Although the economic model was conceived as part of workstream 1 (), it was developed in parallel with other workstreams and drew on data and findings from workstreams 2 and 5 to inform the model structure and to populate the model with data. Accordingly, the methods and results are reported here rather than in Epidemiology of psoriasis and its association with risk of cardiovascular disease.
Workstream 1: relationship of study 1.iii to other IMPACT programme workstreams.
The data used in the model come primarily from workstreams 2 and 5 plus data from the economic review and other published sources.
Economic model
Population, intervention, comparator, outcome, time (PICOT) statement:
The target population for the economic model was patients with mild, moderate or severe psoriasis. The model used the sample of patients in the cohort study in IMPACT workstream 2, utilising the rich data set containing 287 patients. This assumes that patients recruited to the study were representative of the target population.
The intervention comprised the PSO WELL clinician training and patient information materials as a package of care.
The comparator was treatment as usual (TAU), defined as the usual management of psoriasis patients in primary and secondary care without any PSO WELL patient information materials or clinician training taking place.
The primary outcome of the economic analysis was the incremental cost-effectiveness ratio (ICER). This was estimated as the net cost of the intervention divided by the net health benefit. The measure of health benefit for the primary analysis was the quality-adjusted life-year (QALY).
The time horizon for the evaluation was 10 years from entry to the model.
Approach
The economic model aims to provide an assessment of the costs and outcomes of usual care and an exploratory evaluation of the cost-effectiveness of the PSO WELL intervention compared with TAU. The perspective is the UK health and social care system. This was used to define the range of costs included in the analyses. The time horizon for the primary analysis was 10 years, the maximum of previous analyses identified in the literature review. A longer horizon would stretch the limited evidence available too far, whereas shorter time horizons are assessed in scenario analyses. The key outcomes were incremental health benefit, measured in QALYs, and incremental costs. These were used to calculate the ICER, which is the primary outcome.
Model development
The model was intended to be a simplification of psoriasis care in real life, rather than to incorporate all the possible characteristics and interactions of patients, health-care professionals and treatment. Accordingly, the model was designed to incorporate key features identified in the different components of the IMPACT research programme.
Published economic evidence and guidelines61,182–188 combined with the early evidence from the qualitative research components of the IMPACT research programme were used to characterise current management of psoriasis and develop a preliminary version of the IMPACT programme cost-effectiveness model. Key decisions about model goals, structure and assumptions were made by the multidisciplinary IMPACT programme research team through a structured programme of consultation. The process involved discussion papers, short surveys, one-to-one and small group meetings and workshops at FIMs over the course of the IMPACT programme [see Appendix 10 for further details and for the completed International Society for Pharmacoeconomics and Outcomes Research (ISPOR) guidelines checklist189]. The model was also discussed with the Independent Expert Advisory Committee at key points in development. This evidence-led fine-tuning helped ensure that the economic model structure provided a meaningful and accurate characterisation of the key components of the psoriasis care pathways from the perspectives of patients and clinicians.
Psoriasis is a complex condition with great diversity of potential patient pathways. Although the individual treatments used for psoriasis appear simple, they are prescribed in a complex package of care and interactions between patient and clinician. A simple characterisation of the pathway through different treatments indicates patients starting with topical therapies and stepping up to phototherapies and/or systemic non-biological and biological treatment if symptoms are not adequately managed. These steps in treatment are also referred to as the treatment ladder, reflecting both the move from first- to second- and third-line therapies and the increasing toxicity associated with second- and third-line treatment.
In reality, progression from one treatment to the next is determined by a number of other factors. For example, the point at which a person with psoriasis seeks treatment varies – this may be when symptoms first appear, when topical therapy is an appropriate first line of treatment or when symptoms have progressed to be more severe, requiring early systemic treatment. Patients starting with topical therapy may disengage with treatment and not seek further care until the symptoms require treatment with the more advanced systemic therapies.
In addition, evidence provided in earlier chapters suggests that physical, social and emotional challenges are important aspects of psoriasis, but clinicians do not always broach these topics during consultations.111,112 This means that the interaction between patients and clinicians can also play an important role in prescribing and medication usage as well as the success of psoriasis management. Finally, treatment depends on the interface between primary and secondary care and interactions between generalist and specialist health-care professionals.
