Presentation by David Wonderling, Head of Health Economics at National Guideline Centre, Royal College of Physicians and Lauren Ramjee, Senior Health Economist, Royal College of Physicians.
This workshop outlines the principles of health economic evaluation for the NHS.
2. Who are we?
ā¢ David Wonderling, Head of health economics
ā¢ Lauren Ramjee, Senior health economist
ā¢ National Guideline Centre
ā Hosted by Royal College of Physicians
ā Commissioned to develop guidelines by National Institute of Health and Care
Excellence (NICE)
ā¢ NICE guidance
ā Guidelines
Ā» Clinical, public health, social care, service delivery
ā Technology Appraisals
ā Medical Technologies and Diagnostics
ā (Interventional procedures)
5. Aims
1. Understand what health economics is about
2. Understand what an Economics Evaluation is
3. Identify different types of economic evaluation
6. Dispelling Myths
Myth 1: Health economics is about saving the government money
ā¢ Economics is the study of how to best allocate scarce resources in order to
maximise benefits to society
ā¢ Health economics helps NHS use its limited budget to maximise health
outcomes for the whole population
ā¢ Identify interventions which offer the best value for money
7. Dispelling Myths
Myth 2: A cheap intervention is cost-effective
ā¢ Surgery A costs Ā£500 while surgery B costs Ā£800 butā¦
ā¢ More re-operations after Surgery A
ā¢ And/or more complications after Surgery A
ā¢ Eventually Surgery A generates more costs
8. Dispelling Myths
Myth 3: Expensive interventions are not cost-effective
ā¢ A strategy which is very expensive might generate substantial future
savings and/or improve health substantially
ā save more lives
ā Fewer adverse events
ā Better quality of life
9. Dispelling Myths
Myth 4: Health economics is only concerned with expensive drugs &
high technology
ā¢ Health economics is relevant for any clinical question if there could be a
difference in resource use:
ā What should staff use to wash their hands?
ā What is the best platform to store patient information/records?
ā Should point of care tests be used?
ā Should monitoring be conducted remotely (via tele medicine)?
ā What sequence should tests be done?
ā Should follow-up be conducted in hospital or in primary care?
ā What types of rehab should be offered after stroke?
10. Challenges to the NHS
ā¢ Everyone wants better health and healthcare
ā¢ But resources are limited
ā Staff e.g. doctors and nurses
ā Facilities e.g. hospitals
ā Equipment e.g. MRI scanners
ā Consumables e.g. drugs
ā¢ The NHS does not have limitless spending therefore there is an
āopportunity costā to spending i.e. the value of the best alternative use of
resources
ā¢ If the NHS spends more on one thing it has to spend less on something
else. But how can we decide?
11. Treatment A
(usual care)
Treatment B
(new treatment)
Costs
Health outcomes
Costs
Health outcomes
Economic Evaluation
ā¢ ā... the comparative analysis of alternative courses of action in terms of
both their costs and consequences.ā (Drummond et al. 2005)
12. Types of Economic Evaluation
Type of analysis
Value of
resources
Value of health gain
Cost-utility Ā£ Combined index:
Quality Adjusted Life Years (QALY)
Cost-effectiveness Ā£
Single indicator:
Weight loss (kg),
blood glucose control (HbA1c)
deaths averted,
life years savedā¦
Cost-consequences Ā£ Multiple indicators
Comparative cost /
Cost minimisation
Ā£ None
Cost-utility analysis is desirable.
Cost-consequences or cost minimisation
might be more pragmatic and sufficient
13. Cost effectiveness vs. resource/budget impact
Cost-effectiveness Resource/budget Impact
Is it value for money? How much it will cost?
Costs, savings and health outcomes Costs and savings
Inform policy / purchasing For planning & implementation
Time horizon up to lifetime Time horizon of 1 to 5 years
Cost-effectiveness ratio
e.g. Cost per quality-adjusted life-
year (QALY) gained
Cost per patient
Total cost (for Trust/CCG/England)
You might also want to conduct a budget
impact analysis
14. Conclusions
ā¢ Health economics is about the optimal allocation of scarce resources
ā¢ There is an opportunity cost for each and every decision
ā¢ Different types of economic evaluation can be conducted
ā¢ NICE prefers
ā Cost-utility analyses for Guidelines and Technology Appraisals
ā Cost-consequences analysis (or sometimes comparative cost analysis) for
Medical Technologies
16. Aims
1. Understand what makes a product efficient for use in the NHS
2. Understand ways that cost-savings can be achieved
17. What makes a product efficient / value for money for the NHS?
ā¢ There are four possibilities that could make a product efficient:
ā Cost saving + clinical benefits
ā Cost saving + clinical equivalence
ā Cost neutral + clinical benefits
ā āCost-effectiveā
23. Treatment recovery is faster
= reduction in hospital length of
stay
Filling out patient records is
faster
= reduction in staff time needed
for specific activity
Time to diagnosis is faster
= reduction in costs of treating
complications prior to a diagnosis
Staff training is faster
= reduction in cost of training
staff
Faster
24. New IT system that centralises
information
= reduction in staff time required
for administrative tasks
More operation are a success
= reduction in long term health
costs
Tele-monitoring
= reduction in cost of monitoring
Less repeat tests required
= reduction in cost of diagnosis
Better
25. Fewer staff injuries
= reduction in staff costs
Fewer adverse events during
surgery
= reduction in long term health
costs
Fewer adverse events after
surgery
= reduction in long term health
costs
Lower radiation dose during
imaging
= reduction in long term health
costs
Safer
26. Conclusions
ā¢ The most attractive products for the NHS are those that improve clinical
outcomes and reduce costs
ā These products DOMINATE current practice
ā¢ If your product is not dominant you will need to demonstrate the trade-off
between health gained/lost and costs increased/saved
ā¢ Products can be cost saving by being faster, better, cheaper or safer than
current practice
28. Aims
1. Specify clearly the question you want the Economic Evaluation to answer
2. Choose appropriate comparator(s)
29. Specifying the question
Population(s)
What specific patient population(s) are we interested in?
