IQuEST Work Package 1: Economic Modelling Do not reproduce without permission
Jon Tosh
2
Presentation • What is an economic analysis? • What is modelling? • Complex modelling • How will modelling inform the IQUEST project? • The modelling process • The model
15/03/2011 © The University of Sheffield
3
What is an economic evaluation?
• A new intervention/treatment/service structure? • A fixed budget to spend? • Need to evaluate the impact: • on resources (costs) • on patients (health benefits)
15/03/2011 © The University of Sheffield
4
What is an economic evaluation? • An economic evaluation is a widely used method to evaluate costs and health benefits • If a new intervention is more costly AND more effective, is it “value for money?”
• There is only so much money (pie) to go around
• Difficult decisions, but economic evaluations help 15/03/2011 © The University of Sheffield
5
What is modelling? • A economic evaluation will require evidence
• This evidence may come from a number of sources • Modelling brings evidence together • A model is a simplification of reality • No offense Kate Moss 15/03/2011 © The University of Sheffield
6
What is modelling? A simple model: • “Usual care” or “New drug” for terminally ill patients? • Evidence required? • Costs/use of resources • Life expectancy (effectiveness) of both treatments
• Quality of life whilst on treatment • Adverse events
• Evidence sources? • • • •
Costs = NHS data Life expectancy = trial data? Quality of life = trial data/mapping? Adverse events = trial data?/observational studies
15/03/2011 © The University of Sheffield
7
What is modelling? • “Essentially, all models are wrong, but some are useful?” G. Box 1987 • However, this does not mean that complex services cannot be modelled
• In fact, this is exactly when things should be modelled! • This is what we are trying to do in the IQUEST project 15/03/2011 © The University of Sheffield
8
Complex Modelling
15/03/2011 © The University of Sheffield
9
Complex Modelling • The current service of care for patients with long-term depression is complex
• “Simple” modelling is unlikely to be useful • Modelling a small part of the system will not capture any impacts “down stream”
• New methods are being developed to inform modelling “Whole Disease Modelling” (Tappenden - University of Sheffield) • The Whole Disease Modelling methodological framework has been used in this project 15/03/2011 © The University of Sheffield
10
Complex Modelling • Why build a Whole Disease Model (WDM)? • Depression is a chronic condition, what happens now will have an impact on the future • WDM models a patients‟ lifetime • WDM models the complete service
• WDMs are adaptable • Can potentially be used to answer any economic evaluation question • This means future modelling will be more efficient 15/03/2011 © The University of Sheffield
11
Complex Modelling To undertake a Whole Disease Model 1.
Develop a Conceptual Model •
How does the current system work?
•
Draw from literature and experts
2. Develop a Mathematical Model •
Model the natural history of the disease
•
Evaluate the system with simulated patients
•
Draw from literature, data and experts
15/03/2011 © The University of Sheffield
12
Conceptual Model (understand the current system)
15/03/2011 © The University of Sheffield
13
Mathematical Model
(use the conceptual model to develop the mathematical model)
15/03/2011 © The University of Sheffield
14
How will modelling inform the IQUEST project? •
Conceptual modelling of the current system has value
•
Improving psychological therapies is likely to have: •
An impact on patients
•
An impact on resources
•
A model will allow us to test improvements to the system
•
Also “what ifs?” can be tested •
What if we could improve X? What is the impact?
15/03/2011 © The University of Sheffield
15
How will modelling inform the IQUEST project? Q: “Wow, that sounds great! But isn‟t it a bit ambitious?” A: “Yes! But the methods and tools are there to be used. It is a challenge, but we have the opportunity to improve care for patients”
15/03/2011 © The University of Sheffield
16
How will modelling inform the IQUEST project? •
Today, I‟ll explain how the model currently works
•
We can discuss how the model could be improved
•
We can think of some potential ways to improve the system
•
I can go away and put them through the improved model
•
We can come back in April, consider the results and decide which ones are best
•
Work Package 3 will take these and test them in the bench-marking study and the research service
15/03/2011 © The University of Sheffield
17
The model • A conceptual model has been developed • This provides an agreed basis for how the service for patients with long term depression is organised • The mathematical model is an extension of the conceptual model
• The mathematical model is in development • Searches for literature evidence are taking place • Data are being obtained from the Trust
15/03/2011 © The University of Sheffield
18
The model • Patients are simulated • Characteristics (age, gender, socio-economic status)
• History (2 previous depressive episodes or diagnosed dysthymia)
• The model has two „halves‟ 1.
Disease half
2.
System half
•
Represents the „patient‟ and the „service‟ •
The service impacts on the patient
15/03/2011 © The University of Sheffield
19
The disease half
15/03/2011 © The University of Sheffield
20
The disease half • Assume there is no health care system • The model assumes that a patient will have an episode of depression, followed by a period of feeling more „normal‟ • Relapsing/remitting model • Apologies if terms are a little crude
• The model treats dysthymia in the same way but • much longer disease episodes 15/03/2011 © The University of Sheffield
21
The disease half • The patient can be in one of three health states: 1.
Depressive episode
2.
Normal
3.
