1 a cost-effectiveness framework for profiling hospital efficiency justin timbie academyhealth...
TRANSCRIPT
1
A cost-effectiveness framework for profiling hospital efficiency
Justin Timbie
AcademyHealth Annual Research Meeting
June 5, 2007
Walt Disney World “Dolphin”
2
Acknowledgements
Sharon-Lise Normand1,2
Joe Newhouse1
Meredith Rosenthal3
1Department of Health Care Policy, Harvard Medical School2Department of Biostatistics, Harvard School of Public
Health3Department of Health Policy & Management, Harvard
School of Public Health
3
Context
• Interest in efficiency measurement following growth of P4P.– 42% of commercial HMOs use cost information
(Rosenthal, 2005)
• DRA of 2005 requires Medicare to implement value based purchasing for hospital services by FY’09.– Efficiency measures to be included in FY’10-11.
• Measuring appropriateness and efficiency are both challenges.
4
Examples of efficiency metrics
• Dartmouth Atlas: population-based efficiency:– Medicare spending (last 2 years of life)– Resource inputs: beds, physician FTE inputs – Utilization: hospital/ICU days, physician visits
• Leapfrog Group: risk-adjusted LOS, readmission rates within 14 days.
• National Quality Forum: focusing on LOS and readmission.
• Medicare: MEDPAC considering publicly reporting hospital readmission rates.
5
Measurement challenges
• Defining efficiency: Focus on payment or resource use (LOS, readmission rates, RVUs).
– DRG-based payment makes hospital efficiency profiling different.
– Limited ability to measure inpatient resource use.
• Duration of efficiency, quality measurement.
– Longer duration is desired.
– Causes attribution difficulties (PAC providers).
• Weighting of cost vs. quality.
– Binary (threshold) scoring approaches weight domains equally.
– Measuring performance continuously allows tradeoffs.
6
Study design
• Objective: Compare efficiency of hospital care following acute myocardial infarction (AMI).
• Motivation: Channeling patients to high-value hospitals for specific conditions.
• Outcomes: In-hospital survival, hospital costs. • Data source: Massachusetts all payer data.
– 69 hospitals (11,259 patients) in FY’03.
Efficiency = Health benefit relative to cost
7
Methods - Cost measurement
• Used total hospital charges and global cost-to-charge ratios. – Costs derived from charge data remove price
variation.– Use of global cost-to-charge ratios may confound
estimates due to differential markup across revenue centers.
• Used in-hospital outcomes, although 30-day outcomes are preferred.
• Lacking post-acute care costs, costs of procedures.
8
Methods - Estimation
• Link inter-hospital transfers to create inpatient “episodes.”
• Estimate “predicted” outcomes.– Fit hierarchical models.– Condition on hospital-specific effect, risk factors.
• Estimate “expected” outcomes.– Condition on population mean effect, risk factors.
9
Methods - Combining measures
• Incremental outcomes:
ΔEi = Predicted survivali – Expected survivaliΔCi = Predicted costi – Expected costi
• Incremental Net Health Benefits (INHB):
• Estimate P(INHB > 0)• Identify efficient hospitals using relative or absolute
threshold.
INHBi = ΔEi – ΔCi/
where = WTP/ΔE = $5M/Life saved
10
15000 20000 25000 30000 35000
87
88
89
90
91
92
93
94
Standardized Cost (Dollars)
Sta
nd
ard
ize
d S
urv
iva
l (%
)
91.34
17,846
Results – Threshold Scoring
Standardized Cost (dollars)
Sta
nd
ard
ized
Su
rviv
al (
%)
15,000 20,000 25,000 30,000 35,000
88
9
0
9
2
94
)YY,YP(Y (C)(C)(S)(S)
)YY,YP(Y (C)(C)(S)(S)
1115000 20000 25000 30000 35000
87
88
89
90
91
92
93
94
Standardized Cost (Dollars)
Sta
nd
ard
ize
d S
urv
iva
l (%
)
91.34
17,846
Results - Cost-effectiveness
Standardized Cost (dollars)
Sta
nd
ard
ized
Su
rviv
al (
%)
15,000 20,000 25,000 30,000 35,000
88
9
0
9
2
94
12
Sensitivity of INHB estimates to 0
.00
.20
.40
.60
.81
.0
Willingness to Pay (Million $/Life Saved)
P(I
NH
B >
0)
0 1 2 3 4 5
Willingness to Pay Threshold
(Million $/Life Saved)
0 1 2 3 4 5
P (
INH
B >
0)
0.0
0.2
0.4
0.6
0.8
1
.0
0
λi
ii
ΔCΔEP0INHBP
13
Summary
• Proposed an economic approach to measuring efficiency using a composite measure.
• Theoretically strong and objective weighting mechanism.
• Results will differ from threshold model due to ability to incorporate tradeoffs.
• Difficult to agree on single WTP value.– LY and QALY measures of benefit are more
promising.
14
Future work
• Longitudinal analysis.• Inclusion of AMI process measures, quality of
life.• Developing willingness to pay values that reflect
multiple outputs (benefits).• Refining cost measure to include RVUs.• Exploring a composite measure of hospital
efficiency.