Page 186 - 16Neonatal Jaundice_compressed

Appendix C: Economic evaluation of testing strategies for hyperbilirubinaemia

Diagnostic tests are usually evaluated according to their sensitivity and specificity and these
characteristics can be used to generate probabilities in decision-analytic models. Initially, we
intended to compare the alternative strategies using such an approach. However, the decision-
making process in this context is far more complicated than that implied by the outcomes for a
‘two by two table’. Rather than the test result dividing the patient population neatly into
positives and negatives for hyperbilirubinaemia, ‘raised bilirubin level’ is measured on a
continuum from normal to severe disease and different test thresholds are used to stratify
patients into groups requiring immediate treatment, further monitoring or transfer back to
routine care. Decision-making is affected implicitly in a Bayesian manner by the impact of the
bilirubin level on the post-test probability of disease. The decision-making is complicated further
as a number of other factors, such as family history of jaundice, will also be taken into account.
Furthermore, monitoring can occur at many points in time and this temporal aspect is important
because thresholds for clinically significant jaundice change and the evidence base to track
changes in diagnostic accuracy over the relevant time periods is lacking. Therefore, it was
ultimately decided that there was not sufficient published evidence to populate such a decision
model. Furthermore, it was felt that the GDG would not be able to estimate the vast array of
model parameters to reflect the actual micro decision-making process that occurs in actual
clinical practice.
The GDG has set a higher bilirubin threshold as a basis for treatment and a lower bilirubin
threshold for further monitoring. The rationale for this is to avoid unnecessary phototherapy (i.e.
a high specificity or false positive rate in terms of treatment) while avoiding missed cases by
continued monitoring in babies who have an intermediate bilirubin level (i.e. a high sensitivity
or false negative rate in terms of monitoring). While the TCB is not thought to be reliable at high
bilirubin levels (hence the need for TSB if TCB is positive), it is nevertheless thought to be
accurate at the more intermediate levels.

The GDG opinion is that, using the thresholds defined in this guideline, either method of testing
would be effective in detecting hyperbilirubinaemia and avoiding new cases of kernicterus.
Therefore, the cost-effective strategy was estimated using a cost-minimisation approach that
assumes no difference in effectiveness between testing strategies. As noted earlier, there is
insufficient evidence to estimate the incremental benefit of moving from ‘current practice’ to a
more intensive testing regime, although evidence on the limitations of visual examination
suggests that some benefit is likely. Therefore, threshold analyses were undertaken to determine
the number of kernicterus cases that a more intensive testing approach would have to avert in
order for this to be considered cost-effective.

C.5 Model parameters and assumptions

The cost analysis was undertaken from the perspective of the NHS and personal social services,
which is in accordance with the NICE guidelines manual
(www.nice.org.uk/guidelinesmanual). 237 The costs were estimated using a bottom-up or
‘ingredients’ approach, which involves detailing the physical quantity of resources used in
providing treatment alongside the unit cost of those resources. From this it is possible to
estimate the total cost of treatment.
It was assumed that visual examination is undertaken in the first instance in all strategies. In the
‘current practice’ strategy, it was additionally assumed that visual examination is used to
determine the severity of hyperbilirubinaemia with a proportion of these having a TSB blood
test as a precursor to possible phototherapy. No cost has been applied to the visual examination
as it is assumed this would occur as part of the standard home visit carried out by a midwife or
as part of standard hospital care if the baby had not been discharged.
The population characteristics for this analysis are shown in Table C.1. Economic parameters
used in the assessment of cost benefit (other than the costs of the test strategies) are shown in
Table C.2. It should be noted that the base-case values for the cost and QALY loss of a
kernicterus case should not be considered point estimates in a conventional sense. Considerable
uncertainty surrounds both values and therefore, as part of sensitivity analysis, both are varied
simultaneously to derive a cost-effectiveness threshold across a wide combination of values.

155