I thank Dr Adamiak for her careful reading of the glossary.
I believe the confusion may be motivated by the fact that the sentence she
refers to is not as clearly stated as it could have been. By "in the case
of continuous dependent variables" I meant more precisely "when the
response variable is normally distributed and the link function is
identity (ie models usually referred to as linear models)". By "i...
I thank Dr Adamiak for her careful reading of the glossary.
I believe the confusion may be motivated by the fact that the sentence she
refers to is not as clearly stated as it could have been. By "in the case
of continuous dependent variables" I meant more precisely "when the
response variable is normally distributed and the link function is
identity (ie models usually referred to as linear models)". By "in the
case of non-normally distributed variables (for example, logistic models)"
I meant more precisely "in the case of other models such as logistic
models (where the response variable is the log odds of a binary outcome)".
Of course, as Dr Adamiak notes, linear regression could be performed
using dichotomous variables, and continuous variables can be dichotomized
and modelled using a logistic model. The essence of the statement however
is correct: In the linear model (as defined above), regression
coefficients derived from random effects and population-average models are
equivalent but in the logistic model they are not (see references 46 and
47 in the glossary for more details). I apologize for the confusion and
thank Dr Adamiak for the opportunity to clarify this.
It would be interesting to know the overall health condition of the
subjects. Were they average weight, average stressed occupations, diets.
What other environmental conditions could have contributed to their heart
problems, if any.
The Authors draw conclusion that the use of acute hospital beds does
not increase as the population ages, which is a result from a seven year
cohort study in Germany. The problem is however that there are supply
related factors, which can strongly affect the results. It is known that
there are substitutions between different forms of care.[1] The
days spent in hospitals and the mortality rates, in part...
The Authors draw conclusion that the use of acute hospital beds does
not increase as the population ages, which is a result from a seven year
cohort study in Germany. The problem is however that there are supply
related factors, which can strongly affect the results. It is known that
there are substitutions between different forms of care.[1] The
days spent in hospitals and the mortality rates, in particular among the
oldest populations, are affected by the access to elderly care, e.g.
providers of palliative care in the terminal stage of life.[2] In Sweden the reform of the care for the elderly meaning transfer
of responsibility for the care of the elderly to municipalities led to an
increase of death rates in some of the nursing homes providing long term
care and after care to the elderly people.[3,4] Also, the supply of beds at geriatric departments might be
associated with the number of inhospital deaths as opposed to deaths
occurring in other places.
The stable patterns in inhospital death rates among German inpatients
might be to a large degree explained by the growing supply of other forms
of care of the elderly, which authors do not appear to take into account
in their calculations and the following discussion. The impact of supply
of various kinds of inpatient beds is an important confounder in studies
that calculate inhospital mortality and beds use as proxies for morbidity
and related mortality (or health outcomes) without considering effects of
supply and proximity known by all health economists and epidemiologists.
Proxy measures or performance indicators such as use of hospital beds are
conceptually different from outcome indicators or measures of health care
needs.[5-8] Use or supply is often confounded with needs and a
serious methodological problem.[9,10]
The knowledge about the supply inducement is used as a rationale
behind many management decisions in health care, so called "chains of
care" and by HMO in USA. Patients treated for acute conditions are
transferred to other forms of after care when their condition stabilizes.
As Ashton et al.[11] point out physicians will discharge patients
prematurely if there is a safety net outside hospitals, which can provide
prompt aftercare. Some patients are requiring to be discharged and die at
home, which is a consequence of free choice, often used by terminally ill
cancer patients. Thus, the conclusion about health care needs of acute
care among the German as well other populations based on consumption of
hospital days as proxy for health care needs is strongly biased. It is
rather a matter of provision of inpatients beds and integration between
various providers.[4] Inpatient beds might in fact be
located outside hospitals or in patient homes, and those deaths are not
registered in hospital discharge data bases, which often serve as the
primary source of information in many studies. It must be more appropriate
to calculate total death rates to make inferences about changing health
care needs among populations. A health care system perspective[12] and account into total population instead solely into use of
acute care hospitals and inhospital deaths would be more fruitful in the
context of projections of health care needs.
