CAPÍTULO 3. VALIDACIÓN DEL PROCEDIMIENTO PROPUESTO MEDIANTE SU
3.3 Validación del procedimiento mediante el juicio de expertos
implementing a Fully Remote Online Pathway for
Chlamydia
The methods finally selected reflect the stage of technology development, and specific methodological issues identified in the consideration of the costs and benefits of asymptomatic chlamydia testing and treatment. There are two linked pieces of research presented in this thesis, firstly a preliminary evaluation of the costs and outcomes for the delivery of an OCCP which is presented in Chapter 7.
This included some data from an exploratory study undertaken in London by the eSTI2 consortium. Secondly, Chapter 8 outlines an
economic model for a fully remote online pathway (Pathway E in figure 2.1) to explore the costs and consequences of the introduction of this. The model considered the longer-term impact of the health complications associated with untreated chlamydia.
Methods for costing health services and their strengths and weaknesses are well documented in the literature (Drummond et al., 2015, Gray et al., 2011, Mogyorosy and Smith, 2005). There is a consensus that there are three distinct steps within the process of costing– identification, measurement and valuation (Drummond et al., 2015, Gray et al., 2011). However, Mogyorosy & Smith suggest two further steps prior to commencing costing that have a critical role to play in setting the perspective of the costing and framing the service to enable the accurate identification of costs. Their five step approach is summarised in figure 3.5 and further information on the detail of each stage is outlined in 7.2.1.
Figure 3.5 - Process of Costing, adapted from Mogyorosy & Smith, 2005
A primary costing study was undertaken to enable the calculation of costs to be included in the economic model and to ensure that costs are reflective of current service delivery models. Drummond and colleagues identify a scale of precision to healthcare costing from average daily cost to micro-costing (Drummond et al., 2005), whilst Brouwer and colleagues recognise that in practice the majority of economic evaluations use a combination of approaches from this scale (Brouwer et al., 2001).
Definition of decision problem and objectives of costing Description of service being
costed
Identification of resource items and units of measure
Measurement of resource use
Assignment monetary value to resources
A pragmatic approach was adopted to costing the OCCP and comparator pathways, in line with the type of economic evaluation for stage II of technology development identified by Sculpher and colleagues. They recognise that in respect of the identification of costs there is greater access to “individual patient data on the costs and outcomes of the new technology” however that “stage II estimates of cost-effectiveness are unlikely to be definitive” (Sculpher et al., 1997:27). Further information on the detail of the costing methods used is contained in section 7.2.
Careful consideration was given to the selection of the economic model used to assess the likely costs and benefits of a fully integrated online testing and treatment service for chlamydia. The use of static versus dynamic models in considering the cost- effectiveness of interventions for infectious diseases is widely debated. A literature review was undertaken, and the result presented in section 7.4.4. This identified that, for evaluating the cost-effectiveness of chlamydia testing and treatment, both types of models are in us: static models (Althaus et al., 2014, Hislop et al., 2010, Turner et al., 2011, Turner et al., 2014) and dynamic models (Adams et al., 2007, Looker et al., 2015, Low et al., 2007). A separate published literature review exploring the cost- effectiveness of chlamydia screening identified that six out of ten studies used dynamic models and four used static models (ECDC, 2014).
Economic models can be broadly categorised as illustrated in figure 3.6. Brennan and colleagues argue that it is the responsibility of the model developer rather than the policy maker to determine the type of model being used, and “advocate the use of simple models that still accurately reflect disease progression and health care delivery to the extent needed by a given decision problem”
This view is supported by Pitman and colleagues who note that dynamic models are important when the intervention impacts on the transmission of the disease, however they suggest that static models are acceptable where “their projections suggest that an intervention is cost-effective and dynamic effects would further enhance this” (Pitman et al., 2012:829), proposing the use of dynamic modelling to supplement static models which indicate borderline cost-effectiveness (ibid.).
Figure 3.6 - Taxonomy of Model Structures, Source: Brennan et al., 2006:1297
A B C D
Cohort/ aggregate level / counts Individual Level Expected value, continuous state, deterministic Markovian, discrete state, stochastic Markovian, discrete state, individuals Non- Markovian, discrete state, individuals 1 No Interaction Allowed
Untimed Decision tree rollback Simulated decision tree Individual sampling model (ISM); simulated patient-level decision tree (SPLDT) 2 Timed Markov model (evaluated deterministically) Simulated Markov model
ISM: Simulated patient level Markov model (SPLMM) (variations as in quadrant below for patient level models with interaction) 3 Interaction Allowed Discrete Time System dynamics (finite difference equations) Discrete time Markov chain model Discrete time individual event history model Discrete individual simulation 4 Continuous Time System dynamics (ordinary differential equations) Continuous time Markov chain model Continuous time individual event history model Discrete event simulation
Roberts concluded that in order to avoid misleading results, transmission dynamic models should be used to evaluate the cost- effectiveness of chlamydia screening programmes. To illustrate the difference between the static and dynamic modelling approaches, Roberts created three models, two static and one transmission dynamic to compare the cost-effectiveness of non-selective proactive screening, with no organised screening (Roberts, 2008). This identified base case results from the three models as follows: Static 1 - £8,474 per major outcome averted (MOA), Static 2 - £13,344 per MOA and Dynamic - £19,300 per MOA. Roberts explains the difference in results as being attributable to the comparator option not being the same – the static models had a comparator of no screening, whereas the dynamic model assumed a background level of opportunistic screening, plus the difference in approach to the application of discounting between static and dynamic models (ibid.).
