2.4 Secuencias de iniciación
2.4.1 Importancia de la secuencia de iniciación
The conceptual model developed by the External Assessment Group was relatively simplistic because of the lack of appropriate data. A decision tree approach was adopted with a lifetime horizon and discounting undertaken at 3.5% per annum.
The NICE diagnostic reference case141requests that cost-effectiveness is presented in terms of cost per
QALY gained. This has been adhered to although the authors highlight the considerable uncertainty in any estimate because of the lack of robust data on key components of the calculation.
The cost per QALY can be divided into the incremental costs incurred and the incremental QALYs gained. The incremental costs should consider:
l the cost of each test/comparator
l the net effect on hospital length of stay for both ICU and non-ICU patients, noting that rapid tests
could be detrimental to the patient as well as beneficial
l the net effect on the costs of antimicrobial treatment
l any net cost impact associated with the potential impact on antimicrobial resistance.
The incremental QALYs would ideally consider:
l the impact on sepsis-related mortality
l the impact of net effect on hospital length of stay for both ICU and non-ICU patients, noting that rapid
tests could be detrimental to the patient as well as beneficial
l any net QALY impact associated with the potential impact on antimicrobial resistance.
Although the costs of the tests and comparators can be estimated relatively well from current data, there are no conclusive data on any other parameter listed in the bullet points above that were identified in the
External Assessment Group’s review. Therefore, a scenario analysis was undertaken in which these values
Within the model it was assumed that the rapid identification of a pathogen could result in changes in four key outcomes: 30-day mortality rates; the length of stay in an ICU; the length of stay in the hospital; and the costs associated with antimicrobial treatment. Of these, changes in the mortality rate were assumed to affect QALYs only, with the remaining categories assumed to affect costs only. This is a simplification in that, for example, additional time in an ICU may be associated with slightly lower QALYs, but the impact of such omissions was assumed not to affect the overall conclusions. In all scenarios the potential impact of better antimicrobial stewardship in terms of drug resistance was not evaluated because of both the complexity of such a task and the absence of information on how the tests would reduce antimicrobial use.
It was assumed that negative tests would not impact on any of the four key outcomes. This assumption was supported by the clinical experts to the External Assessment Group. The decision to ignore negative tests was because of the potential fatal consequences if treatment was withdrawn from a patient with sepsis. Acknowledged reasons for a false-negative result include the test being unable to detect the
pathogen or the quantity of the pathogen being below the test’s limit of detection. Similarly, tests that
would be denoted as failures were assumed to have no impact on the four key outcomes. Both negative tests and failures would, however, be associated with the cost of the test.
A pictorial representation of the conceptual model is provided inFigure 11. The net cost impact and the
net QALY impact of rapid identification were used to estimate a cost per QALY gained ratio.
Rapid identification of a pathogen (costs) 30-day mortality (QALYs) ICU length of stay (costs) Hospital length of stay (costs) Antimicrobial treatment (costs) Aggregated result Change in costs Change in QALYs
From these values, the cost per QALY gained could be estimated
It was assumed that negative results from rapid identification would not impact on patient management
It was assumed that failures would not impact on patient management
Both would, however, be associated with costs
1. Base case 1: an analysis based on currently published evidence.
2. Base case 2: an analysis in which parameter values were populated by estimates from clinical experts in order to estimate the cost-effectiveness of each test. This has a benefit in that if, in base case 1, the rapid tests offered little or no benefit compared with the comparators, based on the absence, or lack of statistical significance, of the required data, then clinical beliefs could be incorporated.
3. Threshold analyses were undertaken to guide decision-makers on the likelihood of the interventions having cost per QALY gained values of £20,000 or lower and of £30,000 or lower, as it was assumed that experts in the field would be more confident in providing an indication of whether the value of a parameter was greater than, or less than, a threshold value than in estimating a value in the absence of data (as was requested in base case 2). The variables assessed within the threshold analysis in the threshold base case were the number of mortalities within 30 days that were prevented and the reduction in days in ICU. For simplicity, and to allow thresholds to be presented purely in terms of net 30-day mortality or net cost, it was assumed that no additional QALY gain was associated with a reduction in ICU duration of stay. The results are presented allowing for a mixture of both net
mortalities and of net reduced ICU stay. In an alternative analysis, the thresholds of both the reduction in the net number of ICU days and the net reduced costs of antimicrobial treatment are also presented. 4. Analyses comparing the interventions with MALDI-TOF MS based on published literature.
5. Analyses of data taken from studies in which more than one intervention were compared directly. Given the large divergence in results produced by base cases 1 and 2, the External Assessment Group decided that probabilistic sensitivity analyses would provide spurious accuracy with respect to the decision being undertaken. As such, only deterministic answers have been provided. If robust data are produced in relation to the efficacy of the interventions on key patient outcomes, then probabilistic sensitivity analyses should be conducted.
The lack of evidence for heterogeneous diagnostic accuracy among subgroups resulted in the External Assessment Group providing only an overall measurement of cost-effectiveness rather than by subgroup.
Although the cost-effectiveness may differ among subgroups–for example, a neonate would be expected
to accrue more QALYs than an adult–these do not affect the fundamental uncertainty of whether or not
the interventions would be associated with any key patient outcome.
For all but the threshold analyses, the incremental cost per test has been calculated accounting for the net effect on ICU and hospital length of stay, and changes in the costs of antimicrobial treatment. The rate of positivity for each test must also be known, as it has been assumed that only positive intervention tests would result in a change in management.
As an illustrative example, assuming that the cost of a test was £400, that each positive test was associated with a 0.1-day reduction in ICU length of stay, a 0.3-day reduction in hospital stay and a £50 reduction in antimicrobial treatment, and a 20% rate of positivity, the incremental costs would be estimated to be:
£400+(f(−0:1 × £1057)+½(−0.3− −0.1) × £275+−£50g× 20%)=£357.86. (11)
Note that these values are for illustrative purposes and are not necessarily those used in the modelling exercises that are detailed in later sections.
The incremental QALYs have been calculated assuming 11.32 discounted QALYs per 30-day mortality
avoided (seeThe quality-adjusted life-year gains associated with preventing a 30-day mortality). Thus, if an
intervention was assumed to reduce 0.01 deaths per test, then the discounted QALYs gained would be 0.1132. To calculate the numbers of deaths avoided, data are required on the assumed underlying mortality rate at 30 days, the estimated reduction in the rate of 30-day mortality associated with each test
and the rate of positivity. As an example, if it was assumed that the underlying mortality rate was 13%, the reduction following a positive test was 5% and the rate of positivity was 20%, then the estimated number of deaths prevented would be 13% × 5% × 20% per test, which equals 0.0013.
The principles outlined above in calculating incremental costs and QALYs have been maintained throughout the analyses undertaken in this report. For simplicity, the example provided above did not distinguish between the assumed impacts of the interventions when a subsequent blood culture was either
negative or positive although, as detailed inModel parameters assumed for base case 1, separate values
for these were provided by the clinical experts.