Grazyna Adamiak
References
(1) Culyer AJ. The morality of efficiency in health care – some
uncomfortable implications. Economics of health care systems. Health
Economics 1992;1:7-18.
(2) Fisher J, Wennberg JE, Stukel TA, Sharp SM. Hospital readmission
rates for cohorts of medical beneficiaries in Boston and New Haven.
New England Journal of Medicine 1994;331:989-95.
(3) The Federation of County Councils. Place of death project (Plats för
död-projektet) http://www.lf.se
(4) Andersson G, Karlberg I. Integrated care for the elderly. The background
and effects of the reform of Swedish care of the elderly. International
Journal of Integrated Care (IJIC) 2000. http://www.ijic.org/
(5) Morrill RL, Earickson R. Hospital Variation and Patient Travel
Distances. Inquiry 1968;5:26-34.
(6) Wennberg JE, Barnes BA, Zubkoff M. Professional uncertainty and the
problem of supplier-induced demand. Social Science and Medicine
1982;16:811-834.
(7) Hornbrook MC. Practice Mode and Payment Method. Effects on Use,
Costs, Quality and Access. Medical Care 1985;23:484-511.
(8) Giuffrida A, Gravelle H, Roland M. General Practice. Measuring
quality of care with routine data: avoiding confusion between performance
indicators and health outcomes. British Medical Journal 1999;319:94-98.
(9) Carr-Hill RA, Jamison JQ, O'Reilly D, Stevenson MR, Reid J, Merriman
B. Risk adjustment for hospital use using social security data: cross
sectional small area analysis. British Medical Journal 2002;324:390.
(10) Gibson A, Asthana S, Brigham P, Moon G, Dicker J. Geographies of need
and the new NHS: methodological issues in the definition and measurement
of the health needs of local populations. Health and Place 2002;8:47-60.
(11) Ashton CM, Wray NP. A conceptual framework for the study of early
readmission as an indicator of quality of care. Social Science and
Medicine 1996;43(11):1533-1541.
(12) Nutting PA, Shorr GI, Burkhalter BR. Assessing the Performance of
Medical Care Systems: A Method and its Application. Medical Care
1981;19(3):281-296.
I have found that under the heading Population-average models (page
592), when comparing the multilevel models to population-average models,
the Author is stating that in the case of continuous dependent variables
the coefficients are mathematically equivalent in the marginal models. In
the next phrase the Author suggest "...but in the case of non-normally
distributed variables (for example, logistic...
I have found that under the heading Population-average models (page
592), when comparing the multilevel models to population-average models,
the Author is stating that in the case of continuous dependent variables
the coefficients are mathematically equivalent in the marginal models. In
the next phrase the Author suggest "...but in the case of non-normally
distributed variables (for example, logistic models) etc.".
Linear regression might be performed using dichotomous variables as
well as dependent variables. The issue of mathematical equivalence is some
what tricky, the value of dichotomous variables, if 1, might be also
regarded as continuous whether mathematically there is infinity and we can
continue to count 0.999999999 etc.
There is no contradiction between the first statement and the second,
after "but". Continuous variables might also be non-normally distributed,
which is quite common in the context of measurements in health care, for
example age distribution among inpatients is usually strongly skewed.
Logistic regression might be performed on normally distributed variables
as well.
The logistic regression is based on the assumption of binomial
distribution, which means that it does not matter if the variables are
normally distributed or not. Both in the multiple linear and the logistic
regressions there is an assumption about the variance of the outcome
variable. In the linear model there is an expectation that the variance of
the outcome variable is equal around the mean whether in the logistic
regression the variance depends only on the mean. The linear models model
the mean value of the outcome whether the logistic the logarithm of the
odds of the outcome (referred to as logit). The relationship of multiple
independent variables to outcome in the linear regression is such that the
mean value of outcome changes with linearly with multiple independent
variables. In the logistic model, the logit of outcome changes linearly
with multiple independent variables. Thus, both are based on the
assumption of linearity and averages appear to play a role in both.
Grazyna Adamiak
Reference
(1) Katz MH. Multivariable analysis. A practical guide for Clinicians. Cambridge:
Cambridge University Press.
