searching these for evidence of 12 other
clinical problems, only 3% of the men and 7% of the women
had hypertension as a single disorder.
Socially responsible clinicians in countries with populations
registered for care (rather than undefined numbers of customers)
are responsible for everyone. They cannot exclude people because
their problems seem too complicated, though many seem in
practice to limit the problems they recognize to whatever they
feel competent to handle. Those excluded from most trials are
precisely those in most urgent need of care, and for whom care
might be most cost-effective. As Barbara Starfield has said, this
is the most important and most neglected field of research for
primary care.7 For routine practice to converge with routine
research, researchers must stop making things so deceptively
easy, and move closer to the real world.
The populations are too different, and so are the circumstances
in which professionals both choose and are compelled to do their
work. Research trials provide (usually within narrow limits)
planned, proactive care for defined populations, implemented
by clinicians with elementary public health skills and orientation.
Routine practice provides unplanned, essentially reactive care,
usually for undefined populations, often with resources of time,
staff and training which are insufficient to their task. Where
patients are still customers shopping around in a care market,
rational planning of any kind becomes almost impossible. To
bring research theory and routine practice together, the circumstances and attitudes of both sets of professionals must change.

Can We Narrow the Gap Between Trial

Possibilities and Practised Realities?
The short answer has to be yes, but only with immense difficulty.
Trial possibilities need to become more relevant, by ending
the policy of simplifying but falsifying exclusions. And practised
realities must stop using the products of science with little
regard for elementary rules of scientific rigour and integrity.
Doctors in practice inherit a culture that dared not claim
public health gain as its measurable outcome, because this

product would have seemed so small. Instead, it wanted full

marks simply for trying. Whatever merit this may once have
had, advance in medical knowledge has nullified. To mean well
is not enough, we must do well.8 This doing must be measurable
as health gainlonger, healthier lives, and healthier deaths. If
circumstances preclude good care, we share responsibility with
our patients to change them: our patients have votes, and we
have their names and addresses.
For problems amenable to treatment, affordable solutions
depend chiefly on improved anticipatory care of impaired health.
They depend much less than we or the public like to think, on
crisis intervention and salvage.
Like the quality of research, the quality of anticipatory care
depends on cumulative records rigorously maintained. These
become the material and verifiable expression of continuity
essential to effective care.9,10 In this context, continuity should
be defined not as patients seeing the same doctor, but as patients
not being compelled to repeat, simplify and ultimately distort
their stories, to suit all the different and often uncoordinated
professionals with whom they must deal. Real lives can only be
understood (and changed) as cumulative stories. Effective care
must reflect this reality, through clinical records that accumulate information, using personal stories as their stem. Naturally
this is easier for patients to understand, but it is also indispensable for professionals, who need not just the bits they want
for simple caricatures, but the entire context required for shared
judgements resulting in ultimate health gain. If trials come to
accept more representative problems as their material, they will
have to accept these same requirements for patients records.
Professionals actually responsible for delivering care to whole
populations know that the people with most to gain from relatively simple but effective care are those with the least access to
it, who are the most difficult to reach by proactive measures,
and the least attractive to traders in care as a marketed commodity.11 In terms of health gain, the key to rapid improvement in
clinical productivity lies in better treatment for poor, chronically
sick, unprofitable people.12 The most neglected but also the most
productive opportunities for effective care are concentrated
among the poorest people, so that resources concentrated on
them will normally give higher health gains per unit of investment, benefiting not only these patients, but the whole of society.
In the English language literature, I know only three examples
of valid comparison between the results of ordinary reactive care,
prompted by individual complaints, and continuing anticipatory
proactive care, prompted both by complaints and by perceived
personal and community needs. These are: a 2-year study in
New York which finished too soon to prove anything except that
patients liked it more, and it detected many more problems than
the ordinary reactive care then available;13 the US Hypertension
Detection and Follow-up Program (HDFP);1416 and my own
comparison between mortality rates in two South Wales coalmining communities,17 one managed proactively, the other
reactively. All three were sick populations, typical of those
posing our most intractable problems for health care.

