CARDIOVASCULAR JOURNAL OF AFRICA • Volume 25, No 1, January/February 2014
22
AFRICA
savings that may be realised by lowered complication rates and
length of hospital stay. The available MediClinic administrative
database has limited clinical outcomes but sufficient information
to allow for the identification of high-risk cAVR patients who
potentially could have been TAVI candidates, and to compare
data with the patients who did undergo TAVI.
TAVI has been introduced into clinical practice and is
reimbursed in many European countries since its commercial
availability in 2007. In the US, FDA approval arrived in
November 2011 for inoperable patients and was overwhelmingly
(11-0) recommended by an FDA advisory panel in June 2012
for patients at high risk for conventional surgery. The aim of
this study was to compare outcomes between TAVI and cAVR in
South Africa in order to evaluate the costs and benefits of both
treatment options.
Methods
An initial dataset was obtained fromMediClinic, one of the largest
South African private hospital groups, and contained billing
records on 394 patients who had undergone conventional aortic
valve replacement (cAVR) during the period 2009 to 2011 at eight
cardiac hospitals, MediClinics Bloemfontein, Heart Hospital,
Morningside, Nelspruit, Panorama, Vereeniging, Verglegen and
Witwatersrand University Donald Gordon Medical Centre.
From procedural coding we were able to exclude all patients
who had undergone concomitant procedures with cAVR (e.g.
CABG, ascending aorta) in order to compare more appropriately
with TAVI patients who would not electively undergo these
additional procedures. This produced a final dataset of 239
isolated cAVR patients. Over the same period the records of 75
TAVI patients were also available.
The dataset included the total costs per patient to the
healthcare provider (insurer) without professional fees and no
breakdown of cost components was available. Professional
fees vary from centre to centre and in order to avoid adding
uncertainty to the analyses by including estimates, these were not
included. Overall, costs were standardised to 2011 South African
Rand (ZAR) using the South African consumer price index (CPI)
published on the governmental statistics department website.
12
For ease of interpretation and at the time of writing, 10 ZAR was
approximately equal to
€
1 or US$ 1.25.
The database did not provide clinical risk scores, so to
compare cAVR with TAVI we excluded the results of those
patients who would not be considered for TAVI. Age
≥
75 years
provided the single predictive variable and we used ICU and
hospital LoS as surrogates or proxies for indicators of ‘high-risk’
patients.
Statistical analysis
Quantitative continuous variables are described with means
±
standard deviation, and quantitative discrete variables with
absolutes and relatives frequencies. Inference statistics comparing
continuous variables were made using the
t
-test or Wilcoxon rank
sum test as appropriate. To compare discrete variables, Pearson’s
chi-squared test with Yate’s continuity correction or Fisher’s
Exact test (when count data
≤
5) were applied. Two-sided tests
were used and a type I error significance level of 0.05 was
considered. Distributions of quantitative continuous variables are
presented graphically with normalised histograms; the
y
-axis is
given with densities, ensuring the total area equals one.
When representing the distributions between groups, box-and-
whisker plots (boxplots) were chosen. Relationship and linear
correlation between quantitative continuous variables were
populated and tested with Pearson’s product moment correlation
coefficient. Linear models were fitted by ordinary least-square
regression estimates and by robust regression using an M
estimator (package MASS).
13
Coefficient estimates are given
with standard errors. All analyses were performed with the use
of R software, version 2.13.1.
14
Results
We obtained a total sample of 75 TAVI and 239 isolated cAVR
patients from the period 2009–2011. Descriptive and inference
statistics for both TAVI and cAVR groups are presented in Table
1. The mean age for each group was 79.4
±
7.3 versus 62.3
±
15.2 years for TAVI and cAVR, respectively. This difference
was highly statistically significant (
p
<
0.001). Male gender
was more frequent in the cAVR group (59.8%) but less in the
TAVI group (44.0%) and this difference was also statistically
significant (
p
=
0.023).
Due to limitations in the available clinical data, we were
unable to make direct comparisons between groups and it is
impossible to draw strong inferences. Surprisingly, in-hospital
mortality rates were numerically higher for cAVR than TAVI.
Out of the 75 TAVI patients, four (5.3%) died before discharge,
compared to 19 (7.9%) for cAVR, although this difference was
not statistically significant (
p
=
0.613).
Much less surprisingly, TAVI, a less-invasive procedure,
clearly demonstrated faster post-operative recovery. Indeed,
the ICU LoS and the hospital LoS were reduced on average by
47.1% (2.7
±
2.8 vs 5.1
±
6.1 days,
p
<
0.001) and 44.1% (7.6
±
4.9 vs 13.6
±
9.2 days,
p
<
0.001), respectively. These reductions
were also robust and were not impacted on by outliers; indeed,
when considering the median ICU and hospital LoS, the
corresponding reductions were 59.1% (2.0 vs 3.5) and 40.9%
(6.5 vs 11.0), respectively. From an absolute perspective, the
number of ICU days saved per patient with TAVI was between
1.5 (from medians) and 2.4 (from sample means). Similarly, we
could expect a reduction in per-patient hospital LoS of between
4.5 (from medians) and six days (from sample means).
The overall average cost per patient to the healthcare
provider was ZAR 335.5k
±
47.9k for TAVI and ZAR 213.9k
±
87.4k for cAVR (
p
<
0.001). As we can see from Fig. 1, the
cAVR distribution is highly skewed, indicating a presumably
heterogeneous population with a wide range of what may be
low- to high-risk patients (certainly heterogeneous post-operative
outcomes). This seems not to be the case in the TAVI group
Table 1. Descriptive table for tavi and cavr groups
Variable
TAVI (
n
=
75) cAVR (
n
=
239)
p
-value
Age (years)
79.4
±
7.3
62.3
±
15.2
<
0.001
Male sex,
n
(%)
33 (44.0)
143 (59.8)
0.023
ICU LoS (days)
2.7
±
2.8
5.1
±
6.1
<
0.001
Hospital LoS (days)
7.6
±
4.9
13.6
±
9.2
<
0.001
Total costs (ZAR)
335.5k
±
47.9k 213.9k
±
87.4k
<
0.001
In-hospital mortality,
n
(%)
4 (5.3)
19 (7.9)
0.613