Professor Muntaner's cricism of the neglect of power relations in
research on health inequality is well timed. The 2001 Census of England
and Wales has used as its social classification a new measure: the
National Statistics Socio-economic classification (NS-SEC). This measure
has an explicit theoretical basis in the relations and conditions of
employment. It therefore distinguishes those who own pr...
Professor Muntaner's cricism of the neglect of power relations in
research on health inequality is well timed. The 2001 Census of England
and Wales has used as its social classification a new measure: the
National Statistics Socio-economic classification (NS-SEC). This measure
has an explicit theoretical basis in the relations and conditions of
employment. It therefore distinguishes those who own property and
businesses from those who do not, the self employed with and without
employees, and employees with different degrees of power over their own
work and that of others. It makes no reference to individual
characteristics such as skill or education, nor to status or prestige.
Publications are now emerging that use this measure, for example, in
examining inequalities in health.
Modern obstetritions in developed nations would do well to emulate
Traditional Birth Attendents. These women are much less apt to cause
iatrogenic injury, which is so often unknowingly done by OB/GYNs in
modern actively managed births when the umbilical cord is clamped almost
immediately. This prevents normal placental transfusion and causes
hypoxia and hypovolemia in the newborn.
Modern obstetritions in developed nations would do well to emulate
Traditional Birth Attendents. These women are much less apt to cause
iatrogenic injury, which is so often unknowingly done by OB/GYNs in
modern actively managed births when the umbilical cord is clamped almost
immediately. This prevents normal placental transfusion and causes
hypoxia and hypovolemia in the newborn.
I think it is time to renew ideas at PAHO. Drs Roses and Casas had
been high level officers for many years, why did not they implemented
their suggestions? Dr Sepúlveda is a bright public health man from Mexico,
with an oustanding curriculum. Lets work towards improving PAHO's activities with
new people.
The discussion triggered by this paper is a useful one.
In general I agree with McLeod, and Davey Smith that we
have to be careful when we make conclusions regarding
etiology. Socioeconomic factors could certainly confound
the relationships. Adjustment for socioeconomic factors
could also conceal true relationships, however. In this
literature (see the review by Schnall et al. 2000 which
has been referred...
The discussion triggered by this paper is a useful one.
In general I agree with McLeod, and Davey Smith that we
have to be careful when we make conclusions regarding
etiology. Socioeconomic factors could certainly confound
the relationships. Adjustment for socioeconomic factors
could also conceal true relationships, however. In this
literature (see the review by Schnall et al. 2000 which
has been referred to previously in the correspondence
and also Theorell et al. 1998) it has been found that the
demand control model is more important in blue collar
worker strata than in white collars. Adjustment for
social class certainly does not solve the problem that
psychosocial factors may be related to cardiovascular
illness risk in different ways in different social
classes. In addition, if we adjust for social class in
analyses of the relationship between psychosocial
factors at work and cardiovascular illness risk we are
neglecting the possibility that one of the reasons why
people in lower social classes have more cardiovascular
disease might be that they might be humiliated due to
bad working conditions (with resulting accumulating
psychophysiological reactions) more often at work than
those in higher classes.
Still, I agree that it is relevant to do such adjustments
for social class but we have to interpret the total
picture in the best possible way. Accordingly we need to
do analyses that are both adjusted and non-adjusted for
social class. Of course I agree that if adjustment is
made it should be done in the best possible way and we
should try to avoid residual confounding.
There is one passage in their letter that needs to be
commented. They refer to our paper (Orth-Gomir et al.
1994) describing our intervention as a two-day course.
The two-day course was only part of the total program. I
quote from our own paper:
"In the third component of the stress management
program, self-initiatives were solicited from workers to
improve the work stress situation according to the three
aspects of this theory (demand-control-support). This
was stimulated by having each group of workers set up an
action plan consisting of a a list of problems to
attack and factors to improve."
Thus, the third component of the intervention lasted for
several weeks. One of the aims of the initial course was
to increase awareness in employees regarding what the
relevant factors might be and to stimulate them to make
suggestions for improvements. These suggestions were
followed up systematically and did result in some
organisational changes. In these particular work sites
high work load was not the main factor but rather other
structural problems. This study was not ideal because it
was based upon small samples and the intervention had a
mix of both structural and individual components.