US Hypertension Detection & Follow-up

Project 19761979
In 1976, when the HDFP began, treatment of mild to moderate
hypertensives in the diastolic range below 105 mmHg was


unsupported by controlled trials. Without such evidence, 78%

of doctors in New York State were already routinely prescribing
antihypertensive drugs from a diastolic threshold of 90 mmHg.
By 1979 (when HDFP results were available) this had risen
to 90%.18 It was therefore ethically indefensible to randomize
patients either to active treatment or placebo. Instead, the HDFP
first identified a large (11 000) hypertensive subset randomly
sampled from generally poorer populations (with most to gain
and therefore most likely to co-operate) and then allocated them
randomly either to whatever care was usually available from
the US care market (usual care) or to free, planned proactive
care by specially trained nurse practitioners (stepped care).
The results represent the only randomized controlled trial we
are ever likely to see, between marketed reactive care and
socialized proactive care in terms of their efficiency in producing
net health gain, measured as mortality from all causes, on an
intention-to-treat basis.
As expected, results after five years showed 26% fewer
cardiovascular deaths for people randomized to proactive care
than for people randomized to usual reactive care. Completely
unexpected, and still unconvincingly explained, this group also
had 13% fewer non-cardiovascular deaths. These results implied some substantial advantage for free, proactive organized
care for a whole population according to need, compared with
operation of the usual US commodity market in medical care.
The reality of this conclusion was accepted (though not in these
terms) at the 89th annual meeting of the Association of Life
Insurance Medical Directors of America.19 Retrospectively, it seems
that many, probably most, of the nurse practitioners giving
stepped care had helped patients in many ways other than
by controlling their blood pressures, rightly believing that
continued compliance largely depended on comprehensive approaches to care and willingness to listen to the wider concerns.
Usual care from usual doctors remained predominantly reactive
and narrowly focused. As the trial was not planned to address
such questions, this interpretation is necessarily speculative, but
it seems the most obvious and plausible.
Another anomaly was that whereas in virtually all trials
benefits of treatment are greatest in patients with most deviant
indices of disease, in the HDFP differences in death rates
were greatest for hypertensives in the lowest diastolic range
90104 mmHg. In this group there was no significant difference
between results of usual care and proactive care for white
women, but for black women deaths from all causes were 28%
lower in the proactive care group.15 Overall, 77% of the difference in deaths between proactive and reactive care occurred in
black patients, though they composed only 44% of the trial
population,20 and only about 11% of the whole US population.
Though the trial population was drawn from screened random
samples, these were selected from localities served by US
medical schools, which are generally poorer and sicker than the
US population as a whole.
They were also much poorer, and very much sicker, than the
UK populations sampled by the superficially similar Medical
Research Council (MRC) trial of treatment for mild hypertension, whose results appeared six years later, showing barely
significant benefits from treatment in this diastolic range. With
economic incentives to minimize rather than maximize their
patients needs, UK doctors were more sceptical than their
US colleagues, and less eager to act beyond the evidence.21


The MRC trial could therefore be a straightforward comparison

between screened people in the diastolic range 90104 mmHg
randomized to active or placebo treatment, double blind within
the same care system.22 Death rates during this trial, both from
stroke and from all causes, were almost one-third, and coronary
deaths were less than half, of those in the HDFP. Mortality and
morbidity differences between these two trials were probably
socioeconomic rather than racial.23 The HDFP was selective for
poor communities, and thus may have exaggerated benefits
for the US population as a whole. The MRC study was selective
for affluent communities, and may well have understated
potential benefits for poorer UK communities.
Community management of high blood pressure was, and
still is, a useful model for community management of all chronic
disorders. In the US, over 45% of non-institutionalized citizens
have one or more chronic disorder. Their direct costs account for
70% of US health care expenditure,24,25 and the UK proportion
is probably similar. Most of these patients are neither elderly
nor disabled, though obviously prevalence and complexity rise
steeply with age. Effective and economic management of all
chronic problems and combinations of problems depends on
enlisting patients as co-managers of their problems and coproducers of their own health, rather than passive and credulous
consumers.26 No amount of technically excellent care can maximize health outcomes if patients are not actively engaged in
managing their own problems, particularly chronic problems,27
with continuing support. These factors were evidently provided
more consistently by free proactive care through the HDFP than
by usual care through the medical market.

