Decision-analytic evaluation of the clinical effectiveness and cost-effectiveness of management programmes in chronic heart failure.
ABSTRACT While management programmes (MPs) for chronic heart failure (CHF) are clinically effective, their cost-effectiveness remains uncertain. Thus, this study sought to determine the cost-effectiveness of MPs.
We developed a Markov model to estimate life expectancy, quality-adjusted life expectancy, lifetime costs, and the incremental cost-effectiveness of MPs as compared to standard care. Standard care was defined by the EuroHeart Failure Survey for Germany, MP efficacy was derived from our recent meta-analysis and cost estimates were based on the German healthcare system. For a population with a mean age 67 years (35% female) at onset of CHF, our model predicted an average quality-adjusted life expectancy of 2.64 years for standard care and 2.83 years for MP. MP yielded additional lifetime costs of euro1700 resulting in an incremental cost-utility ratio (ICUR) of euro8900 (95% CI: dominant to 177,100) per quality-adjusted life year (QALY) gained. Sensitivity analyses demonstrated that the ICUR was sensitive to age and sex.
MPs increase life expectancy in patients with CHF by an average of 84 days and increase lifetime cost of care by approximately euro1700. MPs improve outcomes in a cost-effective manner, although they are not cost-saving on a lifetime horizon.
- [show abstract] [hide abstract]
ABSTRACT: Measurement of health care costs is a crucial task in health economic evaluation. Various guidelines with different amount of details have been set up for costing methods in economic evaluation which, however, do not precisely stipulate how to value resource consumption. In this article we present a proposal for the standardisation of the monetary valuation of health care utilisation occurring in the follow up period after the actual intervention to be evaluated. From a societal perspective the primary direct and indirect cost components are considered, such as outpatient medical care, pharmaceuticals, non-physician health services, inpatient care, days of sick leave and early retirement due to sickness. The standard costs are based on administrative charges and rates or on official statistics. They are based on the most current data sources which are mainly from 2002 and 2003. This system of standard costs aims at an average valuation of resource consumption. This makes for the comparability of different health economic studies. Most standard costs are not based on market prices but on administratively specified charges and rates. This implies that institutional changes which are quite common in the health care system, may also affect the valuation rates, for example the introduction of DRGs. This should be taken into account when updating the system of standard costs.Das Gesundheitswesen 11/2005; 67(10):736-46. · 0.94 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Heart failure is a common and serious condition requiring extensive health care resources. The aim of this study is to estimate the total treatment costs of heart failure in Sweden. The study is a prevalence-based cost-of-illness study. It includes costs of institutional care (hospitals and nursing homes), outpatient care, surgery and drugs. The costs are estimated based on official Swedish statistics, and on various clinical and epidemiological studies. The results are expressed in 1996 prices. The total annual treatment costs for heart failure are approximately Swedish kronor (SEK) 2000-2600 million, or nearly 2% of the Swedish health care budget. Institutional care is the single largest component, amounting to SEK 1300-1900 million, or about 65-75% of the costs of heart failure treatment. The results from this study indicate that heart failure is a costly condition. Efforts to develop effective management programmes that can reduce the need for expensive institutional care, without a negative impact on quality of life, morbidity and mortality, should be given high priority.Journal of Internal Medicine 10/1999; 246(3):275-84. · 6.46 Impact Factor
- Journal of The American College of Cardiology - J AMER COLL CARDIOL. 01/2004; 44(4):810-819.
Decision-analytic evaluation of the clinical effectiveness and
cost-effectiveness of management programmes in chronic heart failure
Alexander Göhlera,b,c, Annette Conrads-Franka,c, Stewart S. Worrella, Benjamin P. Geislera,c,
Elkan F. Halperna, Rainer Dietzb, Stefan D. Ankerb, G. Scott Gazellea,d, Uwe Sieberta,c,d,⁎
aInstitute for Technology Assessment and Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, USA
bDivision of Cardiology, Charité Campus Virchow-Klinikum, Berlin, Germany
cDepartment for Public Health, Medical Decision Making and Health Technology Assessment, University for Health Science,
Medical Informatics and Technology, Hall i.T., Austria
dDepartment of Health Policy and Management, Harvard School of Public Health, Boston, USA
Received 27 October 2007; received in revised form 30 April 2008; accepted 24 July 2008
Available online 28 August 2008
Background and aims: While management programmes (MPs) for chronic heart failure (CHF) are clinically effective, their cost-effectiveness
remains uncertain. Thus, this study sought to determine the cost-effectiveness of MPs.