However, it was not primarily an effort to make the
employees think that they had more say at work than they
had! We certainly need more intervention research in
this field.
References
(1) Hallqvist J, Diderichsen F, Theorell T, Reuterwall C,
Ahlbom A.: Is the effect of job strain on myocardial
infarction risk due to interaction between high
psychological demands and low
decision latitude? Results from Stockholm Heart
Epidemiology Program (SHEEP).
Soc Sci Med 1998 Jun;46(11):1405-15.
(2) Orth-Gomir K, Eriksson I, Moser V, Theorell T and
Fredlund P. Lipid lowering through work stress
reduction. Int J Beh Med 1994 1:204-14
(3) Schnall PL, Belkic K, Landsbergis P och Baker D. The
workplace and cardiovascular disease. State of the Art
Reviews. Occ Med 2000; 15(1).
(4) Theorell T, Tsutsumi A, Hallquist J, Reuterwall C,
Hogstedt C, Fredlund P, Emlund N, Johnson VJ, and the
SHEEP Study Group. Decision Latitude, Job Strain, and
Myocardial Infarction: A Study of Working Men in
Stockholm. American Journal of Public Health 1998; 88
(3):382-8.
We are grateful to Peter and colleagues for taking the trouble to
respond to our e-letter. They raise several issues that we will try to
address briefly.
First, they say effort-reward imbalance had no relation to socio-economic status in their study. We are unclear how they assessed this –
the only SES measure they mention in their paper is a white-collar/blue-
collar binary distinction. Effort–r...
We are grateful to Peter and colleagues for taking the trouble to
respond to our e-letter. They raise several issues that we will try to
address briefly.
First, they say effort-reward imbalance had no relation to socio-economic status in their study. We are unclear how they assessed this –
the only SES measure they mention in their paper is a white-collar/blue-
collar binary distinction. Effort–reward imbalance may well have varied
with SES without such variation being apparent in relation to such a crude
measure. Furthermore the fact that effects were apparent amongst
“homogenous socio-economic groups” does not rule out residual confounding
as a possible explanation. For confounding to be possible the exposure
under assessment simply needs to be a better proxy for social position
than the stratification measure by which these “homogenous” groups are
defined.[1] Again, if the groups are defined as broadly as “white-collar”
and “blue collar” then residual confounding seems a distinct possibility.
Second, we don’t understand the basis for their suggestion that we
failed to consider experimental evidence as a criterion for causality in
this context and were thus in some way guilty of bias. A substantial part
of our letter related to this very point – as Peter and colleagues appear
to concede in the first sentence of their fourth paragraph. The
experimental evidence that we cited was that examining effects in relation
to objective cardiovascular endpoints[2] – as opposed to intermediate
outcomes of questionable significance or subjective outcomes that might be
influenced by response bias. None of the publications cited by Peter and
colleagues appear to relate to any additional evidence of this former
type. For example one of the specific studies they mention described a two
-day educational intervention for civil servants aimed at increasing
perceptions of workplace control.[3] Perceptions of control were
increased in the intervention group, but other than a marginal change in
apolipoprotein B to A1 ratio, objective parameters were unaffected. The
nature and magnitude of these effects (or non-effects) are typical of
those seen with interventions targeting work stress.[4] We applaud the
efforts of Peter and others to evaluate such interventions more
rigorously. However till these evaluations demonstrate more impressive
results, we stick with our original suggestions as to the current most
evidence-based strategy to improve population cardiovascular health and
reduce health inequalities.
References
(1) Phillips AN, Davey Smith G. How independent are independent
effects? Relative risk estimation when correlated exposures are measured
imprecisely. J Clin Epidemiol 1991;44:1223-31.
(2) Louis AA, Manousos IR, Coletta AP, Clark AL, Cleland JG. Clinical
trials update: The Heart Protection Study, IONA, CARISA, ENRICHD, ACUTE,
ALIVE, MADIT II and REMATCH. Eur J Heart Fail 2002;4:111-6.