Afan Valley 25-year Study

From 1961 to 1987 I was responsible for organizing most of the
primary care for a population of around 1800 in the coal-mining
village of Glyncorrwg, remaining as research observer to 1992.
Starting from a local census in 1965 to establish an accurate
denominator, and screening of the whole population aged
2079 for blood pressure in 1968,28 all hypertensives meeting
contemporary evidence-based criteria were offered treatment and
indefinite follow-up at not more than 3-monthly intervals with
proactive recall.29 Over the following 20 years we gradually
extended systematic case-finding throughout this population
(with proactive search for the small minority of non-consulters
over a 5-year period) for other chronic disorders, including smoking, chronic lung diseases, diabetes, obesity, and alcohol excess.
In 1991 we compared mortality under 65 for 19811986 in
this population, with that in Blaengwynfi, a neighbouring population of similar social composition receiving usual reactive care
of a high standard. Ratio of actual to expected deaths (derived
from mean UK rates) was 21:22 for Glyncorrwg (proactive care)
and 48:30 for Blaengwynfi (reactive care). Using Townsend Index
as a measure of social deprivation, Glyncorrwg and Blaengwynfi
both ranked in the worst 5% of the distribution in 55 electoral
wards of West Glamorgan County. For age-standardized mortality under 65, Glyncorrwg ranked fourth lowest, Blaengwynfi
32nd lowest out of 55.30 Note that though continuing proactive
care seemed more effective than good reactive care, good reactive
care did seem to have an effect.
So far as it goes, this evidence suggests that in poor communities
which retain high social morale, planned, proactive continuing



care within a public health-oriented strategy can compensate for

some of the lethal effects of social deprivation, and do so more
effectively than good reactive care alone. The italicized proviso
is critical. Since defeat of the miners in 1985 and decline of most
UK manufacturing industry, there has been a disintegration
of social morale, particularly in young men, creating a new
and for this purpose untested situation. This has occurred in
Glyncorrwg as much as any other former mining community,
with health consequences, which though no longer measured,
are only too obvious.
This study was limited by small numbers, wide confidence
limits, and opportunistic design. However, it provided detailed
evidence on the extent of hitherto unmet needs,31 about how
work was organized, the staff numbers, skills and attitudes required to perform it, and measures of exactly what was achieved
in changing (or failing to change) indicators of risk. For example,
male smokers fell from 61% in 196870 to 36% in 1985 (verified
by urine cotinine measurement), but women showed no
significant change (43% to 42%). There were also no significant
changes in body mass index or total blood cholesterol (this was
the pre-statin era).
All three examples suggest ways in which actual care can
converge with scientific experiment. Epidemiological studies
and trials need to learn from best practice in continuing primary
care, with many more long-term cohort studies integrated with
routine care, and far more active roles for patients in research
as well as in care processes.32 Primary care practice must cease
to be the personal fiefdom of GPs, and must become properly
staffed and resourced to perform its full social function as the
principal site for ambulatory and continuing care.

The gap between what research suggests may be possible, and
what is actually achieved in routine practice, can and should
be reduced. To some extent this can be done by designing trials
that are socially more realistic, and addressing the problems of
clinicians and patients on the ground, which means both these
groups should be represented at the design stage. Chiefly it
depends on redefinition of medical professionalism, to include
social responsibilities once marginalized to Public Health as a
subordinate and often homeless speciality, and a large redistribution (as well as expansion) of resources to allow this redefinition
material expression.
Whatever its value compared with other social factors, the
relative contribution of medical care is growing and will continue to do so. As Bunker says, this means that failure to
implement existing medical knowledge for all who can benefit
is becoming more rather than less important as a cause of social
class differences in health. I am not convinced that attempts
to quantify the relative contributions of medical care or of other
social variables are useful, meaningful, or perhaps even possible.
In practice, wider availability of good medical care has gone
hand in hand with rising wages, rising employment, wider provision of social housing, and more generous provision of free
public education. All these public services have advanced together,
and in periods of social retreat, all have suffered. Voluntary
retreat by one has not led to preferential advance for another,
but simply reduced the overall burden of public services on
those too rich to need them.

