A Model to Assess the Cost-Effectiveness of Early Intervention

Early Intervention Services:
The Economic Case
Paul McCrone,1 A-La Park,2 Martin Knapp1,2
1 Institute
of Psychiatry, King’s College London,
2 PSSRU, London School of Economics
Background
• Deinstutionalisation in UK started in 1980s and is
largely complete
• Community mental health teams (CMHTs)
developed often using case-management
techniques
• In 2001 the government stating that specialist teams
should be provided throughout England
– Assertive community treatment (ACT)
– Crisis resolution (CRT)
– Early intervention (EI)
• Are these services a good investment?
Early Intervention in Psychosis
Services
• Intervening early is encouraged in other clinical areas
(e.g. cancer, heart disease)
• Onset of psychosis frequently not recognised
• Duration of untreated psychosis (DUP) can be up to 2
years
• Longer DUP is associated with poorer outcome
• EI services provide rapid care using a multidisciplinary
team approach
• Varied interventions
– Medication
– Psychological therapies
– Vocational support
• EI is generally time limited (around 3 years in England)
Why Consider Cost-Effectiveness?
• Increasing number of studies evaluating
EI services
• New services clearly require scarce
resources and therefore economic
evaluation is essential
• Are the extra costs of EI offset by
reduced costs elsewhere in the system?
• Is EI cost-effective?
Interpretation of Results from Economic
Evaluations
Outcomes
Higher
‘Equal’
Lower
Worse
‘Equal’
Better
N
N
N
?
?
Y
?
Y
Y
What Type of Evidence?
Randomised
controlled trials
Long-term follow-up
observational studies
Decision models
Decision Models
• A way of assessing costs and cost-effectiveness
• Alternative or supplementary to trial
• Advantages:
– Results can be produced quickly
– Models can be adapted to aid generalisability
– Allows a focus on certain key parameters of interest
• Disadvantages
– Models are by definition an abstraction from reality
– Data are required for probabilities and costs and these are
not always available
Initial Model
Base Case Model (EI subtree)
Base Case Model (SC subtree)
Data Required for Model
• Probabilities
– clinical trials (LEO)
– audit data (Worcestershire and Northumberland EI
services)
– routine data (28-day readmission rates)
– expert judgement
• Costs
– existing economic studies of EI
– economic studies in other areas
– non-economic studies
Base Case Data: Probabilities
Parameter
EI
Standard
Care
Formal admission (first cycle)
0.23
0.44
Informal admission (first cycle)
0.25
0.23
Discharge to CMHTs (all cycles)
0.10
NA
Remain with EI team (first cycle)
0.42 (D)
NA
Formal admission (subsequent cycles)
0.06
0.13
Informal admission (subsequent cycles)
0.06
0.07
CMHT treatment (subsequent cycles)
NA
0.80 (D)
Remain with EI team (subsequent cycles)
0.78
NA
D = default probability
Base Case Data: Costs
Parameter
Cost
EI input over 2 months
£388
Standard community services over 2 months
£233
Formal admission (61 days)
£10492
Informal admission (33 days)
£5871
Base-Case 1-Year Costs
£14394
Expected costs (£s)