Methods and results: We developed a Markov model to estimate life expectancy, quality-adjusted life expectancy, lifetime costs, and the
incremental cost-effectiveness of MPs as compared to standard care. Standard care was defined by the EuroHeart Failure Survey for
Germany, MP efficacy was derived from our recent meta-analysis and cost estimates were based on the German healthcare system. For a
population with a mean age 67 years (35% female) at onset of CHF, our model predicted an average quality-adjusted life expectancy of
2.64 years for standard care and 2.83 years for MP. MP yielded additional lifetime costs of €1700 resulting in an incremental cost-utility ratio
(ICUR) of €8900 (95% CI: dominant to 177,100) per quality-adjusted life year (QALY) gained. Sensitivity analyses demonstrated that the
ICUR was sensitive to age and sex.
Conclusion: MPs increase life expectancy in patients with CHF by an average of 84 days and increase lifetime cost of care by approximately
€1700. MPs improve outcomes in a cost-effective manner, although they are not cost-saving on a lifetime horizon.
© 2008 European Society of Cardiology. Published by Elsevier B.V. All rights reserved.
Keywords: Hearth failure; Management programme; Cost-effectiveness analysis; Markov model
Despite therapeutic advances, chronic heart failure (CHF)
remains a medical condition of increasing importance in
public health. CHF causes or complicates approximately
20% of all hospitalisations in people in the developed world
older than 60 years of age . With a crude 1-year mortality
rate of 33% to 61% in high risk patients , CHF has a
prognosis similar to many cancers. Moreover, treatment
costs for CHF in Europe and the United States account for
1–2% of total health care expenditures  with hospitalisa-
tion costs accounting for two-thirds of this amount .
In the hope of improving CHF outcomes, management
programmes (MPs) have been developed to standardize and
optimize CHF treatment. These programmes focus on
disease education for patients and continuing support after
hospital discharge. Recent meta-analyses have demonstrated
that MPs for CHF are clinically effective; however, the cost-
effectiveness of MPs remains uncertain .
European Journal of Heart Failure 10 (2008) 1026–1032
⁎Corresponding author. Cardiovascular Research Program, MGH Insti-
tute for Technology Assessment, Harvard Medical School, 101 Merrimac
Street, 10th Floor, Boston, MA 02114-4724, USA. Tel.: +1 617 724 4445;
fax: +1 617 726 9414.
E-mail address: email@example.com (U. Siebert).
1388-9842/$ - see front matter © 2008 European Society of Cardiology. Published by Elsevier B.V. All rights reserved.
by guest on June 9, 2013
The objective of this study was to assess the long-term
clinical and economic consequences of MPs in the treatment
of CHF patients in Germany.
2.1. Description of the Markov model
We developed a five-state Markov model that is identical
in structure for both strategies. The model simulation begins
at the time when a patient hospitalised for CHF is discharged
alive and moves through subsequent rehospitalisation states
(Fig. 1). The number of rehospitalisations is a valid and
generally accepted proxy for disease progression in patients
with CHF . Since an individual's number of hospitalisa-
tions prior to trial enrolment is usually not known to the
investigator, we defined the hospitalisation at which the
patient was enrolled into the management programme as the
index hospitalisation and counted the number of hospitalisa-
During each one month cycle, patients may remain in the
current hospitalisation state, experience a rehospitalisation or
die from cardiovascular or non-cardiovascular causes. The
simulation was carried out until all patients who had not
already died reached an age of 120 years.
2.2. Risks of hospitalisation and death
We utilized Weibull regression and logistic regression
models  to derive transition probabilities that allowed us
to simulate the natural history of chronic heart failure from
original patient-level data obtained in the Beta-Blocker
Evaluation of Survival Trial (BEST) . Briefly, this multi-
centre trial showed no overall survival benefit in 2708
patients with New York Heart Association (NYHA) class III
or IV heart failure, who were randomly assigned to
bucindolol or placebo.
The regression analyses allowed us to derive input param-
eters for the Markov model that are not available in the
published literature, and thus to design the model in order to
reflect the health care context of MP.