The modelling studies included in the systematic review typically used either decision tree models or Markov models. Decision tree models compare ‘branches’ of treatment strategies, made up of all possible treatment decisions. Characterising the complexity of treatment and interactions noted above would make a decision tree unwieldy, with each possible long-term event or treatment decision exponentially increasing the overall number of branches and treatment strategies being compared.
A Markov model is made up of a number of unique health states between which simulated patients move over time. Each health state is associated with certain health or cost outcomes, permitting a simulated patient cohort to move between these. Typically, any individual patient’s route to a particular health state is not considered and takes little or no account of their characteristics or how long they have lived with the disease. It is clear from the outcomes of workstreams 3 and 4 that this would be a limiting assumption in a model for psoriasis care, as patient history is highly important; for example, did their clinician provide adequate information? Did the patient see the same clinician or a different one at follow-up? These historical events and characteristics have a bearing on a patient’s probability of experiencing another event; this restricts the usefulness of decision tree or Markov models.
These factors point to the need for an approach that can incorporate the complexities of psoriasis and psoriasis care. Agent-based modelling imitates key actions and interactions of individuals or agents (e.g. patients and health-care professionals) and organisations. The aim is to assess what the effects of these interactions might be on the system as a whole. DES models allow an individual patient to take a random walk through a range of possible events, explicitly taking account of the diversity of potential patient pathways in psoriasis. A hybrid model that incorporates both agent-based and DES approaches was chosen. This allows a large number of potential courses of disease and treatment and allows repeated interactions between the different individuals (agents) in the model.190
Overview of economic model structure
The structure of the economic model is illustrated in and briefly described in this section. Further technical details about the structure are given in Appendix 11.
Economic model structure.
The model compares TAU with the PSO WELL intervention plus TAU. The patient information materials and clinician training developed and evaluated in workstream 5 (see Psoriasis and Wellbeing (PSO WELL): developing patient materials to broaden understanding of psoriasis as a long-term condition, psoriasis-associated comorbidities and the role of self-management and Psoriasis and Wellbeing (PSO WELL): developing a patient-centred approach using motivational interviewing skills with dermatology clinicians to support healthy living in people with psoriasis) are considered together as a single PSO WELL package. It is assumed that the distribution of patient materials and training of clinicians occur right at the start of the simulation. The key differences between the PSO WELL arm and the TAU arm is the assumption that (1) the patient materials could change the likelihood that patients make beneficial changes to their lifestyle and health behaviours and (2) training makes clinicians more likely to provide appropriate treatment and use motivational interviewing techniques. These, in turn, influence patient health, need for health care and health-care costs.
At the start of each simulated patient journey, the patient enters the model with a profile of sociodemographic and health characteristics. The sociodemographic characteristics include age, gender, ethnicity and Townsend deprivation score for the local area where the patient lives. The health characteristics include the severity of psoriasis, PsA, anxiety and depression, BMI, type 2 diabetes, MI and risk of future MI or stroke. Treatment history attributes include prior and current psoriasis treatments. Lifestyle behaviours included are alcohol consumption, extent of physical activity and smoking status.147
Psoriasis severity determines what type of clinician a person sees routinely and what treatments are prescribed. The interactions between patients and practitioners determine many events that happen during a patient’s journey through the model. The model focuses on interactions between the patient and their GP and between the patient and the dermatology service and also includes the wider costs and health impacts of the comorbidities associated with psoriasis. The patient uses health-care resources (e.g. interacting with clinicians at clinical consultations, receiving treatment) until either the model time horizon is reached or the patient dies. The number of health-care services used and the costs depend on each person’s characteristics and how they interact with the health-care system. At the same time, the patient may or may not make changes to their lifestyle. These changes will then influence their health and need for health care.
Outcomes are recorded at the end of the simulation to calculate mean and total costs and QALYs across the whole simulated population, for usual care alone and for usual care plus the PSO WELL interventions.
The model includes a number of assumptions to simplify the components and the patient’s journey through the model and to estimate parameters the data used. The model assumptions are described in detail in Appendix 11, Key components and assumptions, and the data assumptions are included in Appendix 11, Tables 26–30. The key assumptions about the patient’s journey include:
A proportion of clinicians will act in accordance with clinical guidelines. This will be higher for clinicians who receive the training intervention than for those who do not receive the training intervention.