15000
10000
£9422
5000
0
EI
Standard care
Sensitivity Analyses (1)
• Key parameters increased/decreased by
50%
– probability of initial formal admission
– probability of initial informal admission
– probability of readmission
– probability of remaining with EI team/CMHT
Expected 1-year cost (£s)
Sensitivity Analyses: Results (1)
16000
Std care
14000
12000
10000
EI
8000
6000
4000
2000
0
Low est
High est
Prob
formal
adm (EI)
Prob
infomal
adm (EI)
RR (EI)
Prob
formal
adm (std
care)
Prob
infomal
adm (std
care)
RR (std
care)
Sensitivity Analyses (2)
• Probabilistic sensitivity analysis
– all parameters varied simultaneously
– Monte Carlo analysis
– data drawn from parameter distributions
– 100,000 resamples
– cost distributions generated
1-year cost (£s)
0.06
25
70
0
26
70
0
23
70
0
24
70
0
21
70
0
22
70
0
19
70
0
20
70
0
17
70
0
18
70
0
15
70
0
16
70
0
13
70
0
14
70
0
11
70
0
12
70
0
97
00
10
70
0
87
00
77
00
67
00
57
00
47
00
37
00
Probability
Probabilistic Sensitivity Analyses
(1-Year costs)
0.12
0.1
0.08
EI
SC
0.04
0.02
0
Impact of EI on Vocational
Outcomes
Vocational Model: Structure
Vocational Model: Parameters
Parameter
Employment
Education
Not economically active
Full employment (if employed)
Wage rate
Lost productivity costs/year
EI
SC
0.36
0.27
0.20
0.065
0.44
0.67
0.58
0.52
£5.80
£9744
Sources: Garety et al, 2006; Perkins & Rinaldi, 2002; Major et al, 2010
Vocational Model: Results
8000
7111
Annual cost (2008/9 £s)
7000
6000
5024
5000
4000
3000
2000
1000
0
EI
SC
Homicide Model: Structure
Homicide Model: Parameters
Parameter
Homicide rate
Lifetime cost of homicide
physical and emotional
lost productivity
service costs
Annual cost of homicide
year 1
subsequent years
EI
SC
0.011% 0.17%
£1.72 million
59%
31%
10%
Sources: Nielssen & Large, 2008; Home Office, 2004
£54,079
£50,260
Homicide Model: Results
100
92
90
Annual cost (2008/9 £s)
80
70
60
50
40
30
20
10
6
0
EI
SC
Suicide Model: Structure
Suicide Model: Parameters
Parameter
Suicide rate
Lifetime cost of suicide
physical and emotional
lost productivity
service costs
Annual cost of suicide
year 1
subsequent years
EI
SC
1.3%
4.0%
£1.6 million
69%
29%
3%
£34,412
£33,442
Sources: Melle et al, 2006; Robinson et al, 2010, McDaid & Park, 2010; Platt et al, 2006
Suicide Model: Results
1600
1376
Annual cost (2008/9 £s)
1400
1200
1000
800
600
459
400
200
0
EI
SC
Summary of Savings
Per person
Services
Year 1
Years 2-5
Year 6-10
(£)
(£)
(£)
-5,777
Productivity
Intangibles
Total
By sector
NHS
-5,777
(£m)
-2,408
-60
-2,052
-1,912
-314
-628
-4,774
-2,600
(£m)
(£m)
-39.1
-16.0
0
-0.8
-0.6
-0.4
Productivity
0
-14.2
-13.2
Intangible
0
-2.2
-4.3
-39.9
-32.9
-17.9
Other public sector
Total
Long-Term Model
Scenarios for Long-Term Model
• Scenario 1. Readmission rates are constant
throughout all the 48 cycles for both EI (12%)
and standard care (20%).
• Scenario 2. Readmission rates for EI for the
first three years are constant, and then
suddenly become the same as for standard
care.
• Scenario 3. Readmission rates for EI after
three years gradually become similar to those
for standard care.
