CARDIOVASCULAR JOURNAL OF AFRICA • Volume 25, No 1, January/February 2014
AFRICA
25
Overall, with limited available data, and by using proxies to
derive a subset of high-risk patients, we concluded that TAVI
is likely to be cost-effective versus cAVR. Using the most
conservative estimates, we predict a small number of patients
who could benefit from TAVI versus cAVR, with a lower
mortality rate at, on average, lower costs.
Discussion
In our sample of TAVI and cAVR hospital records from an
administrative database, we found that TAVI patients were on
average older, had reduced ICU and hospital LoS, non-statistically
significantly reduced mortality rates and higher costs than cAVR
patients. However, by trying to identify those cAVR subjects who
would most likely have been candidates for TAVI, we projected
that a small group would have clinically benefitted from TAVI
(lower mortality rate) at a reduced cost to the funder.
The sample’s mean TAVI costs were higher than those
of cAVR, but the latter group was more heterogeneous and
presumably included a wide range of patients in terms of
pre-operative severity levels, which means a direct comparison
between the two groups is difficult. In-hospital mortality rate
was lower in the TAVI group, which was not expected, as cAVR
has a much broader utility and would be expected to be used in
lower-risk patients.
The average of the upper 21st percentile cAVR cost distribution
was equal to the average TAVI cost, so from a purely economic
perspective, we could assign one patient out of five with a novel,
less-invasive treatment, ensuring faster recovery with a reduction
of over 40% in ICU and hospital LoS at no additional cost. This
substantial reduction was expected despite the difference in age
between the two groups and reflects the more rapid and typically
less-invasive nature of the procedure. It is also supported by
numerous clinical data, although it may be confounded by local
guidelines and operational practice, depending on the vagaries of
the healthcare system (e.g. reimbursement practices, discharge to
alternative local facilities).
When estimating the high-risk cAVR costs, defined by
patients staying more than six and 13 days in ICU and hospital,
respectively (as was found in the Partner Cohort A trial), and
being over 75 years of age, the estimate increased just above the
TAVI mean cost, and the in-hospital mortality rate increased to
21.4%, a four-fold increase compared with the TAVI group, and
a rate similar to that of the French MoH level 4 severity group.
Our findings are not dissimilar to others in the literature.
For example, Arnaoutakis
25
studied the relationship between
STS score and hospital charges. In their analysis the authors
showed that the median hospital charges for patients with risk
scores above and below 10% were US$ 88 241 and US$ 42 785,
respectively, a relative difference above 100%. In a multivariate
regression model they found that each 1% increase in STS risk
score was associated with an additional US$ 3 000. Additionally,
Toumpoulis
26
found a significant correlation between additive
EuroSCORE and hospital LoS, as to be expected, whereby a
higher-risk score leads to higher costs and extended LoS.
There were several limitations with this work, most notably,
the restricted clinical data available to us from an administrative
database. The task of making any comparison between patients
according to typical clinical criteria was therefore more difficult,
but also
post hoc
matching of records has its own limitations. Of
enormous benefit and in contrast to most economic analyses, we
had access to precise billing data. Typically, other analyses use
proxy cost data from other studies to estimate the total cost of
a procedure, but here we had the total cost from admission to
discharge, albeit without details of professional fees. It should
be noted that the omission of professional fees favours cAVR,
as these fees are likely to be higher than for TAVI. An additional
limitation is that we do not have any data on re-admission after
discharge. No differences are expected between the groups
but additional analyses would help to quantify this important
outcome.
The transferability of the conclusions from this study are also
limited, in common with those from any economic evaluation.
We acquired our data from a South African administrative
database and the clinical practice underpinning the costs derived
may differ from other settings. However, this also provided a
major strength of the study in that the data we had was derived
from South African clinical practice and the results were not
confounded by the use of unrelated data.
The adoption of any new technology is challenging and
especially if it is a ground-breaking intervention that challenges
treatment paradigms. However, the data on TAVI from South
Africa show that TAVI costs are much more predictable than
cAVR, which should greatly aid planning and implementation.
At best it would appear that TAVI could reduce costs to funders
and improve outcomes in appropriately selected individuals.
Conclusion
Within the context of limited clinical data, we performed the first
attempt at cost-effective analysis of TAVI versus cAVR in South
Africa. Treatment of aortic stenosis with cAVR in a
post hoc
defined high-risk patient segment is more expensive than TAVI
in South African centres. Despite common perceptions on costs,
adoption of TAVI as an alternative, less-invasive therapy that has
been proven clinically and recommended by an FDA advisory
panel (Partner A) to be at least as effective as cAVR has a viable
economic argument in appropriate patients.
We thank the MediClinic Southern Africa private hospital group and
particularly Mr Roly Buys, Ms Wanda de Beer and Mr Ryan Wedlake. Their
administrative database was valuable and they advised us wisely and spent a
consistent and generous amount of time making sure it was used accurately.
Without their help this research could never have been performed.
We also thank Hox-Com Analytic, particularly Mr Gilbert Caranhac,
who provided the French MoH PMSI database. His valuable time ensured
the accuracy of our findings on real-world cAVR clinical and economical
outcomes in France.
Table 3. Total cost and in-hospital mortality rate for
our three proxies
Proxy
Patients
Hospital
mortality
(%)
Total costs
(ZAR)
n
%
ICU LoS
>
6 days
47
19.7
19.1 320.2k
±
136.9k
Hospital LoS
>
13 days
85
35.6
9.4 276.9k
±
116.3k
Age
>
75 years
49
20.5
10.2 236.2k
±
88.7k
All proxies combined
14
5.9
21.4 337.9k
±
80.9k