Whilst it is recognised that dynamic models are superior to static models for modelling infectious diseases (Barton et al., 2004, Roberts, 2008) it also recognised that they are more complex, costlier and time consuming to develop. In the present study, a decision analytic model was selected for the following reasons:
• This is an early stage evaluation of a new technology and therefore the objective is to demonstrate the likely impact on costs and outcomes to inform future research and development
• Data for parametrising a model about the new technology are somewhat limited, with no data on self-testing (Stages 1 and 2 in pathway E, figure 2.1) and some initial exploratory study results and costings for OCCP (Stages 3 to 5 in pathway E, figure 2.1).
• It is not yet known how the availability of self-tests may influence sexual behaviours, risk taking and testing patterns, all of which would be material considerations within a dynamic model to inform parameters such as partner change rate.
In terms of quantifying benefits, Drummond and colleagues categorise outcomes into intermediate and final, with intermediate outcomes representing a measure which indicates a change in health outcome e.g. improvement in CD4 count (a measure of how well the immune system is working in patients with HIV) versus a final outcome e.g. survival or health related quality of life (Drummond et al., 2015). Process measures may also be an important consideration in the evaluation of a complex intervention (Moore et al., 2015). Process measures typically include aspects in relation to “service organisation, delivery and use” (Bowling, 2009:11). Use, i.e. testing and treatment uptake, are key parameters impacting both costs and outcomes within the economic evaluation undertaken in this thesis.
The outcome metric for economic evaluation preferred by NICE is the quality adjusted life year (QALY) as it a common unit of measure which enables comparison of outcomes between different health care interventions (Gray et al., 2011). QALYs are calculated by multiplying the health state outcome (represented on a scale of 1 for perfect health to 0 which is death) by the time spent in that state (Drummond et al., 2015). The difference in QALYs for an intervention is shown in figure 3.7:
Figure 3.7 - QALYs gained from an intervention, taken from Drummond et al., 2015:9.
Whilst it is recognised that QALYs are the outcome metric preferred by NICE, the ECDC argue that use of cost per QALY to measure the effectiveness of chlamydia screening is inappropriate because “undiagnosed asymptomatic chlamydia infections do not affect quality of life. The complications of chlamydia are also rarely fatal. The impact of chlamydia is therefore mainly through morbidity and decreases in quality of life resulting from PID and its sequelae” (ECDC, 2014:44).
A recent systematic review of economic evaluations has examined the use of QALYs and valuation of health states associated with chlamydia and the consequences of untreated infection (Jackson et al., 2014). Of the 19 included studies in this review, 11 studies cited the same source of QALY information from an Institute of Medicine study (ibid.). The reviewers highlight methodological concerns associated with valuing short-term health states for chlamydia as those highlighted by the ECDC and also draw out the issue of delayed (long term) complications of the disease occurring, in many
Jackson and colleagues argue that further research is required to enable more robust health state measurements to enable economic evaluations for chlamydia screening to be conducted in accordance with NICE standards (ibid.). Owing to the concerns highlighted by the ECDC and Jackson and colleagues, a proxy measure of health outcomes (health complications of untreated chlamydia) has been selected over QALYs as the outcome measure in this analysis. The health complications included in the economic model are:
• Pelvic Inflammatory Disease (PID)
• Infertility
• Ectopic Pregnancy
• Pre-term Rupture of Membranes (PROM)
• Neonatal Conjunctivitis
• Neonatal Pneumonia
• Epididymitis.
Figure 3.8 outlines the research methods chosen to parametrise the economic model, recognising the early stage of technology development. Given the absence of a self-test, a hypothetical scenario was used in order to demonstrate the impact of variance on test performance characteristics.
Figure 3.8 - Research methods used to parameterise the model
A key feature of early economic evaluation is that it is indicative rather than definitive and significant parameter uncertainty is one of the main reasons for this (Sculpher et al., 1997). The aim of early economic evaluation is to provide an indication of the likely costs and benefits of a new technology and to identify areas for further consideration for technology developers. Therefore, sensitivity analysis has been undertaken as part of both the costing study presented in Chapter 7 and the economic modelling in Chapter 8. Reference was made to both the NICE MTEP methods guide (NICE, 2011) and the ISPOR good research practices for parameter estimation and uncertainty (Briggs et al., 2012) to inform the selection of the methods. One way sensitivity analysis was selected as the method so that the impact of varying individual parameters can be seen on the key outcomes to provide insight into the impact on both costs and outcomes. It was concluded that an understanding of the impact at an individual level would be most beneficial in future
Detailed information on the methods for the development and parametrisation of the model, and validity checks are included in section 8.2.