3.3.5 Health-Related Quality of Life
HR-QOL is assessed using both generic and COPD-specific measures. Health status is represented by a specific measure of HR-QOL: the SGRQ. The SGRQ is designed to measure health impairment in patients with asthma and COPD . From a clinical point of view, a significant change in health status can lead to treatment switching. The user can therefore select a change in SGRQ as a criterion for switching in the Control sheet. It was assumed that a worsening of four units in SGRQ (i.e. ?4, since the higher the SGRQ, the worse the health status) can trigger a change in patient medication. SGRQ scores increase with time and this was reflected in the current assumptions.
The model is also designed to be populated with generic measures of HR-QOL such as utilities based on the EQ-5D or the SF-6D. Baseline utilities associated with each disease severity state are applied at each cycle. Whenever a patient experiences an exacerbation, a utility decrement is subtracted from the baseline utility value corresponding to the cycle where the event had occurred. An algorithm that predicts EQ-5D utility from the SGRQ was recently developed to allow the use of utilities from the SGRQ.
3.3.6 Costs
Costs of drugs, exacerbations and maintenance are set by the user according to the setting of interest. In each cycle, the cost of maintenance is determined by disease severity. Exacerbation costs are added if a patient experiences an exacerbation episode within that cycle. It is assumed that there are no other costs associated with symptoms occurring to patients. The model adopts the healthcare payer’s perspective, therefore indirect costs are not considered.
3.4 Model Outputs
The model produces three sets of outputs: clinical, cost and cost effectiveness. These will help to inform decisions regarding the adoption of competing interventions by healthcare payers. Although these decisions are not based on cost-effectiveness criteria only, an intervention is considered to be cost effective when the incremental costeffectiveness ratio is below a predefined accepted threshold representing the willingness to pay by the healthcare payer. This threshold is usually defined as a cost per measure of effectiveness. The decision rules for economic evaluation are described elsewhere . A summary of model inputs and outputs is available in Table .
3.4.1 Uncertainty
In order to characterize parameter uncertainty (or secondorder uncertainty), probability distributions were fitted to some of the model’s inputs. Gamma distributions were fitted to treatment efficacy (initial FEV1 boost) and to cost parameters; a log-Normal distribution was fitted to exacerbation parameters; and, finally, Beta distributions were fitted to utilities, withdrawals and mortality. The merits of assigning these distributions to some of the model’s parameters are described elsewhere . Ninety-five percent confidence intervals (CIs) are presented around the results to represent this uncertainty. This also allows for the development of cost-effectiveness acceptability curves (CEAC) but the current software format emphasizes transparency rather than speed of computations; therefore, the inclusion of a sufficient number of iterations would require a correspondingly large amount of time.
3.4.2 Model Validation
The model was subjected to a comprehensive validation process. Firstly, it was tested to ensure that all mathematical calculations were accurate and consistent with its specifications. Secondly, the model was tested by being populated with four datasets, three of which were from clinical trials and one a natural history of disease study . Our aim was to assess if the model was able to replicate the results of these studies with an acceptable degree of accuracy, ensuring internal and external validity. Table shows
the inputs extracted from the four published studies used to assess the model validity.
The parameters tested varied by the availability of information in each publication, but focused on clinical
parameters such as mortality and exacerbation rates, as well as FEV1 decline. The model benefited from the clinical and methodological scrutiny of key opinion leaders in order to ensure not only that the model was consistent and coherent with the clinical pathway of COPD but also to certify that the methods used were robust. For the validation exercise, the micro-simulation model was run for 500 patients and 50 cohorts.