This will be our denominator for costs and effects
P
Intervention
What is our investigational intervention?
Specify dose, timing, supportive interventions, etc.
I
Comparison(s)
What are the main alternatives to compare with the product?
a) current usual practice (minimum)
b) alternative interventions that could potentially be as cost-effective as usual practice.
c) no intervention
C
Outcome(s)
What do we intend to accomplish, measure, improve or affect?
a) NHS costs (minimum)
b) patient-relevant health outcomes and/or process/resource impacts
O
30. Time horizon
ā¢ Time horizon
ā Time period over which costs and benefits are measured
ā Need same horizon for both costs and effects.
ā¢ How long?
ā Long enough to include all meaningful differences between the comparators
in terms of their costs and benefits
ā¢ Long time horizon (e.g. lifetime of patients)
ā If thereās a difference in patient survival
ā If the impact on patients is far reaching (e.g. hip replacement)
ā¢ Short time horizon
ā If the impact on patients is short term
ā The study follow-up was short?
ā If the intervention is already cost-effective in short-term and extending the
time horizon would only make it more so.
31. Specifying the question
Are sound generators cost-saving compared to no treatment for people with
Tinnitus?
Are sound generators cost-saving
for people with Tinnitus?
compared to no treatment
Outcome
Intervention
Comparison
Population
32. Specifying the question
What is the most cost-effective way to deliver rehab for people with stroke?
What is the most cost-effective
for people with stroke?
way to deliver rehab
Outcome Intervention
Comparison
Population
33. Protocol
Is GP-AF case finding software cost effective in helping to diagnose people with
atrial fibrillation?
Population People (diagnosed and undiagnosed) with Atrial Fibrillation
Intervention GP-AF case finding software platform
Comparison Compared to opportunistic testing at general practice
Outcomes ā¢ Additional cases of atrial fibrillation detected
ā¢ Additional cases treated with anticoagulation
ā¢ Strokes avoided
ā¢ Quality adjusted life years gained
ā¢ Cost per patient
34. Conclusions
ā¢ For an economic evaluation you need to specify the:
ā POPULATION(S)
ā INTERVENTION
ā COMPARATOR(S)
ā¢ Including usual NHS practice
ā OUTCOME(S)
ā¢ Costs and benefits need to be consistently measured over a time horizon
that captures all important differences between intervention and
comparators
36. Practical
ā¢ Now frame an economic evaluation for your product
ā¢ Specifying the essentials
ā¢ What are the details of the:
ā Population(s)?
ā intervention?
ā comparator(s)?
ā¢ What are the key impacts (Intervention vs comparator)?
ā¢ What time horizon is necessary to capture key impacts?
38. Aims
1. Understand which costs you should be considering
2. Understand how to estimate costs
3. Learn where to get information for costs from
39. Rules for conducting health economic evaluations
ā¢ England and Wales - NICE
ā Methods of Technology Appraisal
https://www.nice.org.uk/about/what-we-do/our-programmes/nice-
guidance/nice-technology-appraisal-guidance/process
ā Developing Guidelines
https://www.nice.org.uk/about/what-we-do/our-programmes/nice-
guidance/nice-guidelines/how-we-develop-nice-guidelines
ā Medical technologies evaluation programme methods guide
https://www.nice.org.uk/about/what-we-do/our-programmes/nice-
guidance/medical-technologies-guidance/how-we-develop
ā¢ Scottish Medicines Consortium
ā https://www.scottishmedicines.org.uk/making-a-submission/
ā¢ Other Countries
ā https://www.ispor.org/PEguidelines/index.asp
40. Which costs should you include?
ā¢ Possible perspectives
ā Trust/Practice
ā CCG
ā NHS (for most NICE evaluations)
ā Public sector
ā¢ NHS + local authorities + central government
ā Societal
ā¢ Public sector + productivity losses/gains + all other impacts
Most likely
41. NICE methods guidance for economic evaluation of drugs, medical
technologies & clinical guidelines
Non-NHS costs
Cost to other government bodies may be included in exceptional circumstances as a sensitivity
analysis.
Costs borne by patients should not be included unless they are reimbursed by the NHS
Patients productivity gains/losses should not be included, even as a sensitivity analysis, as this
would mean we would be prioritising people in work (over the elderly, chronically ill etc.).
Where a technology extends life
NHS costs related to the condition of interest and incurred in additional years of life gained as a
result of treatment should be included in the reference-case analysis. NHS Costs that are
considered to be unrelated to the condition or technology of interest should be excluded.