(3) Orth-Gomer K, Eriksson L, Moser V et al. Lipid lowering through work
stress reduction. Int J Behav Med 1994;1: 204-14.
(4) van der Klink JJL, Blonk RWB, Schene AH, van Dijk FJH. The benefits of
interventions for work-related stress. Am J Public Health 2001;91:270-6.
We agree with John Macleod concerning the importance of a better
understanding of social inequalities in health. Yet, the paper to which he
refers (Peter et al., 2002) [1] was not primarily intended to explain social
inequalities, but rather to demonstrate an improved prediction of CHD risk
by combining standardized measures of two innovative theoretical concepts
of psychosocial stress at work.
M...
We agree with John Macleod concerning the importance of a better
understanding of social inequalities in health. Yet, the paper to which he
refers (Peter et al., 2002) [1] was not primarily intended to explain social
inequalities, but rather to demonstrate an improved prediction of CHD risk
by combining standardized measures of two innovative theoretical concepts
of psychosocial stress at work.
Macleod’s claim that confounding may invalidate reported findings and that
"the two psychosocial measures were proxies for social position" is not
justified in our case.
First, the association between job stress and SES, while significant for
demand-control, was non-existing for effort-reward imbalance. If there is
an association between effort-reward imbalance and SES it may be analysed
in terms of effect-measure modification.
Second, associations of job-stress with CHD and cardiovascular risk
factors (e.g. hypertension, atherogenic lipids) were reported for
different, heterogeneous as well as homogeneous socio-economic groups,
thus ruling out confounding as a major explanation (see e.g. Bosma et al.
1998,[2] Theorell et al. 1998,[3] Lynch et al. 1997,[4] Peter et al. 1997, 1998.[5,6]
We agree that controlled intervention studies are needed as an additional
source of evidence when drawing causal inferences in epidemiology. In
fact, at least a few such studies, not cited by the author, have been
conducted, partly with promising findings (e.g. Orth-Gomer et al. 1994,[7]
Kristensen 2000,[8] Theorell et al. 2001[9]). It is hard not to interpret
Macleod´s comments as biased as he does not consider additional criteria
of causality in epidemiology, for instance experimental evidence of
psychosocial stress and changes in cardiovascular parameters (for review
of a substantial body of evidence e.g. Schnall et al. 2000,[10] Stansfeld
& Marmot 2002[11]).
In contrast to Macleod we conclude that interventive and preventive
activities to reduce social inequalities in health should not be
restricted to established forms of behavioural and physiological factors.
Although knowledge on exposure and resources related to specific SES
groups needs to be improved psychosocial factors including job stress
should be part of preventive activities more often in the future.
Richard Peter
Johan Hallvqist
Johannes Siegrist
Töres Theorell
References
(1) Peter R, Siegrist J, Hallqvist J, Reuterwall C, Theorell T and the SHEEP Study Group. Psychosocial work environment and myocardial
infarction: improving risk estimation by combining two alternative job
stress models in the SHEEP Study. J Epidemiol Community Health 2002;56:294-300.
(2) Bosma H., Peter R., Siegrist J., Marmot M. Two alternative job stress
models and the risk of coronary heart disease. Am J Publ Health 1998;88:68-74.
(3) Theorell T, Tsutsumi A, Hallqvist J, Reuterwall C, Hogstedt C,
Fredlund P, Emlund M, Johnson JV, and SHEEP Study Group. Decision
latitude job strain, and myocardial infarction: a study of working men in
Stockholm. Am J Publ Health 1998;88:382-8.
(4) Lynch J, Krause N, Kaplan GA, Tuomilehto JT, Salonen JT.
Workplace conditions, socio-economic status, and the risk of mortalitiy
and acute myocardial infarction: The Kuopio Ischemic Heart Disease Risk
Factor Study. Am J Publ Health 1997;87: 617-22.
(5)Peter R & Siegrist J. Chronic work stress, sickness absence, and
hypertension in middle managers: general or specific sociological
explanations? Soc Sci Med 1997;45:1111-20.
