d6 isaranuwatchai cadth 12_apr16
TRANSCRIPT
YOU ARE
HERE
THINKING OUTSIDE RCTs
Wanrudee Isaranuwatchai, PhDCADTH Symposium
12 April 2016
Marcus Tan, MDKate Butler
Tony Zhong, MDJeffrey S. Hoch, PhD
RCTs
How a person-level cost-effectiveness analysis from an observational study can show
“value for money” of a health intervention?
No conflict of interest that may affect this presentation
Outline
Observational studies Net benefit regression (NBR) Case study of breast reconstruction procedures Summary
Observational Studies
Effectiveness Validity Affordability
Lack of randomization Confounding Selection bias …
Concato, Shah, Horwitz. NEJM, 2000; 342; 1887-1892
Propensity score matching Instrumental variable
Multivariate regression
Net Benefit Regression
Compared to UC, is TX cost-effective?
ICER = ΔC/ΔE Cost-effective = ICER < λ
ΔC/ΔE < λ ΔC < λΔE 0 < λΔE – ΔC
Cost-effective = INB > 0
(= INB)
How to get to INB: Data
NBi = λEi – Ci
Hoch et al. Health Economics, 2002; 11: 415-430
The regression in NBR – SLR
NBi = β0 + β1(TX)i + εi
β1 = NBTX – NBUC
= ΔNB = INB
The regression in NBR – MLR
NBi = β0 + β1(TX)i + βj(X)i,j + εi
β1 = INB = NBTX – NBUC Adjusted for X
Cost-effective: ICER < λ or INB > 0 β1 = INB
p
j = 1
Outline
Observational studies Net benefit regression (NBR) Case study of breast reconstruction procedures Summary
Case Study
Population All women receiving either DIEP or MS-TRAM between 2008 and 2012 in one hospital
Intervention Deep inferior epigastric perforator (DIEP) flaps
Comparator Muscle-sparing transverse rectus abdominis myocutaneous (MS-TRAM) flaps
Outcome Patient-reported satisfaction with outcome (BREAST-Q)
Perspective Hospital
Costs Operating room; hospital staysMedication; allied health careMedical imaging; overhead
Time horizon 2 years
Objective
To compare the cost and outcome of DIEP flap to MS-TRAM flap in autologous breast reconstruction from the hospital perspective
Statistical Analysis
Descriptive analysis
Net benefit regression Adjusted for age, chemotherapy, radiation, laterality,
timing, income, and ethnicity
Cost-effectiveness acceptability curve (CEAC)
Descriptive AnalysisTX (N = 180) UC (N = 47)
Age ± SD 50.3 ± 8.7 52.7 ± 8.6Chemotherapy 106 (59%) 29 (62%)Radiation therapy 89 (49%) 30 (64%)Household income Low ($0 - $39,999) Medium ($40K - $99,999) High (≥ $100,000)
23 (13%)68 (38%)89 (49%)
7 (15%)12 (25%)28 (60%)
Ethnicity Caucasian Others
139 (77%)41 (23%)
40 (85%)7 (15%)
Laterality 101 (56%) 16 (34%)Timing 94 (52%) 29 (62%)Cost ± SD* $15,344 ± $4,728 $16,681 ± $4,289Satisfaction with outcome ± SD 73.8 ± 27.9 71.0 ± 28.6
Net Benefit Regression
Willingness-to-pay(for 1 unit of outcome)
INB(Incremental Net Benefit)
λ = $0 $575
λ = $1,000 $2,924
λ = $5,000 $12,320
λ = $10,000 $24,066
λ = $50,000 $118,029
NB from λ of $0 to $50,000 for each patient Run regression models for each λ
CEAC
$0 $1,000 $5,000 $10,000 $50,000 0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Willingness-to-pay for 1 more unit of outcome
Prob
abili
ty th
at D
IEP
is co
st-e
ffecti
ve
Discussion
Compared to MS-TRAM, DIEP could be an economically attractive option λ from $0 to $50,000: INB > 0 p(TX=CE) ~ 70%
Limitations Single site One perspective One outcome Unknown confounders
THANK [email protected]
-300
0-2
000
-100
00
1000
2000
delta
_c
-20 -10 0 10 20delta_e
Bootstrap adjusted delta C and delta E