Dysthymic episode
•
The patients „cycle‟ between these states through their lifetime, before moving to the final state
•
The model can capture varying levels of severity •
However requires detailed evidence
15/03/2011 © The University of Sheffield
22
The disease half • The model contains the functionality to record a patient‟s history • Number of previous episodes • Length of previous episodes
• This could be useful, because it allows data to be analysed by these variables • QUESTION FOR LATER – appropriate source of evidence for parameters? 15/03/2011 © The University of Sheffield
23
The service half
15/03/2011 © The University of Sheffield
24
The service half • Now assume there is a health service established • Patients have the choice now to go and receive care
• The impact of the service is on the disease half of the model • Treatment will hopefully mean shorter, less severe and less frequent episodes • The model records the resources used as the patient accesses services
• In reality there is also a private sector and third service • Outside the possibilities of this model – a limitation • If patients don‟t enter the system, or drop out, we cannot truly know what happens to them
15/03/2011 © The University of Sheffield
25
The service half • In this model, resources are not fully constrained • Patients can access services, and are not bound by a shortage of clinicians/nurses etc • However the time taken to access services and move through the system is incorporated • If patients progress slowly through the system, then improvements may be due to the natural progression of their disease
15/03/2011 © The University of Sheffield
26
The service half EXAMPLE • Patient is simulated to have a depressive episode for 2 months
• Patient enters system (sees GP) after 1 month • Receives counseling, which is ineffective, is assessed and begins low intensity IAPT at 2 months, but depressive episode has just finished • Quicker entry to system, quicker initiation of treatment(s) may have benefited the patient • Completing treatment may not have a short term impact if episode ends naturally, but may have longer term impact • QUESTION FOR LATER – is this realistic? 15/03/2011 © The University of Sheffield
27
The service half • This is an evidence challenge, it may explain (or be confounded by) patients „dropping out‟. They may feel better and stop attending • However some patients will drop out for other reasons • Assumption at the moment is that if patients do not complete a course of treatment, then the length of the current episode is determined by the disease half of the model • QUESTION FOR LATER - Is this a fair assumption? • Probably a conservative assumption, receiving part of a treatment course may have some benefit
15/03/2011 © The University of Sheffield
28
The service half • More likely to occur is that patients are simulated a long depressive episode, and fail to respond to treatments • This failure to respond triggers „step ups‟ through alternative treatments and secondary/tertiary services • These step ups have been identified in the conceptual model and NICE Clinical Guidelines
15/03/2011 © The University of Sheffield
29
The model • Disease half and Service half explain the overall concept of how the disease and service is modelled • The model retains specific sections of the service (primary, secondary and tertiary care)
15/03/2011 © The University of Sheffield
30
Disease Half
Service Half
15/03/2011 © The University of Sheffield
31
15/03/2011 © The University of Sheffield
32
Presentation • The model simulates patients presenting to primary care • Once in the service patients can be referred between secondary and tertiary sectors • When discharged, a patient can only re-enter via primary care
Primary Care 15/03/2011 © The University of Sheffield
Secondary Care (CMHTs)
Tertiary Care (SPS)
33
15/03/2011 © The University of Sheffield
34
Primary Care • Primary Care includes a GP consultation, and then 4 possible treatments:
•
1.
Medical treatment
2.
Counseling
3.
IAPT (low intensity)
4.
IAPT (high intensity)
The model contains the functionality that patients will be tried on different treatments sequentially, but not in combination
15/03/2011 © The University of Sheffield
35
Primary Care • Evidence limited on which order treatments are usually given
• QUESTION FOR LATER – what might this be? • Variation likely between practices, and this can be modelled. A decision rule could be explored: • “Is it cost effective that all patients with symptoms/history of long term depression receive X first?”
15/03/2011 © The University of Sheffield
36
15/03/2011 © The University of Sheffield
37
CMHTs • Community Mental Health Teams (CMHTs) • A referral to CMHT by Primary Care is assessed, and can be rejected (signposted to other services or returned to referrer) • A referral can be routine or urgent (difference in timing)
• A range of potential treatments for patients with LTD: • In general these are combinations of care coordination, medical treatments and individual CBT
15/03/2011 © The University of Sheffield
38
CMHTs • The model allows for monitoring of patients whilst under CMHT maintenance care • Maintenance therapy aims to prevent relapse/recurrence of depression
• If patient has responded to CMHT maintenance care then referred back to GP • Evidence of the effectiveness of these maintenance treatments will be used to model the treatments offered by CMHTs 15/03/2011 © The University of Sheffield
39
15/03/2011 © The University of Sheffield
40
SPSs • Specialist Psychological Services (SPS) • A referral to SPS comes from CMHT • Patients with complex or severe depression • Patients who have not responded to CMHT treatment
• Individual and group CBT offered • Other individual and group psychological therapies offered
15/03/2011 © The University of Sheffield
41
Conclusion • A whistle-stop guide to modelling, and the modelling work for the IQUEST Project
• Please feel free to ask questions now • We have provided worksheets with some questions for you • Please discuss these in your groups and we will have a period of feedback
• Thank you!
15/03/2011 © The University of Sheffield