42. Which costs should you include?
Immediate cost of
investment + Recurrent costs
Savings from
reduced resource
use-
COSTS SAVINGS
43. How to estimate costs
1) Estimate resource use per patient for each intervention
ā E.g. numbers of GP visits, outpatient visits, tests, drug use (HES data, activity
data, audit data)
ā Sometimes reported in clinical trials or other studies
ā May need assumptions from the Committee or other experts
2) Multiply by unit costs for each resource
ā Some standard national sources (e.g. BNF for drugs)
ā Sometimes available from clinical studies
ā May sometimes have to use local estimates
44. Resource use for Health Economic Evaluations
Cost Type Examples
Technology costs ā¢ Medication
ā¢ Medical devices
ā¢ Diagnostic tests
ā¢ IT software
Costs of healthcare service use ā¢ Days in hospital
ā¢ Medical procedures
ā¢ Outpatient visits
ā¢ Appointments / staff time
ā¢ GP, Nurse, physio
ā¢ Emergency services
ā¢ Days in intensive care
Other costs ā¢ Healthcare consumables
ā¢ Training
ā¢ Administration
ā¢ Overheads
45. Data ā How to find and use freely available national unit costs
Type of cost Source for the cost
Drugs NHS Drug tariff
http://www.nhsbsa.nhs.uk/prescriptions
British National Formulary
http://www.bnf.org
Other technologies NHS Supply Chain Catalogue
http://my.supplychain.nhs.uk/catalogue
Staff time āUnit costs of health and social careā
http://www.pssru.ac.uk/project-pages/unit-costs/
Hospital
procedures/stays,
outpatient visits, tests
Department of Health
Tariff and NHS reference costs
https://www.gov.uk/government/publications/payment-by-results-pbr-
operational-guidance-and-tariffs
https://www.gov.uk/government/collections/nhs-reference-costs
46. Using NHS Reference costs
https://www.gov.uk/government/collections/nhs-reference-costs
ā¢ Cost of Admission
ā¢ Cost per Excess Bed day
ā Example: Total knee replacement with no comorbidities or complications
ā¢ Converting from OPCS/ICD10 to HRG
ā¢ See āCode to groupā Workbook for relevant year, e.g.
https://digital.nhs.uk/services/national-casemix-office/downloads-groupers-and-tools/costing-
hrg4-2017-18-reference-costs-grouper
Currency
Description
Cost Unit cost
Very Major Knee
Procedures for Non-
Trauma with CC
Score 0-1
Non Elective long
stay Ā£5692
Excess bed day
Ā£315
47. Using āUnit costs of health and social careā
http://www.pssru.ac.uk/project-pages/unit-costs/
Cost Value Source
GP appointment Ā£37 (Ā£31 without
qualifications)
PSSRU (Curtis 2017)
(p172)
Example
48. Unit costs ā other sources
ā¢ Published studies
ā HEED (health economics evaluation database up to 2014), HTAs
ā¢ https://www.crd.york.ac.uk/CRDWeb/
ā Medline/Embase
ā¢ Manufacturers
ā¢ Trusts
49. Discounting
ā¢ Different interventions give rise to costs and benefits incurred at different
time points
ā¢ People prefer benefits today and costs in the future
ā Costs and benefits incurred today are therefore valued higher than those in
the future
ā¢ We ādiscountā costs and benefits to account for this time preference
ā we can calculate the āpresent valueā of future costs and benefits
ā¢ NICE discounts both costs and health benefits at 3.5% per annum
ā For example, a cost of Ā£1,035 incurred in one years time would be valued at
Ā£1,000 today (Ā£1,000 is therefore the present value)
50. Annuitizing costs
ā¢ Products have an upfront costs and expected lifetimes
ā¢ As we are interested in cost per patient we need to
ā annuitize the upfront costs ā calculate the implicit rental value
ā divide by the expected caseload / output
ā¢ Example
ā GP-AF case finding software platform costs Ā£1,000 upfront
ā Expected lifetime = 5 years (after which a new licence needs to be purchased)
ā Software can help identify 10 people with AF per year
ā¢ Annuitize, accounting for the discount rate (3.5%)
ā Cost of software per year = Ā£221
ā¢ Divide by the outcome
ā Cost per case detected = Ā£22
51. Formulae for discounting / annuitizing
ā¢ Net present value of cost occurring in future
ā ššš = š¶
1
(1+r) š”
ā Where C=cost incurred at time t; r=discount rate (time preference, e.g. 3.5%)
ā¢ Annual equivalent cost of up-front expenditure
ā Annuity factor: š“ =
1ā(1+r)āš”
r
ā¢ Where t=life expectancy of equipment; r=discount rate
ā Annual equivalent cost= š“šøš¶ =
Kā(S/(1+r) š”)
A
ā¢ Where t=life expectancy of equipment, K=cost at time of purchase;
S=resale value at time t; r=discount rate; A= annuity factor
52. Conclusions
ā¢ Costs include both resource use (number of times used) and the cost itself
ā¢ Costs included should reflect the perspective of the target audience
ā¢ There are available sources for many NHS unit costs
ā¢ Costs that occur in the future need to be discounted
ā¢ Up-front costs need to be annuitized
54. Aims
ā¢ Understand what health outcomes you should be measuring
ā¢ Understand whether you can calculate quality adjusted life years
ā¢ Understand the decision rules of economic evaluations
55. Measuring health gain
Type of analysis
Value of
resources
Value of health gain
Cost-utility Ā£ Combined index:
Quality Adjusted Life Years (QALY)
Cost-effectiveness Ā£
Single indicator:
Weight loss (kg),
blood glucose control (HbA1c)
deaths averted,
life years savedā¦
Cost-consequences Ā£ Multiple indicators
Comparative cost /
Cost minimisation
Ā£ None
56. Quality Adjusted Life Years
QALYs =
Life expectancy
(life years)
x Quality of life
(utility)
ā¢ A measure of overall effectiveness (overall health)
ā Length of life
ā Quality of life (utility): 0 (death) to 1 (full heath) scale