If a clinician acts in accordance with clinical guidelines, then the model assumes that this clinician’s treatment decisions are made in line with evidence-based guidelines.
135If a clinician does not act in accordance with clinical guidelines, then the clinician will, in the model, prescribe phototherapy if required; the specific type of phototherapy regimen is selected at random by the model. If a patient with severe psoriasis has previously received phototherapy, then a systemic therapy is prescribed; again, the type of systemic therapy is selected at random by the model.
A clinician’s propensity to use motivational interviewing skills during consultations depends on their knowledge of and ability to use motivational interviewing skills and their assessment of the relevance of these for the patient and for treatment decision.
A patient’s time until next consultation depends on their current psoriasis treatment –
routine GP appointments occur every 6 months to review progress with topical therapy
an appointment is scheduled to take place 2 weeks after referral from a dermatologist, a psoriasis flare or patient help-seeking following the PSO WELL intervention
routine dermatologist consultations occur every 10–16 weeks depending on the specific complex treatment regimen.
The treating clinician may identify the need for a formal lifestyle or comorbidity intervention. To manage an identified comorbidity, it is assumed that patients will receive the next available intervention in the relevant treatment sequence derived from clinical guidelines
182–187The patient will agree to a lifestyle intervention only if it is a behaviour that they are intent on changing.
Key variables and data
The variables, data sources and parameter estimates are detailed in Appendix 11, Key variables; Appendix 11, Parameter estimates and data sources for the economic model; and Appendix 11, Tables 26–32. The model includes psoriasis treatment effects that were identified from the economic systematic review [Dermatology Life Quality Index (DLQI) and PASI]. These were converted to a common measure for the model using published algorithms (see Appendix 11, Key variables, Treatment effects). The model includes the outcomes of behaviour change and comorbidity interventions, which were identified from clinical guidelines61,182,183,185–187,191 and the sources used to develop the guidelines.
Psoriasis flares occur in the model, potentially causing people with mild disease to progress to severe psoriasis. A flare will cause an increase in psoriasis symptoms and/or decrease in treatment outcomes. It is assumed that even if patients with severe psoriasis have their symptoms effectively managed, they cannot return to having mild or moderate psoriasis (see Appendix 11, Key variables, Psoriasis flares).
Quality-adjusted life-years were used as the measure of total health benefit experienced by simulated patients over the course of their journey through the model. A QALY measures the length of survival and adjusts this to take into account any ill health during that time. For example, a person in full health for a year will have a QALY of 1, whereas a person who is in less than full health may have a QALY of 0.25. Death is assigned a value of zero. The model allows a simulated patient’s health and QALYs to vary as events occur and their attributes change (see Appendix 11, Key variables, Quality-adjusted life-years). Time to mortality was predicted from all-cause mortality.192 The only cause of early mortality is fatality due to MI193 or stroke (see Appendix 11, Key variables, Mortality).194
Various health-care resources and costs are required throughout the simulation (see Appendix 11, Key variables, Costs and resource use). These included the costs of the PSO WELL interventions and other health-care costs. The costs of the clinician training element depended on the number of clinicians trained as well as the number of psoriasis patients managed by each clinician. For the primary analysis, we conservatively assumed that each clinician managed only one patient, which increases the unit cost per patient of the training. A sensitivity analysis was used to explore the impact of increasing the number of patients treated by each clinician.
Other health-care costs included the costs of GP and dermatologist consultations, topical treatment, phototherapy, non-biologic and biologic systemic therapies and the costs of interventions to support lifestyle changes and to manage comorbidities.