Probabilities of all-cause rehospitalisation and CV mor-
tality (without hospitalisation within the previous 30 days)
were estimated as a function of age, sex, number of previous
hospitalisations, and time since last hospital discharge. Cor-
relation among those parameters was accounted for using the
Cholesky decomposition method .
Hospitalised subjects were assigned an increased 30 day-
mortality risk as a function of their age, sex and number of
previous hospitalisations. We thought a time period of
increased mortality was superior to differentiating between
in-hospital mortality and post-discharge mortality, since
hospitals may use various admission and discharge criteria in
CHF patients. The probability of dying from non-cardiovas-
cular causes without prior hospitalisation was based on
German life tables and only depended on age and sex .
Monthly transition probabilities for the base-case cohort are
given in Table 1. For further details on the underlying
statistical methods that were used to derive the input param-
eters, please refer to our accompanying online material (see
2.3. Efficacy data
MP efficacy estimates were based on the meta-analysis
that we performed using 36 randomised controlled studies
from 13 countries with data from a total of 8341 patients. We
performed an extensive literature search to identify studies
that address ‘chronic heart failure’ and ‘disease management
programme’ and reported all-cause mortality or all-cause
rehospitalisation rates. Follow-up of the individual studies
varied between 3 and 18 months with a median follow-up
time across studies of 9 months. Detailed methods are
described elsewhere .
Our meta-analysis yielded a pooled relative risk of 0.81
(95% CI 0.70 to 0.93) for all-cause mortality and a pooled
relative risk of 0.84 (95% CI 0.77 to 0.92) for all-cause
hospitalisation, both favouring MPs over standard care. The
effect of age and sex on the overall effect of MPs was based
on the results of our meta-regression analysis . The
beneficial effect of MPs decreased relatively by 1% for each
1 year increase in age (4% to 0%) in the case of mortality and
by 3% (95% CI 8% to −1%) in the case of rehospitalisation.
Fig. 1. Schematic diagram of Markov model.
Transition probabilities and utilities by rehospitalisation stratum
to die (in
to die (out-of
Index hospitalisation 0.067
1027 A. Göhler et al. / European Journal of Heart Failure 10 (2008) 1026–1032
by guest on June 9, 2013
For men, MP was 92% (95% CI 297% to −92%) as effective
with regard to mortality-reduction and 81% (95% CI 549%
to −49%) as effective with regard to reduction in rehos-
pitalisation when compared to women of the same age.
In the model, MP was assumed to influence the rate of the
first rehospitalisation, all-cause mortality, and length of hos-
Given the difficulties in comparing interventions, cost-
utility analyses were performed. Utilities by rehospitalisation
stratum were derived from a subsample (1628 patients) of the
Eplerenone Post-AMI CHF Efficacy and Survival Study
(EPHESUS) and the utilities were then assessed using the
EQ-5D questionnaire based on German preferences  (see
2.5. Resource use and cost data
Costs of hospitalisations, physician visits, medications,
and MP implementation were reported in 2007 Euros. The
different cost components varied between the treatment
groups but were not affected by the specific Markov state.
Since MP reported an effect on all-cause hospitalisation
rather than CHF-specific hospitalisation, we estimated the
average cost of a hospital day using information from the
BEST trial and the German G-DRG grouper 2006 . We
calculated the average cost per hospital day for each disease
category from the German G-DRG grouper 2006 using linear
regression and translated the value into 2007 Euros. The
mean cost for a hospital day was estimated by taking a
weighted average of all disease categories in which the
weight of each category was based on the disease category
distribution for the first hospitalisation in the BEST trial. The
mean cost per hospital day was €560 (SE 41.6) and the
average length of hospital stay was 7.8 (SE 1.1) days. During
the simulation, costs for hospitalisations were event costs,
i.e., they were only accounted for when a patient actually
experienced a hospitalisation.
Outpatient health resource utilization frequencies were
derived from a survey of cardiologists and general practi-
tioners in Berlin. A detailed questionnaire regarding
healthcare utilization by CHF patients was mailed to 220
physicians, most of whom were based in outpatient clinics.
These physicians were asked to estimate the frequency of
utilization for common outpatient procedures used in the
care of CHF patients. Costs were then derived from the
German healthcare reimbursement database ‘EBM’. In the
standard care group, monthly cost parameters for general
practitioners were €69 (range: €42 to €96) and for cardio-
logists were €114 (range: €86 to €142).