I am grateful to Tony Beddow, Nick Gould, Ben Hart, Sheila Ryan
Johansson, Morton Warner and Graham Watt for necessary and
helpful criticism.

1 Bunker JP, Barnes BA, Mosteller F. Costs, Risks and Benefits of Surgery.

New York: Oxford University Press, 1977.

2 Codman EA. The product of a hospital. Surgery, Gynaecology & Obstetrics

1914;18:49196. Also available in: White KL, Frenk J, Ordonez C et al.

Health Services Research: An Anthology. Washington DC: PAHO, 1992.
3 Bunker J. Commentary: The role of medical care in contributing to

health improvements within societies. Int J Epidemiol 2001;30:126063.

4 McKee M. For debatedoes health care save lives? Croatian Med J

5 van den Bos GAM. The burden of chronic diseases in terms of

disability, use of health care and healthy life expectancies. Eur J Public
Health 1995;5:2934.
6 Schellevis FG, van de Lisdonk EH, van der Velden J et al. Consultation

rates and incidence of intercurrent morbidity among patients with

chronic disease in general practice. Br J Gen Pract 1994;44:25962.
7 Starfield B. New paradigms for quality in primary care. Br J Gen Pract

8 Eisenberg L. Science in medicine: too much, or too little and too

limited in scope? Am J Med 1988;84:48391.

9 Riddle MC. A strategy for chronic disease. Lancet 1980;ii:73436.
10 Kaplan SH, Greenfield S, Ware JE. Assessing the effects of physician-

patients interactions on the outcomes of chronic disease. Medical Care

11 Hart JT. The Inverse Care Law. Lancet 1971;i:40512.
12 Watt GCM, Ecob R. Analysis of falling mortality rates in Edinburgh

and Glasgow. J Public Health Med 2000;22:33036.

13 Goodrich CH, Olendski MC, Reader GG. Welfare Medical Care: An

Experiment. Harvard University Press, 1970.

14 Hypertension Detection & Follow-up Program Co-operative group.

Five-year findings of the HDFP. I. Reduction in mortality of persons

with high blood pressure, including mild hypertension. J Am Med Assoc
15 Hypertension Detection & Follow-up Program Co-operative group.

Five-year findings of the HDFP. II. Mortality by race, sex and age. J Am
Med Assoc 1979;242:257277.
16 Hypertension Detection & Follow-Up Program Cooperative Group.

Mild hypertensives in the Hypertension Detection and Follow-Up

Program. Ann NY Acad Sci 1978;304:25466.
17 Hart JT, Thomas C, Gibbons B et al. Twenty five years of audited screen-

ing in a socially deprived community. Br Med J 1991;302:150913.

18 Guttmacher S, Teitelman M, Chapin G et al. Ethics and preventive

medicine: the case of borderline hypertension. Hastings Center Report,

February 1981.
19 Alderman MH. Management of mild hypertension: a review of

relevant studies. In: Rafter JG (ed.). Viewpoints on the treatment of

hypertension: panel discussion. Transactions of the Association of Life
Insurance Medical Directors of America 1980;64:13045.
20 Peart WS, Miall WE. MRC treatment trial for mild hypertension.

Br Med J 1979;ii:48.
21 Hart JT. Expectations of health care: promoted, managed, or shared?

Health Expectations 1998;1:313.

22 Medical Research Council Working Party. MRC trial of treatment of

mild hypertension: principal results. Br Med J 1985;291:97104.



23 Tyroler HA. Race, education, and 5-year mortality in HDFP Stratum I

29 Hart JT. The marriage of primary care and epidemiology: continuous

referred-care males. In: Gross F, Strasser T (eds). Mild Hypertension:

Recent Advances. New York: Raven Press, 1983, pp.16376.

anticipatory care of whole populations in a state medical service

(Milroy lecture). J R Coll Physicians London 1974;8:299314.