All costs and QALYs were discounted at an annual rate of 3.5% to account for social time preferences as recommended for the UK.195
Analysis
The model produced simulated pairs of net costs and net outcomes. These were used to generate cost-effectiveness acceptability curves, as recommended by NICE for health technology appraisals,195 and to estimate the probability that the PSO WELL intervention is cost-effective compared with TAU. The cost-effectiveness acceptability approach revalues effects or benefits in monetary terms. However, in the UK there is no universally agreed monetary value for the types of health benefit measures (such as QALYs or relapse-free years) typically used in cost-effectiveness analyses. An approach used in health care is to ask the question: what is the maximum amount that decision-makers are willing to pay to gain one unit of health benefit? The simulated net QALYs were revalued using a range of maximum willingness-to-pay (WTP) values that a decision-maker may be willing to pay to gain 1 QALY. The WTP thresholds used ranged from £1 to £50,000 to gain 1 QALY. This range of values was based on the range of WTP values historically implied by NICE decisions.196
The model consists of an inner loop that simulates one patient journey for each simulated patient. There is an outer loop that repeats the inner loop 250 times. Each of these iterations uses a new set of values for the model variables. Therefore, if one patient journey is modelled, a total of 500 unique patient experiences are simulated (250 for the TAU arm and 250 for the PSO WELL arm). This addresses the uncertainty that comes from the differences between patients in terms of their sociodemographic and health characteristics. The patients included in the model were the 287 people included in the screening study of CVD risk in patients with psoriasis (study 2.i). This gives 71,750 unique patient experiences (250 journeys × 287 patients) for the TAU arm and for the PSO WELL arm of the model. This adds in the uncertainty in the data used to estimate the variables or parameters included in the model. All results presented are therefore probabilistic.
The model structure and simplifying assumptions introduce another source of uncertainty owing to structural or design decisions. Threshold and scenario analyses were used to assess the impact of these on the results. The threshold analysis explored the impact of increasing the number of patients who could benefit from the PSO WELL intervention. The scenario analyses looked at the impact of changing key assumptions used in the model. The technical development and data analysis were undertaken in R version 3.2.0 (The R Foundation for Statistical Computing, Vienna, Austria).
Results
Primary analysis
The detailed results of the primary and sensitivity analyses are given in Appendix 11, Tables 32 and 34, and Appendix 11, . In the primary analysis, the model predicts that TAU is associated with 4.74 QALYs (95% CI 4.66 to 4.83 QALYs) per patient over 10 years (see Appendix 11, Table 34). In terms of costs, TAU is expected to incur a total cost of £26,242 (95% CI £25,852 to £28,452) per patient. The PSO WELL arm of the model generates 4.75 QALYs (95% CI 4.64 to 4.82 QALYs) per patient, with a mean cost of £27,801 (95% CI £24,902 to £27,466) per patient.
Overall, the model predicts a small gain in health benefit of 0.01 QALYs for PSO WELL (95% CI –0.10 to 0.13 QALYs) and with a mean net cost of £838 compared with TAU (95% CI –£1106 to £2593). This difference is smaller than the clinician intervention cost per patient (£1133) and may indicate that minor cost savings are made elsewhere to offset it. However, the 95% CI for the difference in costs crosses zero, indicating that the difference in cost could have occurred by chance.
The ICER is £72,802 per QALY. If decision-makers are willing to pay £20,000 to gain 1 QALY, then the probability that the PSO WELL intervention is cost-effective is 34%. This means that if only one patient per clinician trained benefits from the PSO WELL intervention, it is unlikely to be a cost-effective use of NHS resources (see Appendix 11, Table 34, and Appendix 11, and ).
Threshold analysis
Threshold analysis was used to explore whether or not the relative cost-effectiveness of the PSO WELL intervention package improved if more patients could benefit from clinician training. This indicates that if decision-makers are willing to pay £20,000 to gain one QALY, the PSO WELL intervention may be cost-effective if a trained typical clinician group (average of 10 clinicians per group) manages ≥ 22 psoriasis patients (see Appendix 11, ). This equates to each clinician applying their new knowledge and skills to two patients in the year following training, thereby generating more health gains and costing less per patient. PSO WELL dominates TAU, simultaneously improving health outcomes and causing a net saving if 40 patients per clinician group (or four patients per clinician) could benefit from the clinician training.
Scenario analysis
Scenario analyses were used to assess whether or not the probability that the clinician training was cost-effective changed if alternative assumptions about key model inputs were changed (see Appendix 11, Table 35). For comparability of results, we assumed that 10 patients stand to benefit from training a typical clinician group of 10 clinicians.
The PSO WELL intervention is more likely to be cost-effective if the effect of training is maintained for > 1 year. If the impact of the clinician training is maintained for the 10-year time span of the model, then incremental costs fall and the probability that PSO WELL is cost-effective increases to 44%. If the number of patients benefiting per clinician group trained is increased to 16, the probability that PSO WELL is cost-effective increases to 52%.