(6) Peter R, Alfredsson L, Hammar N, Siegrist J, Theorell T,
Westerholm P. High effort, low reward and cardiovascular risk factors in
employed Swedish men and women: baseline results from the WOLF-Study. J
Epidemiol Community Health 1998;52:540-7.
(7) Orth-Gomer K, Eriksson L, Moser V et al. Lipid lowering through
work stress reduction. Int J Behav Med 1994;1: 204-14.
(8) Kristensen TS. Workplace intervention studies. In The workplace and cardiovascular diseases Schnall, K.
Belkic, P. Landsbergis P., D. Bakker (Eds). Occup Med State of the Art Reviews 2000;15:293-306.
(9) Theorell T, Emdad R, Arnetz B, Weingarten AM.
Employee effects of an educational program for managers at an insurance company. Psychosom Med
2001;63:724-33.
(10) Schnall P, Belkic K, Landsbergis P, Bakker D (Eds) The workplace
and cardiovascular disease. Occupat Med State of the Art Reviews 2000;15:1-374.
(11) Stansfeld S & Marmot M (Eds) Stress and the Heart. Psychosocial
Pathways to Coronary Heart Disease. 2002. London: BMJ Books.
Dear Editor
I thank Dr Adamiak for her careful reading of the glossary. I believe the confusion may be motivated by the fact that the sentence she refers to is not as clearly stated as it could have been. By "in the case of continuous dependent variables" I meant more precisely "when the response variable is normally distributed and the link function is identity (ie models usually referred to as linear models)". By "i...
Dear Editor
It would be interesting to know the overall health condition of the subjects. Were they average weight, average stressed occupations, diets. What other environmental conditions could have contributed to their heart problems, if any.
Dear Editor
The Authors draw conclusion that the use of acute hospital beds does not increase as the population ages, which is a result from a seven year cohort study in Germany. The problem is however that there are supply related factors, which can strongly affect the results. It is known that there are substitutions between different forms of care.[1] The days spent in hospitals and the mortality rates, in part...
Dear Editor
I have found that under the heading Population-average models (page 592), when comparing the multilevel models to population-average models, the Author is stating that in the case of continuous dependent variables the coefficients are mathematically equivalent in the marginal models. In the next phrase the Author suggest "...but in the case of non-normally distributed variables (for example, logistic...
Dear Editor
Professor Muntaner's cricism of the neglect of power relations in research on health inequality is well timed. The 2001 Census of England and Wales has used as its social classification a new measure: the National Statistics Socio-economic classification (NS-SEC). This measure has an explicit theoretical basis in the relations and conditions of employment. It therefore distinguishes those who own pr...
Dear Editor
Modern obstetritions in developed nations would do well to emulate Traditional Birth Attendents. These women are much less apt to cause iatrogenic injury, which is so often unknowingly done by OB/GYNs in modern actively managed births when the umbilical cord is clamped almost immediately. This prevents normal placental transfusion and causes hypoxia and hypovolemia in the newborn.
For more i...
Dear Editor
I think it is time to renew ideas at PAHO. Drs Roses and Casas had been high level officers for many years, why did not they implemented their suggestions? Dr Sepúlveda is a bright public health man from Mexico, with an oustanding curriculum. Lets work towards improving PAHO's activities with new people.
The discussion triggered by this paper is a useful one. In general I agree with McLeod, and Davey Smith that we have to be careful when we make conclusions regarding etiology. Socioeconomic factors could certainly confound the relationships. Adjustment for socioeconomic factors could also conceal true relationships, however. In this literature (see the review by Schnall et al. 2000 which has been referred...
Dear Editor
We are grateful to Peter and colleagues for taking the trouble to respond to our e-letter. They raise several issues that we will try to address briefly.
First, they say effort-reward imbalance had no relation to socio-economic status in their study. We are unclear how they assessed this – the only SES measure they mention in their paper is a white-collar/blue- collar binary distinction. Effort–r...
Dear Editor
We agree with John Macleod concerning the importance of a better understanding of social inequalities in health. Yet, the paper to which he refers (Peter et al., 2002) [1] was not primarily intended to explain social inequalities, but rather to demonstrate an improved prediction of CHD risk by combining standardized measures of two innovative theoretical concepts of psychosocial stress at work. M...
Pages