The complexity of the allocation
problem in health care:
can current decision rules provide
a useful guide?
Karl Claxton
Questions of fact and questions of value?
• When costs displace health (∆ch)
ch
cc
h


0
k
v
Health
gained
Health
forgone
• Costs fall on both
c
h
 h
k
ch
k
h
v.h  cc  0, or
cc
v
h
Consumption
forgone
• When costs displace consumption (∆cc)
ch
cc
h


0
k
v
Health
forgone
v
v.h  ch  0, or
k
Consumption
forgone

cc
v
0
k
ch  cc
v
 0,
k or
v.h  chv cc 
k h
Fact : k = how much health displaced by increased HCS costs?
Value: v = how much consumption should we give up for health?
Question of value
Claxton et al, 2010 , 2011 and Paulden 2011
• Specify a complete and legitimate SWF?
– v is the measure of social welfare and presupposes a complete SWF
• Health and consumption are the only arguments
• or separable from other arguments
– k is simply an inefficient nuisance preventing welfare maximisation
• Complete and legitimate specification of SW not possible?
– Trade-offs still need to be and are made
– Legitimate social process reveals something about a latent welfare function
– Interpret shadow prices as revealed but partial expression of social value
• k is a revealed expression of social value of health from collective health care
• v is how much of their consumption individuals are willing to give up to improve
their own health
– So good reasons why k ≠ v
– Good reasons to suppose there are other non separable arguments
What it is and what its not
An efficient NHS
Health
1/k1
Underestimate health effect of ∆B
(i.e., k1 is too high)
Current NHS
H1
1/k1
B1
Average productivity would
overestimate health effect of ∆B
(i.e., H1/B1 < k1)
Budget
A scientific question of fact
Martin et al 2008, 2009 and MRC/NIHR 2012
• Previously
– Variations in expenditure and outcomes within programmes
– Reflect what actually happens in the NHS by PBC
Cancer
Circulation
04/05 per LY
£13,137
£7,979
05/06 per LY
£13,931
£8,426
Respiratory
Gastro-int
£7,397
£18,999
• Need estimate the overall threshold:
–
–
–
–
–
–
How changes in overall expenditure gets allocated across all the programmes
How changes in mortality might translate into QALYs gained
More (all) programmes (types of QALYs displaced)
How uncertain is any overall estimate
How it changes with scale of expenditure change
How it changes over time
Budgetary policies and available actions
Hard constraints with uncertain and variable costs and outcomes
•
–
Corner solutions or exogenous parameters
Model budget, policy, information revealed and available actions
•
–
–
Current rules special case of soft constraint
No simple ex-ante rules – more cost-effective if hard constraint
not meet budget at expectation or maximise expected health outcomes
•
6800
5000
Soft budget constraint
6600
4900
Hard budget constraint
Expected total health benefits
Expected total health benefits
Chalabi et al 2008, and
McKenna et al 2010
Complete flexibility
6400
Restrictive action
6200
6000
5800
5600
5400
4800
4700
4600
4500
4400
5200
4300
5000
4200
15
16
17
18
19
20
21
Budget (£ million)
22
23
24
25
9
10
11
12
13
14
15
16
Notional budget (£ million), actual budget = £12m
17
18
Implications for the value of research
How much budget give up to resolve uncertainty?
•
Underestimate value (hard constraints and less available actions)
Overestimate (soft constraint)
–
–
EVI based on current decision rules are a special case
•
Soft constraint and buy as much health as you like at a constant rate (k)
–
Variability and uncertainty matters
•
Approval and research decisions
–
£3,000,000
£10,000,000
Standard EVPI
Hard budget constraint
£9,000,000
£2,500,000
Restrictive action
EVPI in monetary terms
£8,000,000
EVPI in monetary terms
Chalabi et al 2008, and
McKenna et al 2010
£7,000,000
£6,000,000
£5,000,000
£4,000,000
£3,000,000
£2,000,000
Soft budget constraint
£2,000,000
£1,500,000
£1,000,000
£500,000
£1,000,000
£0
£0
0
5
10
15
20
25
30
Budget (£ million)
35
40
45
50
0
5
10
15
20
25
Budget (£ million)
30
35
40
McKenna and Claxton 2011
and MRC/NIHR 2011
Irrecoverable (opportunity) costs
•
•
Irrecoverable per patient treatment costs (NHE profile)
Irrecoverable costs allocated over time (e.g., capital costs of equipment)
Cumulative incremental NHE at population level for EECP, QALY
5,000
0
0
5
10
15
20
25
30
35
40
45
-5,000
Capital cost spread over 10 years
Capital cost incurred in year 1
-10,000
-15,000
-20,000
Technology time horizon
-25,000
Time, years
50
Griffin et al 2011 and
MRC/NIHR 2011
Irrecoverable (opportunity) costs
Research is not possible with approval (incentives and ethics)
Irrecoverable opportunity cost (value of information forgone)
•
•
0.0
Probability that research is conducted
Sufficient condition for Approve
0.2
0.4
Necessary condition for OIR
0.6
4-year design
3-year design
0.8
2-year design
1-year design
1.0
0
1
2
3
4
5
6
Time for research to report, years
7
8
9
Up a creek
Where
doeswithout
this leave
a paddle?
us?
• Cant fully specify SWF anyway
– At best partial reflection of social value (cant claim efficient/optimal)
– Contribute to accountable decisions and progressive change
• No ‘optimal’ simple ex-ante rules
– Depends on budget, policy, what is revealed and when, and remedial
actions available
– Problem of second best (problem for traditional CBA as CEA)
– Understand the limitations and implications
• Account for irrecoverable opportunity costs (price thresholds)
– Reject to approve (only relevant if no uncerinty)
– Reject to OIR, OIR to Approve (research not possible)
– Reject to OIR, OIR to AWR, and AWR to Approve