The cost-effectiveness of PSO WELL improves if all dermatologists always act in a guideline-concordant manner and 20 patients benefit per clinician group (or two patients per clinician trained), at which point PSO WELL has a 50% probability of being cost-effective. This result suggests that a large benefit may be achieved if GPs correctly refer severe psoriasis patients to secondary care when it is guaranteed to be guideline appropriate. PSO WELL is less likely to be cost-effective over shorter model durations (suggesting that cost offsets are realised over longer durations) or if the PSO WELL development costs are included on top of delivery costs.
Discussion
This exploratory cost-effectiveness analysis estimates that the PSO WELL interventions have the potential to be cost-effective for the NHS. The likelihood of this depends on the number of patients who stand to benefit from a typical group of GPs and dermatologists being exposed to the intervention. If a trained group manages more patients, then the clinician intervention cost is spread over a larger potential pool of people who can benefit. If around 22 patients are managed per group of 10 clinicians, our analysis suggests that PSO WELL may be cost-effective. This is less likely to be true if costs incurred in developing the interventions are included in the analysis, rather than considered to be sunk costs.
Despite its exploratory nature, this evaluation possesses a number of strengths. First, as a DES, the model allows for repeated interactions between clinicians and patients, and contains a large number of psoriasis treatment options, health-related behaviours and comorbidities. This would be impractical with alternative models. Second, model development involved regular collaboration between multidisciplinary experts in the IMPACT programme team. Furthermore, a good-practice checklist was used to validate the model throughout its development (provided in Appendix 10).189
Nevertheless, there were a number of challenges to address in building the model. As a consequence, the model required a number of simplifying assumptions to make the evaluation feasible. In addition, limitations with the evidence available meant that some assumptions about the model inputs were required. These mean that the evaluation does have some limitations, which also highlight where further research may be useful.
Our analysis used the sample of participants recruited into the CVD risk factor screening study (in workstream 2) as the source of data about patient baseline characteristics.66 A key reason for this was the availability of detailed information about the study participants’ health and lifestyle characteristics. However, for the reasons noted in Cardiovascular risk in patients with psoriasis, the sample may not be representative of the target population of people with psoriasis for the economic evaluation in terms of CVD risk. If this sample of patients has a higher CVD risk than the wider population of people with psoriasis, then this may overestimate the impact of the intervention on patients’ overall health and QALYs and underestimate the occurrence of events that incur health-care costs. Even so, the analyses of the CPRD data (see Epidemiology of psoriasis and its association with risk of cardiovascular disease) do indicate that people with psoriasis have higher rates of CVD than the general population.
Our model omits a number of features that would ideally be included for the simulation to accurately reflect reality. Simplifications have been made to the psoriasis treatment logic. These include assuming random treatment selection among clinicians who do not follow published national guidelines, excluding contraindications to treatment and excluding the role of dermatology clinic nurses. Further work is needed to understand whether this would under- or overestimate the relative cost-effectiveness of the PSO WELL intervention.
Non-adherence to treatment by patients is also missing from the model, which may have a significant impact on the effectiveness of psoriasis therapies.133 If the patient materials and/or clinician training improve patient adherence with psoriasis treatments or increase engagement with health care, then our evaluation may underestimate the health benefits of the PSO WELL intervention and underestimate its relative cost-effectiveness. However, improved adherence or engagement may also mean that people with psoriasis use services more intensively and/or use a wider range of health-care services. This would increase the costs and reduce the potential cost-effectiveness of the intervention.
The potential role and effectiveness of the patient education intervention is limited in the model. Its only positive effect is that a patient may choose to seek a GP consultation. In reality, psoriasis education might benefit patients more broadly. This could be by helping patients to manage their psoriasis and adhere to treatment. Patients may also be motivated to change their lifestyle. The benefits of these changes are not captured in the model, which may underestimate the cost-effectiveness of the PSO WELL intervention.
The clinician intervention is more influential in the model; however, we assume that its benefit expires after 1 year, a potentially conservative assumption that underestimates the cost-effectiveness of the PSO WELL intervention. The scenario analyses indicated that increasing the time for which the training is effective decreases costs and increases patient health benefit.