ā¢ 2 years in full quality of life = 2 x 1.0 = 2 QALYs
ā¢ 2 years at 50% quality of life = 2 x 0.5 = 1 QALY
ā¢ Life-years can be calculated using life tables or Markov models
57. Utility weights for QALYs
ā¢ A common measure used in Economic Evaluations and NICEās preferred
measure is the EQ-5D health state valuation tool
ā¢ 5 dimensions of health
ā MOBILITY
ā SELF-CARE
ā USUAL ACTIVITIES
ā PAIN / DISCOMFORT
ā ANXIETY / DEPRESSION
58. Sources for utility weights
ā¢ EQ-5D is NICEās preferred method of health related quality of life in adults
ā https://euroqol.org/
ā¢ You can sometimes source utility weights from published literature
ā http://healtheconomics.tuftsmedicalcenter.org/cear4/SearchingtheCEARegistry/SearchtheCEARegist
ry.aspx
ā https://www.scharrhud.org/
ā¢ When EQ-5D data is not available the data can sometimes be estimated by
mapping from other health-related quality of life measures
59. Decision rules of cost effectiveness analysis
ā¢ An intervention is considered to be cost-effective compared to the best
alternative if:
ā It improves health and costs less (is dominant)
ā The incremental cost effectiveness ratio (ICER) is less than the threshold
ā¢ ICER = the difference in mean costs / the difference in mean QALYs
If an intervention requires additional resources these rules
ensure too much health will not be displaced elsewhere in
the health system in order to fund the intervention
Thresholds
-NICE uses Ā£20,00-Ā£30,000 per QALY for most treatments
-DHSC uses Ā£15,000 per QALY for Impact assessments
60. Assessing cost-effectiveness
Treatment cost-effective
in shaded region
Threshold Ā£20,000
per QALY gained
Cost (Ā£)
Effect (QALYs)
Low extra cost
High QALY gain
High extra cost
Low QALY gain
QALY = quality adjusted life year
61. Measuring health gain
Type of analysis
Value of
resources
Value of health gain
Cost-utility Ā£ Combined index:
Quality Adjusted Life Years (QALY)
Cost-effectiveness Ā£
Single indicator:
Weight loss (kg),
blood glucose control (HbA1c)
deaths averted,
life years savedā¦
Cost-consequences Ā£ Multiple indicators
Comparative cost /
Cost minimisation
Ā£ None
62. Cost-consequences / cost-effectiveness analyses
ā¢ The intervention /product improves all clinical outcomes and reduces
costs compared to current practice
ā Then it DOMINATES current practice and should be recommended for use in
the NHS
ā¢ The intervention/product improves some clinical outcomes but not others
and reduces costs OR
ā¢ The intervention/product improves all/some clinical outcomes but
increases costs
ā Then an informal judgement needs to be made about the ācost-effectivenessā
and whether it is considered āvalue for moneyā
E.g. Ā£100 per stroke avoided vs. Ā£1,000,000 per stroke avoided
63. Measuring health gain
Type of analysis
Value of
resources
Value of health gain
Cost-utility Ā£ Combined index:
Quality Adjusted Life Years (QALY)
Cost-effectiveness Ā£
Single indicator:
Weight loss (kg),
blood glucose control (HbA1c)
deaths averted,
life years savedā¦
Cost-consequences Ā£ Multiple indicators
Comparative cost /
Cost minimisation
Ā£ None
Sometimes it is not possible or necessary to
estimate health gain and then a costing
analysis might be the most appropriate!
64. Evidencing the impact of a product
ā¢ For the main effects (health and resource use) you will need good
evidence (ideally from comparative studies):
ā E.g. reduction in major adverse events
ā¢ Other effects may be modelled using case series or national statistics
ā E.g. length of stay associated with the adverse event
ā E.g. proportion of the adverse events that are fatal
ā¢ Economic evaluation best practice
ā Baseline event rates from large case series
ā Relative treatment effects from pragmatic randomised controlled trials
ā¢ Where you use observational data for relative treatment effects, you
should attempt to control for differences in baseline confounding variables
using regression analysis
65. Hierarchy of evidence for relative treatment effectiveness
Meta-analysis of RCTs
RCTs
Cohort studies
Case series
Expert opinion
Lower risk of bias
Sometimes better
- Larger studies
- More generalisable
- Longer time horizon
Might be the best we can do
66. Sensitivity analysis
ā¢ Threshold analyses
ā How much of a change in a key outcomes is needed in order to make the
intervention cost saving/cost-effective
ā¢ One-way and n-way sensitivity analyses
ā Vary individual parameters within plausible ranges
ā Or to extremes
ā¢ Probabilistic sensitivity analyses
ā Vary all parameters simultaneously within plausible ranges
67. Conclusions
ā¢ Cost-utility analyses are the preferred method of economic evaluation for
NICE but are not always feasible
ā EQ-5D is the preferred health state valuation tool
ā¢ The results of cost-effectiveness analyses and cost-consequences analyses
require informal judgements to be made to determine whether the
product/intervention is considered āvalue for moneyā
ā¢ Comparative cost analyses do not estimate a value of health gain
ā¢ Observational evidence typically has a higher risk of bias but might be the
best we can do
ā¢ Sensitivity analysis should be conducted to deal with parameter
uncertainly
69. 3. Case Studies
NICE Medical Technologies guidance
ā¢ A technology is likely to be selected if:
ā relevant* technology with a CE marks (or equivalent regulatory approval) or is expected to get
one within 12 months
ā substantial benefits to patients or the health and care system compared with current practice
ā¢ easily understood, clearly described, plausible, supported by evidence.