24 Hoffman C, Rice D, Sung HY. Persons with chronic conditions: their

30 Kaul SA, Thomas C. Study of Small Area Statistics in West Glamorgan.

prevalence and costs. J Am Med Assoc 1997;277:147379.

25 Holman H, Lorig K. Patients as partners in managing chronic disease.

Br Med J 2000;320:52627.
26 Hart JT. Clinical and economic consequences of patients as producers.

J Public Health Med 1995;17:38386.

27 Kaplan SH, Greenfield S, Gandek B, Rogers WH, Ware JE. Character-

istics of physicians with participatory decision-making styles. Ann Intern

Med 1996;124:497504.
28 Hart JT. Semicontinuous screening of a whole community for hyper-

tension. Lancet 1970;ii:22327.

International Epidemiological Association 2001

Printed in Great Britain

Swansea: Department of Community Medicine, West Glamorgan

Health Authority, 1988.
31 Hart JT. Rule of Halves: implications of underdiagnosis and dropout

for future workload and prescribing costs in primary care. Br J Gen

Pract 1992;42:11619.
32 Hart JT, Smith GD. Response rates in south Wales 19501996:

changing requirements for mass participation in human research.

In: Chalmers I, Maynard A (eds). Non Random Reflections on Health
Services Research: on the 25th Anniversary of Archie Cochranes Effectiveness
& Efficiency. London: BMJ Publishing Group, 1997.

International Journal of Epidemiology 2001;30:12671268

Commentary: Medical care and the wider

influences upon population health:
a false dichotomy
Stephen Frankel

It is a dull fact that where polarized positions exist either there

is no single truth, or the truth lies somewhere between the
fortified positions of the key protagonists. McKeowns presentation of historical mortality patterns1 and Illichs nihilistic
polemic2 came as refreshing antidotes to both the heroic history
of past therapeutic triumphs and breathless optimism about the
capacity of post-war science to conquer common chronic diseases in the way it appeared, then, to have conquered the major
communicable diseases. Works of that sort, with the added spice
of Cochranes iconoclastic view of the scientific foundations
of clinical activity,3 became the tracts that assuaged the public
health doctors lowly medical status, rather as the Bible may
ease the daily humiliation of Jehovahs Witnesses as doors are
slammed in their faces. The comfort that these works offered
was, however, not without cost, as the implicit or explicit assertion
that the medical care of the later 20th century did not contribute to health improvements was so counter-intuitive that
this view may have contributed to the increasingly marginal
position of public health.
The fact that major improvements in health status occurred
prior to the availability of effective measures does not mean
that a significant proportion of recent improvements cannot
be ascribed to medical care. Mainstream opinion responded
to the view that medical care can take no credit for health
Department of Social Medicine, University of Bristol, Canynge Hall,
Whiteladies Road, Bristol BS9 2PR, UK.

improvement largely by ignoring it. John Bunker has been the

main authority to take on this debate in its own terms. Over
many decades he has performed two major services that reflect
his balanced view of the merits of medical treatment. First he
was a key pioneer of the critical, quantitative analysis of the
benefits of medical treatment through the study of treatment
variations. Second, he offered the formal counterbalance to the
views of McKeown and others. This review brings up to date his
ideas on the relative importance of medical care in explaining
20th century increases in life expectancy, and outlines their
policy implications.4
He highlights the two difficulties with ascribing major
health improvements to medical care. First he considers, fairly,
the matter of iatrogenic disease and death. Second he explores,
with less balance, the impact of socioeconomic influences upon
mortality. While acknowledging a 91 2-year difference in life
expectancy between men in professional and unskilled occupations, he neutralizes this issue on the grounds that where the
very rich and the very poor comprise only a small proportion of
each population, such inequalities can have only a small effect
upon overall population trends. His estimate of 1122 years net
effect on life expectancy is probably an underestimate, given,
for example the 4-year life expectancy difference between the
highest and lowest septile for English health authorities for
19921994.5 Such differences are a serious underestimate of the
true attributable health effects of socioeconomic disadvantage,
since they do not take into account lifetime social circumstances