The probability that a clinician acts in accordance with the guidelines affects participants’ health in terms of the effectiveness of the treatments they receive. There was limited information to inform the estimate of whether or not a clinician acted in accordance with the guidelines. For the primary analysis, it was assumed that 78% of PSO WELL-trained clinicians would use treatment guidelines appropriately, compared with 59% of clinicians who did not receive the PSO WELL training.130,164 If the difference in appropriate use of treatment guidelines is lower than assumed, the results will overestimate the cost-effectiveness of the PSO WELL intervention. If the difference is higher than assumed, our results will underestimate the cost-effectiveness of the intervention. The scenario analyses indicate that this is an important variable in the model, suggesting an area for further evaluation of the intervention.
An important consideration is how to appropriately distribute the cost of the clinician intervention to accurately estimate the cost-effectiveness of PSO WELL The true cost-effectiveness depends on the number of psoriasis patients able to benefit. The present primary analysis assumes that each clinician will be able to apply their new skills and knowledge to only one patient, following training, which is a conservative, minimum figure. The threshold and scenario analyses indicate that the cost-effectiveness of the PSO WELL intervention is sensitive to the number of patients who could benefit from clinician training. Further research is needed to clarify the number of patients with whom each clinician could realistically use their knowledge and skills.
Finally, only one intervention is included in the analysis: the package of patient materials and clinician training combined. This is owing to the low cost per patient of the patient information materials component of PSO WELL and limited information about the relative effectiveness of the patient materials compared with clinician training. As further evidence becomes available it will be possible to evaluate each of the components individually.
Despite these limitations, our model provides initial information to inform future research and evaluate the PSO WELL intervention as new evidence becomes available. The model also provides a structure that incorporates the complexities of psoriasis treatment and interactions between patients and clinicians that affect the effectiveness and cost-effectiveness of any treatment. The model can be adapted to evaluate new approaches to psoriasis management that address the range of issues found to affect patient care that were identified by the IMPACT programme. Given this, the analysis presented provides a valuable preliminary estimate of the health economic outcomes associated with the primary care management of psoriasis as a complex condition, and of the PSO WELL interventions.
Summary of the health economic analysis
We undertook a systematic review of cost-effectiveness analyses, which found that there was little consensus about the cost-effectiveness of treatments and care strategies for individuals with psoriasis. The analyses were limited by a lack of quality evidence and frequently provided only a limited reporting of methods. We developed a new microsimulation model to evaluate the potential economic acceptability of the PSO WELL interventions developed. The simulation model utilised inputs from parallel IMPACT programme workstreams, the systematic literature review, clinical guidelines and expert opinion within the project team. The model was used to provide preliminary estimates of the economic outcomes, including cost-effectiveness, associated with the PSO WELL interventions. The exploratory analysis indicated that the interventions developed in this programme of work have the potential to be cost-effective and identified areas where more work is needed.
Key conclusion
The PSO WELL intervention was found to be associated with a small QALY gain per patient and a net cost comparable with TAU; however, the cost-effectiveness of the PSO WELL intervention depends on how many patients are expected to benefit from improvements in clinician knowledge and skills and how long those improvements are expected to be maintained.
Implications
If each of the clinicians trained as part of the PSO WELL intervention routinely managed more than one psoriasis patient and more than one psoriasis patient benefited from that training, the intervention cost per group trained could generate health benefits more efficiently. Our analysis suggests that the combined patient and clinician interventions may be a cost-effective use of NHS resources if a typical group of 10 clinicians trained manages around 22 psoriasis patients.
We recommend research into the following areas to improve the data available for future economic analyses and to reduce uncertainty about the results:
We recommend quantitative and qualitative research, including integrated economic and clinical randomised controlled trials, in the following areas – assessment of the number of patients likely to benefit from clinician training and the time the effectiveness of training is likely to last.
Quantification and modelling of the benefits of the patient information materials on patients’ ability to manage their psoriasis and broader health.
Evaluation of how clinicians who do not follow published national guidelines select treatment and what treatments are typically used in this situation.
Evaluation of the impact of the PSO WELL interventions on patient adherence and engagement with psoriasis treatment and health-care services and clinician adherence/concordance with treatment guidelines.
Extending the model to include new evidence and incorporate contraindications to treatment as well as the role of dermatology clinic nurses.