ā developing guidance would mean faster and more consistent adoption of the technology.
ā¢ Relevance
ā Detailed and transparent
ā Lower threshold of evidence than medicines or guidelines
ā¢ QALYs not necessary
ā¢ Less sophisticated modelling
ā Approved by a national committee
* a medical device (under EC directive 2007/47/EC or 93/42/EEC)
an active medical device (under EC directive 90/385/EEC)
an active implantable medical device, (under EC directive 90/385/EEC)
an in vitro diagnostic medical device (under EC directive 98/79/EC).
70. Case study 1 ā SecurAcath for securing percutaneous catheters
https://www.nice.org.uk/guidance/mtg34
Published June 2017
71. PICO
ā¢ Population: Patients requiring peripherally inserted central
catheters (PICC)
ā¢ Intervention: SecurAcath ā a single-use device used to secure
percutaneous catheters in position on the skin
ā¢ Comparator: StatLock ā standard care
ā¢ Outcomes:
ā¢ NHS costs
ā¢ Catheter-related complications
72. Model essentials
ā¢ Economic evaluation type: Comparative cost analysis
ā¢ Model type: Simple decision tree
ā¢ Time horizon: 25 days
ā¢ Discounting: Not necessary - short time horizon
ā¢ Perspective: NHS hospital costs (England)
73. Technology costs
ā¢ StatLock ā standard care
ā Unit cost Ā£3.47
ā Needs replacing weekly
ā Nurse time for placement 40.8 minutes xĀ£0.60=Ā£24.48
ā Cost over 25 days=Ā£3.47x4+Ā£24.48=Ā£38.36
ā¢ SecurAcath
ā Unit cost Ā£16
ā Does not need replacing
ā Nurse time 20.5 minutes xĀ£0.60=Ā£12.30
ā Cost over 25 days =Ā£16+Ā£12.30=Ā£28.30
Faster
Better
74. Benefits
ā¢ Reduced need for PICC replacement
ā¢ Slightly reduced risk of catheter-related infection
Probability of
complication over 25
days Cost of
treating
complicationSecurAcath StatLock
PICC migration 0.0040 0.0593 Ā£250
PICC malposition 0.0166 0.1097 Ā£250
PICC occlusion 0.1435 0.1200 Ā£250
Thrombosis 0.0369 0.0369 Ā£250
Infection 0.0036 0.0037 Ā£9,900
Sources
Effects
1xRCT (n=105)
2x cohort studies
4x case series
Unit costs
3x published studies
Safer
75. Decision tree - SecurAcath
Ā£28
SecurAcath
PICC migration
PICC malposition
PICC occlusion
Thrombosis
Infection
0.0040
0.0166
0.1435
0.0369
0.0036
No complication
0.7954
Ā£278
Ā£278
Ā£278
Ā£278
Ā£9,928
Ā£114
Multiply the cost of each endpoint with the probability of getting there
(0.7954x28)+(0.0040x278)+(ā¦ā¦.. =
78. Other Analyses
ā¢ Sensitivity analyses did not change outcome
ā Device cost +/-20%
ā Complications +/-20% and no difference
ā Shorter SecurAcath placement time
ā¢ Other population - alternative time horizons
ā 5 days and 25 days
ā Threshold analysis ā SecurAcath is cost saving for PICC use over 15 days
ā¢ Other population - CVC instead of PICC
ā Alternative comparator ā sutures
ā SecurAcath was not cost saving
79. Recommendations
ā¢ āSecurAcath should be considered for any PICC with an anticipated
medium- to long-term dwell time (15 days or more).ā
ā¢ āEstimated cost savings range from Ā£9 to Ā£95 per patient for dwell times
of 25 days and 120 days, respectively.ā
ā¢ āAnnual savings across the NHS in England from using SecurAcath are
estimated to be a minimum of Ā£4.2 million.ā
81. Case study 2
ā¢ PleurX peritoneal catheter drainage system for vacuum-assisted drainage
of treatment-resistant, recurrent malignant ascites
Published March 2012
Model updated February 2018
https://www.nice.org.uk/guidance/mtg9
82. PICO
ā¢ Population: People with treatment resistant, recurrent malignant
ascites (accumulation of fluid in the peritoneal cavity)
ā¢ Intervention: PleurX peritoneal catheter drainage system for
vacuum-assisted drainage
ā¢ Comparator: Repeated large-volume paracentesis (needle drainage
of fluid) inpatient procedures
ā¢ Outcomes: Technical success, resolution of symptoms, perception
of body image, quality of life, adverse events, drainage
frequency, resource use, cost per patient
83. Model essentials
ā¢ Economic evaluation type: Comparative cost analysis
ā¢ Model type: Simple decision tree
ā¢ Time horizon: 26 weeks
ā¢ Discounting: Not necessary - short time horizon
ā¢ Perspective: NHS hospital costs (England)
84. Claimed Benefits
ā¢ Greater patient independence
ā¢ Better symptom control
ā¢ Reduced need for repeated large-volume
paracentesis procedures
ā¢ Resource savings through a reduced need for hospital
physician and nurse time, outpatient visits and
hospital bed days
Cheaper
Better
86. Clinical variables
Parameter LVP PleurX Source
Mean survival
(weeks)
8.45 8.45 Mullan et al
(2011a)
Parameter LVP PleurX Source
Probability of
infection (LVP)
4.5% 2.5% Rosenberg
(2004)
Probability of
catheter
failure (LVP)
3.0% 5.0% Rosenberg
(2004)
Survival
Complications
87. Healthcare recourse use
Parameter Value Source
Bed days for LVP per session 2.8 Mullan (2011a)
Frequency of repeated LVP (per month) 1.22 Mullan (2011a)
Large volume paracentesis
PleurX
Parameter Value Source
Bed days for catheter placement 1.0 Assumed based on Mullan
(2011a)
Probability of re-intervention 4.0% Rosenberg (2004)
Proportion who are self-managed 73.0% Courtney (2008)
Length of nurse visit (hours) 0.25 Assumed