The prescription frequency for cardiovascular medication
in Germany was obtained from the German section of the
EuroHeart Failure Survey . Drug reimbursement prices
were derived from the German 2006 Red List prices  and
the estimated average costs were €40/month/patient.
Initiation costs for MP were derived from the published
literature [14,15] and translated into 2007 Euros. Mean costs
were found to be €420 (range: €210 to €630). Finally, MP
increased outpatient costs by 5% and decreased the hospital
length of stay by 7%.
2.6. Cost-effectiveness analysis
Analyses were performed from a societal perspective
using values based on the recommendations of the German
Working Group on Methods in Health Economic Evalua-
tions . Costs and effects were both discounted at 5%/
annum according to current German guidelines .
Fig. 2. Validation plot of 1-year all-cause mortality as observed in the control arms of the studies included in our meta-analysis (boxes with error bars for 95%
confidence interval)  vs. model prediction (circles).
1028A. Göhler et al. / European Journal of Heart Failure 10 (2008) 1026–1032
by guest on June 9, 2013
2.7. Base-case and sensitivity analyses
For the base case we used data as reported by Germany in
the EuroHeart Failure Survey . The cohort thus had a
mean age of 67 years and comprised 35% women. We
assumed that the benefit of MP remained for 9 months, the
mean follow-up time in our meta-analysis , and then
declined linearly to zero over the next 12 months. We also
estimated incremental cost-utility ratios (ICUR) given
reasonable boundaries for the duration of treatment effects.
The first was that the effect of MP lasts for 9 months and then
immediately drops to zero and the second was that effects
observed in the meta-analysis continue for 21 months. The
effect of MPs on costs remained according to the time period
of the benefit.
Furthermore, we investigated the effect of management
programmes on not only the first but also all consecutive
hospitalisations. In addition, we also assessed the impact of
age and sex by using the results from our meta-regression
analysis . To examine the robustness of the results, we
performed 2 types of sensitivity analyses. In multiple one-
way sensitivity analyses, the following input parameters
were both halved and doubled: costs of hospitalisation,
outpatient costs, MP implementation and maintenance costs,
in- and out-of hospital mortality rate, risk of rehospitalisation
and discount rate. In probabilistic sensitivity analysis, we
accounted for the overall uncertainty in the estimated costs
and effects of each different strategy by using a Monte Carlo
simulation with 25,000 iterations .
3.1. External validation
We externally validated our model using 1-year mortality
data from the control arm of those MP studies included in our
meta-analysis. The input parameters for the model were
selected to reflect the age and sex observed in the studies.
Fig. 2 shows the all-cause mortality at 1 year as observed
in the six studies separately and as predicted by our model.
The model's prediction lies within the 95% confidence
interval of the observed 1-year mortality for each of the six
3.2. Base-case analysis
The expected quality-adjusted life years, life years, and
costs for patients who receive standard care and management
programmes are shown in Table 2. Results are tabulated for
the base case. Mean quality-adjusted life expectancy was
estimated to be 2.64 for standard care and 2.83 for MP,
resulting in a gain of 0.19 years for MP. Mean life expec-
tancy was estimated to be 3.31 years for standard care and
3.54 years for MP, thus yielding a gain of 0.23 life years for
MP. This gain in quality-adjusted life expectancy comes
at the additional cost of €1700, yielding a discounted incre-
mental cost-utility ratio (ICUR) of €8900 (95% Credibility
interval [CI] dominant to 177,100) per quality-adjusted life
year gained. The incremental cost-effectiveness ratio was
Results base-case analysis (values are discounted at 5%/year)
RegimeLife years QALYsCosts per patient (€)
Mean DifferenceMean DifferenceMean DifferenceICER (€/LYG)ICUR (€/QALY)
29,4000.23 0.1917007400 8900
Annotation: QALY: quality-adjusted life year; MP: management programme; ICER: incremental cost-effectiveness ratio; ICUR: incremental cost-utility ratio.
Fig. 3. Cost-effectiveness acceptability curve for base-case scenario.
1029A. Göhler et al. / European Journal of Heart Failure 10 (2008) 1026–1032
by guest on June 9, 2013
€7400 (95% CI dominant to 93,000) per life year gained.
Fig. 3 shows the cost-effectiveness acceptability curves for
the base-case scenario. The graphs represent the probability
that management programmes are cost-effective when
compared to standard care, given a particular willingness-
When accounting for the effect of MPs on subsequent
hospitalisations the ICUR dropped to €4600/QALY (95%
CI dominant to 59,400).