Eight Year Costs of EI and SC
120000
£36,632
£17,427
£27,029
Cost (£s)
100000
80000
EI
SC
60000
40000
20000
0
Scenario 1
Scenario 2
Scenario 3
Cost-Effectiveness of EI:
The LEO Study
Craig et al (2004) BMJ 329: 1067
Garety et al (2006) Br J Psychiatry 188: 37-45
McCrone et al (2010) Br J Psychiatry 96: 377-382
Methods (1)
• Lambeth Early Onset (LEO) service
• Deprived area of inner-London
• For first episode psychosis or those for with second
episode where care was never received
• Patients identified by screening for possible
psychosis
• Randomised controlled trial conducted including 144
patients (71 to EI, 73 to standard care)
• Assessments at baseline, 6 months and 18 months
• Primary outcome measure was relapse and
hospitalisation
Methods (2)
• EI
– Provided ACT
– Focus on maximising engagement, psychosocial
recovery and relapse prevention
– 10 staff members (psychiatrists, psychologists,
occupational therapists, nurses, healthcare
assistants)
– Interventions included low-dose medication, CBT,
family therapy and vocational rehabilitation
• SC (standard care)
– CMHTs with no extra training in dealing with first
episode psychosis
Methods (3)
• 6-month service use measured at each
assessment with CSRI
• Data on hospital admissions available
for entire follow-up period
• Service use data combined with unit
costs
• Cost-effectiveness analysis used
vocational recovery and quality of life
data
Sample
71 randomised to EI and 73 to SC
Mean age: EI 26 years, SC 27 years
Men: EI 55%, SC 74%
First episode: EI 86%, SC 71%
BME: EI 62%, SC 75%
Employment: EI 19%, SC 18%
Schizophrenia: EI 72%, SC 67%
Inpatient Days
60
54.9
52.3
50
44.0
40
35.5
EI
SC
30
20
10
0
Baseline
18m FU
P
ay
O
T
M
ed
Ph care
ys
ic
al
i
R
es p
ca
re
Po
lic
e
D
Ps
yc
h
O
th
do
Ps c
yc
ho
l
H
C
A
C
ou
ns
So
c
w
C
M
HN
G
Use of Services 0-6 months
100
90
80
70
60
% 50
40
EI
SC
30
20
10
0
P
ay
O
T
M
ed
Ph care
ys
ic
al
i
R
es p
ca
re
Po
lic
e
D
Ps
yc
h
O
th
do
Ps c
yc
ho
l
H
C
A
C
ou
ns
So
c
w
C
M
HN
G
Use of Services 12-18 months
80
70
60
50
% 40
EI
SC
30
20
10
0
Inpatient Use and Costs (2003/4 £s) at
Baseline and 18-Month Follow-Up
EI
SC
Inpatient days
52.3
44.0
Inpatient costs
8989
7573
Total costs
9747
8256
Inpatient days
35.5
54.9
Inpatient costs
6103
9442
Other costs
5332
4544
Total costs
11685
14062
Baseline
18-month follow-up
95% CI of cost difference -£8128 to £3326)
Outcomes
Vocational recovery at 18m FU:
EI 33%, SC 21% (p = 0.162)
Quality of life (MANSA): EI 59.3, SC 53.3
(p = 0.025)
EI was dominant – lower costs and better outcomes
Cost-Effectiveness Acceptability Curve 1
Probability that LEO intervention is more cost-effective than routine care
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1000
2000
3000
4000
5000
6000
7000
Willingness to pay for full vocational recovery by 18 months (£s)
8000
9000
10000
Cost-Effectiveness Acceptability Curve 2
Probability that specialised care is more cost-effective than usual care
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
20
40
60
80
100
120
140
Willingnes to pay for one-unit improvement in quality of life (£s)
160
180
200
Conclusions from LEO
Study
• EI resulted in reduced inpatient use
• Costs were lower for EI (although not
significantly)
• When combined with outcomes, EI is
very likely to be cost-effective
Summary
• Initial model has demonstrated savings in
care costs for EI compared to SC
• Large savings due to increased employment
• Small savings due to reduced homicide and
suicide
• Long-term cost savings depend on
convergence in readmission rates
• LEO study revealed lower costs, better
outcomes and (therefore) cost-effectiveness
How do findings compare with
those from other studies?
• Australia - savings of $AUD 7110 (Mihalopoulos et al,
1999)
• Long-term savings of $AUD 6058 (Mihalopoulos et al,
2009)
• Canada – EI $2371, SC $2125 (Goldberg et al, 2006)
• England – 54% fewer bed days (Dodgson et al, 2008)
• Norway & Denmark – weeks in hospital EI 16.4, SC 15.5
(Larsen et al, 2006)
• Denmark – inpatient days in year 1 EI 62, SC 79; year 2
EI 27, SC 35; years 3-5 EI 58, SC 71 (Petersen et al,
2005; Bertelsen et al, 2008)
• Norway – admissions EI 33%, SC 50% (Grawe et al,
2006)
• Sweden – cost savings of 29% year 1, 55% year 2, 61%
year 3 (Cullberg et al, 2006)
Acknowledgements
•
•
•
•
•
•
•
•
•
•
Mike Clark
David Shiers
Swaran Singh
Jo Smith
Tom Craig
Philippa Garety
David McDaid
Other steering group members
IOP/LSE colleagues
DH for funding programme