Nurse visits for catheter use training 2 Questionnaire
Nurse visits per week 3.5 Assumed
Number of drainage kits used (per week) 3.5 6.4.5
88. Healthcare Costs
Parameter Value Source
Hospital bed day Ā£355.00 NHS reference cost 2015-
16
Infection Ā£198.97 NHS reference cost 2015-
16; BNF*
Catheter failure Ā£405.73 NHS reference cost 2015-
16, BNF**
Catheter re-intervention Ā£790.96 Assumed***
Cost per home visit (assisted) Ā£67.89 PSSRU 2016
Cost of travel per visit (assisted) Ā£1.58 Assumed
*Includes: A medical oncology consultant led first attendance visit:
Ā£197; 7 day course of antibiotics (Ciprofloxacin) Ā£1.97
**Includes: A medical oncology consultant led first attendance visit:
Ā£197; vial of Streptokinase: Ā£16.73; Ultrasound lasting <20 minutes:
Ā£51.00; contrast fluoroscopy lasting <20 minutes: Ā£141.00
*** Assumed to be cost of 1st catheter procedure + 1 hospital bed day
89. Costs of consumables
Large Volume Paracentesis
Parameter Value Source
Catheter and pack Ā£33.64 Uplifted from Mullan et al
Connector Ā£7.22 Uplifted from Mullan et al
Drain Ā£5.19 Uplifted from Mullan et al
2L Drainage Bag Ā£0.67 Uplifted from Mullan et al
Procedure costs/sundries Ā£127.21 Uplifted from Mullan et al
Parameter Value Source
Catheter and pack Ā£245.00 Provided by manufacturer
2L Drainage Bag and 1L drainage
kit
Ā£63.75 Provided by manufacturer
Procedure costs/sundries Ā£127.21 Uplifted from Mullan et al
Drainage kit box (10 units) Ā£637.50 Provided by manufacturer
PleurX
90. Key Model Assumptions
ā¢ Effects
ā No change in survival rate in both arms of the model
ā Drainage volume of 9.2 litres per procedure in patients who have large-
volume paracentesis
ā Average drainage volume of 3.5 litres per week using PleurX
ā¢ Resource use
ā Need for 2 nurse visits to train patients to self-manage drainage at home using
PleurX
ā Nurse visit length of 15 minutes for PleurX help with drainage
ā One nurse visit per litre of fluid drained using PleurX
ā Similar levels of treatment monitoring needs in both arms
91. Results
Intervention/Comparator Cost per patient
Inpatient large-volume paracentesis Ā£3,146
PleurX peritoneal catheter drainage system Ā£2,466
Savings = Ā£680 per patient when PleurX catheter
drainage system is used
92. Sensitivity Analysis
ā¢ One-way deterministic sensitivity analysis
ā¢ All variables were tested except population size
ā¢ Variables were analysed using 20% variance regardless of level of
confidence in an input
ā¢ Six key drivers were identified and subjected to further deterministic
threshold analysis to identify point at which PleurX became more costly or
cost saving
93. Key drivers
ā¢ Cost of a hospital bed day
ā¢ Number of bed days per LVP procedure per month
ā¢ Number of bed days for PleurX catheter placement
ā¢ Cost of drainage kit box (10 units)
ā¢ Number of drainage kits used per week per patient
94. PleurX became more costly compared to inpatient LVP when:
ā¢ the cost of an excess bed day is reduced to less than Ā£220 per day
ā¢ the frequency of an inpatient large-volume paracentesis procedure is
reduced to fewer than one per month
ā¢ the average length of inpatient stay after the LVP is decreased to 2.1
days
ā¢ the number of inpatient bed days following the PleurX catheter
insertion is increased to more than 3.1 days
ā¢ the cost of the PleurX drainage kit is increased to more than Ā£915
(per 10 units)
ā¢ more than 5.1 drainage kit units are needed per week
Findings of the Threshold Analyses
95. Recommendations
ā¢ āThe PleurX peritoneal catheter drainage system should be considered for
use in patients with treatment-resistant, recurrent malignant ascites.ā
97. Conclusion: Key things to remember
ā¢ Appropriate comparators should include usual care
ā¢ NHS cost perspective should not cover patient costs or productivity
gains/losses
ā¢ Find evidence for impact on health care resource use
ā¢ Unit costs from standard sources
ā¢ Appropriate time horizon to capture costs and benefits
ā¢ Accounting for time preference, through discounting/annuitizing
ā¢ Types of economic analysis, e.g. cost-consequences analysis
ā¢ Health outcomes to include
ā¢ Simple models, e.g. decision trees, can help
ā¢ Deal with uncertainty through sensitivity analysis
ā¢ Parameter estimates, sources, pathways and assumptions should be
transparent
98. Health economic evaluation resources
ā¢ Textbooks
ā Drummond et al Methods for the Economic Evaluation of Health Care Programmes
(4th edition) 2015
ā Briggs et al Decision Modelling for Health Economic Evaluation 2006
ā¢ Short courses
ā Oxford University ā one day
ā¢ https://www.herc.ox.ac.uk/herc-short-courses/introduction-to-health-
economic-evaluation
ā Brunel University ā 3 day
ā¢ https://www.brunel.ac.uk/research/Institutes/Institute-of-Environment-
Health-and-Societies/Health-Economics/Short-Courses
ā York University
ā¢ https://www.york.ac.uk/che/news/news-2018/che-short-courses/
ā¢ Good practice guides
ā https://www.ispor.org/workpaper/practices_index.asp
100. Practical
ā¢ Costs
ā¢ What are the technology costs?