Deterministic sensitivity analyses of duration of treatment
benefit, costs of hospitalisation, outpatient cost, MP costs,
in- and out-of-hospital mortality rates, risk of hospitalisation,
and discount rate indicated that the results were insensitive to
changes in those parameters as the resulting values for the
ICUR differed by less than 10%.
3.3. Effect of age and sex
Table 3 relates the effects of age and sex to the overall
effect of MPs based on the results of 25,000 simulations. We
report the values of the willingness to pay (WTP) at which
MP was cost-effective when compared to standard care in
50% of the simulations as well as at 2.5% and 97.5% of the
simulations and we then used these values to derive the
appropriate credibility interval. Given the nature of MPs, we
assumed that they could not increase mortality rates.
However, it is possible that MPs could be only equally as
effective as standard care but increase rehospitalisation rates,
resulting in additional costs and thus be considered inferior
to standard care. Since the results of the simulations did not
fall into the third quadrant of the incremental cost-
effectiveness plane (i.e., they were not less effective and
less costly), the 50% threshold marks the point at which a
MP is as likely to be cost-effective as not when compared to
standard care, and thus the threshold at which society should
implement MPs if their WTP is more than the threshold
was €8100/QALY gained, €13,100/QALY gained, and
€24,000/QALY gained, respectively. For women in the
same age groups, the thresholds were €3400/QALY gained,
€3700/QALY gained, and €4000/QALY gained.
Based on our decision analysis, management programmes
for care of patients with chronic heart failure are likely to be
cost-effective. The base-case incremental cost-utility ratio
was €8900/quality-adjusted life year gained. Sensitivity
analyses demonstrated that the ICUR was sensitive to age
and sex but insensitive to cost of care for CHF, the duration
of the MP effect and our modelling assumption regarding
Assumptions about the duration of a MP effect after the
median follow-up time of 9 months had only little impact on
the overall ICUR. This seems to be a reasonable estimate
given that most first rehospitalisations occur within the first
year of follow-up. The result differed substantially when we
assumed that MPs not only affect the first rehospitalisation
but also reduce subsequent rehospitalisations. Five of the 36
studies in our meta-analysis  reported the effect of a MP
on subsequent rehospitalisations and demonstrated an even
stronger effect on subsequent rehospitalisations in favour of
MPs over standard care. For the base-case analysis, we felt
that ignoring this benefit of management programmes on
subsequent rehospitalisations would bias the analysis away
from MPs and consequentially would be a conservative and
According to our sensitivity analyses, age and sex of the
patient population strongly influenced the ICUR. The
threshold (i.e., median WTP) at which the MP was as likely
to be cost-effective as not when compared to standard care
increased with age and was about twice to six times as high
for men as for women. The correlation between an increase
in the threshold and increasing age seems clinically
plausible, since older patients may profit less from the
long-term effects of MP. The more favourable result for
women as compared to men is surprising as the target
population has already developed CHF and thus the
subsequent risks would be expected to be similar for both
sexes. We believe that this observation may result from
multiple factors. One possible explanation might be that
women are more responsive to this type of intervention as
observed for example in the German Interdisciplinary
Network for Heart Failure (INH) study (personal commu-
nication Dr. Angermann on April 3, 2008) . However,
the effect might also be a consequence of a bias due to higher
disease severity resulting from co-morbidities such as renal
failure among women as compared to men. While the
adjustment for NYHA class at baseline did not change the
favourable efficacy estimate for women as observed in the
meta-regression analysis , we did not have sufficient data
to adjust for further co-morbidities. The fact that the
credibility intervals for all age and sex subgroups overlap
indicates substantial uncertainty around these estimates.
Future studies should be designed in order to investigate the
effect of age and sex on the efficacy of MP and thus reduce
In contrast to specific interventions in cardiology, MPs
assume a more holistic approach, affecting all-cause
mortality and all-cause rehospitalisation rather than only
addressing specific cardiovascular events. An important
difference between previous models [18,19] and ours is that
Cost-utility analyses by age and sex
65 years 13,100
75 years 24,000
8100 (Dominant;17,000) 3400
1030A. Göhler et al. / European Journal of Heart Failure 10 (2008) 1026–1032
by guest on June 9, 2013