and are based on comparing large heterogeneous areas, and thus

substantially misclassifying individual experiences of deprivation.5
But the main problem here is the assertion that improvements in public health (are) essentially complete , with the
clear recommendation that medical care must be the overriding
priority for national policy. This overstates the case, and serves
to distract from the fact that the environmental effects seen in
19th century mortality patterns are still available to deprived
populations within industrialized countries. Modelling the wider
determinants of health is also complicated by the proper absence
of trial evidence from randomized individuals on the effects of
population measures; the idiom of evidence-based medicine is
now being applied inappropriately to the consideration of the
merits of broader investments in health.6 The view that income
redistribution can be justified outside any technical discussion
of health status is not considered. The suggestion that government commitments to reducing suicide deaths can be met through
medical care is surprising, given the uncertain effectiveness of
relevant interventions and uncertain access to those who may
be at risk of suicide.7
What is the effect of making the assertion that medical care
accounts for a high proportion of increased life expectancy and
quality of life? The case is here made for more medical care. But
what does this mean? Medical care is not a unitary activity
where having more is unequivocally beneficial. The reasonable
case is for more benefits of health care, rather than more health
care per se.8 In some instances there is no doubt that more is
better. For example, there is little doubt that there is underprovision of cataract extraction in the UK, with consequent
unnecessary disability; we can therefore feel confident that more
cataract extraction is desirable. Many other instances are more
complex. The major discrepancy between the rates of coronary
revascularization procedures between the US and UK is
amongst older people; amongst those aged 7584 the difference
is over tenfold. The question is then, are older people in the UK
being deprived of an unequivocal benefit, or are older people
in the US being exposed to unwarranted risks? It is important
to stress that there are no formal answers to this question at
the moment as the trial evidence that exists is based upon outcomes amongst younger people. This is a specific example of the
general problem in reconstructing the population effects of
medical care where participants in trials differ from the population
experiencing conventional services in terms of age and other

characteristics, including multiple pathology. Not only do international studies not show any mortality benefit from higher
levels of coronary revascularization, there is some evidence that
countries with the most aggressively interventionist approach
have higher rates of other adverse events, such as stroke.9 The
issue is therefore as much a matter of what approach to the
morbidity of declining years individuals in different cultures
may favour; implicit assumptions that more is better are therefore unhelpful.
The time has probably passed when the simple opposition of
medical care and other determinants of health has any value.
The more interesting questions concern which components of
medical care and which influences upon other determinants of
health justify investment. However, the discourse in health
policy has been dominated by the idiom of the health economist
where choices are always hard and costs represent opportunities
forgone.10 If we have to express these complex issues in terms
of a continuum between increasing investment either in the
wider determinants of health or in medical care, the truth, for
the UK at least, lies at both ends.

1 McKeown T. The Role of Medicine. Oxford: Basil Blackwell, 1979.
2 Illich I. Medical Nemesis. London: Calder and Boyars, 1975.
3 Cochrane AL. Effectiveness and Efficiency. London: The Nuffield

Provincial Hospitals Trust, 1971.

4 Bunker JP. The role of medical care in contributing to health improve-

ments within societies. Int J Epidemiol 2001;30:126063.

5 Shaw M, Dorling D, Gordon D, Davey Smith G. The Widening Gap.

Bristol: The Policy Press, 1999.

6 Davey Smith G, Ebrahim S, Frankel S. How policy informs the

evidence. Br Med J 2001;322:18485.

7 Gunnell DJ, Frankel SJ. Prevention of suicide: aspirations and

evidence. Br Med J 1994;308:122733.

8 Frankel S, Ebrahim S, Davey Smith G. The limits to demand for health

care. Br Med J 2000;321:4044.

9 Yusuf S, Flather M, Pogue J et al. Variations between countries in

invasive cardiac procedures and outcomes in patients with suspected

unstable angina or myocardial infarction without initial ST elevation.
Lancet 1998;352:50714.
10 Davey Smith G, Frankel S, Ebrahim S. Rationing for health equity: is

it necessary? Health Econ 2000 (in press).