ā¢ Are there upfront costs? Can you calculate cost per patient?
ā¢ Cost savings? (Faster, better, cheaper, safer)
ā What is the evidence?
ā¢ What cost perspective? Who is the evaluation for?
ā¢ What unit costs?
ā¢ Health gain
ā¢ Does it improve health?
ā What is the evidence?
ā¢ What kind of health economic analysis is required? (e.g. cost-
consequences analysis)
108. Repeat for each intervention and calculate ICER
A B Difference
Expected cost Ā£1,394,575 Ā£2,250,404 Ā£855,830
Expected QALYs 9,286 9,345 59
ICER (Ā£ per QALY) = Ā£14,466
5%
1%
75%
5%
Ā£100 pa
QoL=1
Ā£0 pa
QoL=0
Ā£1,000 pa
QoL=0.6
Intervention A
4%
1%
78%
5%
Ā£200 pa
QoL=1
Ā£0 pa
QoL=0
Ā£1,100 pa
QoL=0.6
Intervention B
109. Limitations
ā¢ Must assume that each āstateā is mutually exclusive, i.e. For one person,
they cannot be in both states at once
ā¢ Models usually have no āmemoryā:
ā Feasible that one person in the model could have multiple events i.e. 6 heart
attacks
ā Assumes that transition values, costs and QALYs are the same (so if you have
the 1st heart attack, same costs and QALYs as 7th)
ā There are ways to solve this issue
111. Cirrhosismodelsstructure F3-TN F3-FP
Comp-
TP*
Comp-
FN*
Var-
Un*
VarTP-
Pr*
Var-
FN*
Decomp
*
Bleed
Trans
2
dcVar
Un*
dcVar
Pr*
Post -
Trans
Dead
i) Shaded states relate to test results
ii) States with an asterisk have a corresponding hepatocellular carcinoma (HCC) 5
year state attached to them (not shown)
iii) From an HCC state patients can either die or get a transplant Non-transplant HCC
survivors either return to their state of origin or if that was a FN to the
corresponding TP state (not shown)
iv) All states transit to death
v) Dashed arrows represent the additional transitions during a cirrhosis retest cycle
112. Steatosis model structure
<F3ā<F3
<F3 TN*
<F3 ā F3
<F3 FP
F3 - F3
F3 TN*
F3 - <F3
F3 FN
F4 ā F3
Comp FN
F4 ā F4
Comp TP
F3 ā F4
F3 FP
<F3 ā F4
<F3 FPc
Cirr test
Cirr test +
Progress
Fib test
i) The first component of the first name depicts the true health state
whereas the second is the fib/cirr test results
e.g. F3 ā <F3 = patients with advanced fibrosis identified by the test
as not having advanced fibrosis. (The second name underneath is as it
appears in the spreadsheet).
ii) All states transit to dead
iii) F4 states transit to cancer and decompensated cirrhosis; Varices
states transit to bleed. See NAFLD cirrhosis model
<F3 as true state F3 as true state F4 as true state Retesting for advanced fibrosis
F4 ā <F3
Comp FN+
Cirr test
Cirr test +
Progress
F4V ā F3
Var FN
F4Vā F4 pr
Var TPpr
Cirr test +
Progress
F4Vā <F3
Var FN+
Cirr test F4V ā F4
VarTPun
Var test + progress
Fib test +
Progress
Fib test +
Progress
Fib test +
Progress
Fib test
Fib test
Fib test
Cirr test
Fib test +
regress
Var test
Non-<F3
Non-FP
<F3-Non
<F3-FN
F3-Non
F3FN+
F4-Non
CompFN++
Non-Non
Non-TN
F4V ā Non
Var FN++
Steat test
Non-F3
Non-FPf3
Steat test
Steat test
Steat test
Steat test
Steat test +
Progress
Steat test +
Progress
Steat test +
Progress
Steat test +
Progress
Fib test +
Progress
Fib test
No NAFLDas true state
Steat test +
regress
Cirr test +
regress
Cirr test +
Progress
Editor's Notes
Delete graphics if preferred
Delete graphics if preferred
This definition of economic evaluation (from Drummond, Stoddart & Torrance) has two key aspects:
EE should always compare one health care intervention with one or more alternative interventions for the same population group.
EE should include both the costs and consequences of interventions - the resources that they consume and the health outcomes that they produce.
Note that the choice of comparator is crucial ā should include all relevant options for a group of patients (including ādo nothingā and ācurrent practiceā).
Analysis should be conducted separately for each subgroup of patients.
Different types of economic evaluation are defined by the choice of outcome unit.
CUA is the preferred option for NICE ā but sometimes QALY estimation is difficult, so CEA may be used.
Most likely that they will want to conduct cost-consequences analyses
We will come back to this concept later on in the session!!!!!!!
