Do intermediate outcome
measures of quality provide
incentives for inefficient care?
Justin W. Timbie, Ph.D.
Rodney A. Hayward, M.D.
Sandeep Vijan, M.D.
Ann Arbor VA HSR&D Center of Excellence
Funding from the QUERI program (QUERI DIB 98-001) and the
Measurement Core of the Michigan Diabetes research & Training Center,
NIDDK (P60 DK-20572), and VA HSR&D IIR 06-253 is gratefully
acknowledged.
1
Background
• Diabetes practice guidelines recommend treating
cardiovascular risk factors to low targets (TTT).
– LDL <100 mg/dl, A1c <7%, BP <130/80
• Patient characteristics are not incorporated into
guidelines.
• Recent trials remind us that treatments can be
risky.
2
Background
• Risk-benefit framework could help mitigate
potential harm from treating to targets.
• Some effect modifiers of TTT are known:
– Treatment effects are NOT additive.
– Treatment harms increase with greater polypharmacy.
– CVD risk varies among diabetics.
3
Objective
• To simulate the benefits of treating to LDL and
BP targets.
• To assess variation in benefit of TTT according
to:
1) baseline CVD risk
2) individual treatments
4
Data
• National Health and Nutrition Examination
Survey (NHANES-III), 1988-1994
– Risk factor distributions, prior complications, current
treatments.
– Relatively untreated population
• Meta analyses of RCTs for model parameters.
5
Methods
• Monte Carlo simulation
– Individual patients underwent intensification
– Risk factor reductions, side effects, and
discontinuation simulated
• Markov model
– Estimate QALYs gained from LDL and blood pressure
reduction.
– 10 year time horizon.
6
Methods: Treatments
LDL target: 100mg/dl
BP target: 130/80
SMV 20
Thiazide
SMV 40
ACE
ATV 40
BB
ATV 80
CCB
SMV80/EZE
CCB2
7
Methods: Treatment-related harms
• Side effects (0.005 disutility each):
– Statin: myalgia (0.01 disutility)
– Antihypertensives: 9-17 side effects
• Polypharmacy
– Inconvenience and drug interactions
– 0.001, 0.002, 0.003, 0.004 disutility for first 4 classes
8
Results: Attaining tight control
Baseline
treatment
LDL
BP
Prevalence
(%)
Risk factor
reduction
Tight control
(%)
None
90
-54.9 mg/dl
78.2
Low-dose statin
4
-47.0
50.5
None
43
-16.1 / -6.6 mmHg
78.2
1 class
32
-15.0 / -4.9
56.8
2 classes
22
-12.2 / -3.6
27.4
3 classes
4
-8.7 / -2.1
19.2
4 classes
1
-4.4 / -1.2
0.3
9
600
500
700
Results: Baseline risk distribution
300
200
300
400
Frequency
500
400
Blood pressure
100
200
0
100
0
Frequency
LDL
0
1
2
3
4
5
0
2
QALYs at risk from poor control
4
6
8
10
Results: QALYs gained by TTT
LDL
Treatment harm
Net benefit
Blood pressure
Treatment harm
Net benefit
11
Benefit by treatment step (BP)
Lowest risk
Highest risk
Treatment:
THI
ACE
BB
CCB
THI2
ACE2
BB2
CCB2
12
Summary of findings
• 25 to 40% of subjects failed to achieve tight
control.
• Benefit is limited to a small subset of the
population.
• Accounting for treatment-related harms can
identify inappropriate intensification.
13
Implications for clinical practice and
quality measurement
• Intensifying therapy based on tolerability ignores
magnitude of benefit.
• Treatment risks might be underweighted.
• Intermediate outcome measures might
encourage use of treatments with low marginal
benefit.
• Need decision support to determine net benefit
of intensifying treatment.
14
15
Benefit by treatment step (LDL)
Treatment:
SMV20
SMV40
ATV40
ATV80
% Treated:
95 96
27 82
19 74
8 53
SMV/EZE
6 44
16
Methods: Treatments
LDL target: 100mg/dl
BP target: 130/80
-32%
SMV 20
Thiazide
-5.7%
-8%
SMV 40
ACE
-4.6%
-20%
ATV 40
BB
-4.2%
-12%
ATV 80
CCB
-3.4%
-6%
SMV80/EZE
CCB2
-0.6% -1.2%
17
Limitations
• No standard treatment regimen exists.
• Additional patient characteristics needed for
blood pressure simulation.
• Limited data for certain model parameters.
– Efficacy of combination therapy
18