Other examples
Mobile ultrasound (MDI medical) iPhone size device with DICOM image standards. Need to model benefit of less hospital referrals and more GP practice/home based scans for a wide range of conditions not just pregnancy
Point of care testing (eBiogen) Point of care rapid (nearly real time) testing of lactate (and other) analytes in blood. Lactate is a strong indicator of sepsis
Pick out the cost of an average GP appointment from the document
The ICER is illustrated by the slope of a line through the origin and the IE/IC point for an intervention.
Compare this with the āthresholdā cost-effectiveness ratio ā assume around Ā£20,000 to Ā£30,000 per QALY for NICE decisions.
SOMETIMEES A SIMPLE COSTING MIGHT BE THE MOST APPROPERIATE
Clinical outcomes
technical success of catheter insertion and drainage procedure
resolution of symptoms (bloating, nausea, acid reflux, reduced appetite, negative
perception of body image and resulting psychological distress)
quality of life outcomes
adverse events (catheter site infections, peritonitis, catheter occlusion, and haemorrhage or bowel perforation when the device is inserted)
drainage frequency
resource use outcomes, for example re-admission rates, re-interventions and duration of hospital stay.
Cost model submitted by the sponsor
Malignant ascites is a sign of peritoneal carcinomatosis, the presence of malignant cells in the peritoneal cavity. While survival in this patient population is poor, averaging about 20 weeks from time of diagnosis, quality of life can be improved through palliative procedures
The conventional management of treatment-resistant, recurrent malignant ascites involves repeated large-volume paracentesis (LVP) procedures that are carried out in hospital. Most commonly this is done as an inpatient procedure, although some centres are able to offer paracentesis as a day-case procedure.
Inpatient paracentesis is carried out when patients have developed troublesome symptoms from recurrent ascites. This can entail someresult in delay while waiting for admission, during which the patient continues to experience symptoms.
The PleurX peritoneal catheter drainage system (UK Medical Ltd) is designed to remain in place indefinitely and patients and carers are trained to perform fluid drainage when needed by attaching the vacuum bottle to the catheter. The use of the PleurX peritoneal catheter drainage system may allow greater patient independence, and lead to resource savings through a reduced need for repeated LVP procedures and hospital bed days.
Clinical outcomes
technical success of catheter insertion and drainage procedure
resolution of symptoms (bloating, nausea, acid reflux, reduced appetite, negative
perception of body image and resulting psychological distress)
quality of life outcomes
adverse events (catheter site infections, peritonitis, catheter occlusion, and haemorrhage or bowel perforation when the device is inserted)
drainage frequency
resource use outcomes, for example re-admission rates, re-interventions and duration of hospital stay.
Give brief detail on the studies Rosenburg and Mullan
Rosenberg (2004) was the one comparative case series study
Limitation is that variables were analysed using a 20% variance regardless of level of confidence in an output
Management with PleurX may result n cost savings of Ā£679 per patient when compared with inpatient large-volume paracentesis. 7.4 hospital bed days saved per patient but 23.5 more community nurse visits needed
Key drivers
Cost of a hospital bed day
Number of bed days per LVP procedure per month
Number of bed days for PleurX catheter placement
Cost of drainage kit box (10 units)
Number of drainage kits used per week per patient
Cost savings of PleurX are heavily dependent on a reduction in hospital stay.
The health economic information booklet explains key concepts and HE methods
NICE run free introductory workshops on health economic methods. Please let us know if you would like to attend one of these
Complete feedback forms
1) Markov models are based on a series of health states that a patient can occupy at a given point in time.
2) The arrows represent possible transitions:
We allocate members of a population to one of a finite number of āstatesā (e.g. āwellā, āsickā and ādeadā).
We then specify the probabilities of moving between the states in consecutive time periods.
3) The probability of a patient occupying a particular state is assessed over a series of time periods called cycles ā need to choose the appropriate cycle length, this which will depend on the disease and interventions.
NB Death is a āabsorbingā state - there is no way out of it!
Now that we have the structure set up we can add data:
1) For example the probabilities of transitioning between the different states have been defined.
2) Also we need to attach costs and QoL values to each of the states.
So just like in the decision tree, these will be used to calculate the expected value.
SO letās say our cycles are yearly and we have assumed a time horizon of 10 years.
Therefore each year patients can move between different health states, which health state they move to depends on the transition probabilities.
Given that we now know how many patients are in each state per cycle, It is a simple case of multiplying the number of patients in each state by the costs and QALYs of each state.
We then sum the costs of each cycle and the QALYs, this will then give us the total costs and total QALYs for each intervention.
In state transition models:
We allocate members of a population to one of a finite number of āstatesā (e.g. āwellā, āsickā and ādeadā).
We then specify the probabilities of moving between the states in consecutive time periods.
NB Death is a āsinkā state - there is no way out of it!
In state transition models:
We allocate members of a population to one of a finite number of āstatesā (e.g. āwellā, āsickā and ādeadā).
We then specify the probabilities of moving between the states in consecutive time periods.
NB Death is a āsinkā state - there is no way out of it!
If one person has multiple events, then the same costs and QALYs are applied if you had the first event or a recurrent event.
Going back to our previous example ā a person could get sick multiple times, however they would still have the same probability of going into that sick state, as well as the same cost and same QALY assigned to that state ā however in reality, someone who has their 7th heart attack for example may have a different decrease in their QoL compared to someone who had their 1st heart attack.
In summary when someone in a model has an adverse event, the model doesnāt know if they have had any previous adverse events or what state they have come from ā which is why it is memoryless.