Transcript New Incentive Approaches For Adherence

New Incentive Approaches to
Adherence
Stephen Kimmel, MD, MSCE
Andrea Troxel, ScD
Center for Clinical Epidemiology and
Biostatistics
The Importance of Clinical Trials to
Study Adherence Interventions
• Observational studies prone to bias
– Patient selection
– Temporal trends
– Unmeasured confounders
• Effects of intervention may not be “symmetrical”
– e.g., increasing copayments for medications may not
have the same effect as reducing copayments
• Logistics
– You can learn a lot just from developing an
intervention
Adherence worsens among those facing copayment
increases…but also among patients without copays
Source: Doshi, Zhu, Kimmel, Volpp et al. Circulation
Reducing co-pays appears to have modest
effects in quasi-experimental studies
Source: Chernew ME, et al. Health Affairs 2008;27(1):103-12.
$5  $0 for generics
$25  $12.50 for preferred drugs,
$45  22.50 for non-preferred drugs
4
But, Randomized Trial of Copay
Reduction: No Impact. . .
•RCT among veterans with poorly
controlled blood pressure
•Average SBP on study entry: 160
•Copayments reduced from $8 per
month to $0
• No significant effect on blood
pressure or medication adherence in
either study
Source: Volpp, Troxel, Kimmel , Doshi et al. 2010
Novel Approaches to Adherence
• Framework: Behavioral Economics
• Most commonly used approaches
– Financial incentives
• patients
• clinicians
– Reminders
– Pre-commitment mechanisms
Adherence to Warfarin
% Population
40% of patients were
>20% non-adherent
Either missed or extra dose
Underadherence by Under-AC
Odds Ratio (CI)
P<0.0001
% Missed Pills
Reference group: No missed pills
Kimmel et al. Arch Intern Med. 2007;167:229-235.
Adherence Changes Over Time
Percent of Days Non-Adherent
30%
20%
10%
0%
1
3
6
9
Month After Beginning Warfarin Use
12
Lottery-Based Incentives: Conceptual
Framework
• Deliver frequent feedback and rewards – ideally at
the daily level, because of present-biased
preferences
• Lotteries give more bang for the buck, in part
because people overweight small probabilities
• Give frequent positive feedback plus the hope of big
payout
– Take advantage of time discounting
– Use variable reinforcement
• Utilize regret as a motivator
Lottery-based financial incentive based on daily
medication use
• Eligible for daily lottery prize if the medication monitoring
device has registered that patient took medication the
previous day
• If patient did not take medication s/he will be informed that
s/he would have won the lottery that day had s/he taken the
medication properly
• High probability small reward & low probability big reward
• Takes advantage of present-biased preferences (immediacy of
receiving coupon/gift card), overestimation of probabilities,
regret aversion, and variable reinforcement
Incentives for medication adherence
using daily lotteries
– Warfarin is an anti-stroke
medication with large
benefits but non-adherence
rates high (chronic,
asymptomatic condition)
– Designed lottery with 1 in 5
chance of winning $10 a
day, 1 in 100 chance of
winning $100 each day IF
took warfarin previous day
Volpp, Loewenstein, Troxel, Doshi & Kimmel, BMC
Health Services Research, 2008
12
The Warfarin INcentives (WIN) Pilot Trial
P<0.001
P=0.4
P-value for
interaction=0.0016
WIN2 Trial
WIN2 Trial Design
• Balanced randomization, stratified by
– Baseline INR status
– Study site
• Five comparisons of interest
– Each active arm against control (3)
– Combined arm against each single arm (2)
– Bonferroni adjustment of alpha = 0.01
WIN2 Power/Sample Size
• Binary outcome: out-of-range (OOR) INR
• Detect a 30% reduction in occurrence of OOR
INRs from 0.43 to 0.3
– Assume 10% drop-out by six months
– Assume ICC of 0.05
– Assume 7 INR assessments per subject
WIN2 Analysis Plan
• Generalized linear mixed models
– Logit link
– Fixed effects for time, treatment, other timeinvariant covariates
– Random intercepts and slopes
– Exploratory modeling of shape of time effect
– Time x treatment interactions
– Explore adjustment for frequency of assessment
Segmenting or Tailoring
• How can we improve effectiveness (and costeffectiveness) of these interventions?
• Should we (can we) only target non-adherent
patients?
• Can we tailor our interventions?
- Upfront screening of RCT patients at baseline for
reason for non-adherence with tailored
intervention
- Sequential Multiple Assignment Randomized
Trials (SMART)
Tailoring
• Evaluate subjects at baseline regarding drivers of
non-adherence (cost, distraction/forgetfulness,
knowledge/beliefs/attitudes)
• Develop strategy tailored to each driver (copay
reduction, lottery-based reminder,
education/instruction)
• Consider randomizing “tailored strategy” vs “onesize-fits-all strategy” vs “usual care”
SMART Designs
Sequential Multiple Assignment Randomized Trials
• Work by Murphy, Robins, Thall, and many others
• Optimize response by varying interventions on the
basis of time-dependent information
- In cancer therapy, randomize patients to initial therapy;
then randomize responders to maintenance therapy and
non-responders to second-line therapy
- In adherence, randomize subjects to primary strategy;
then maintain adherent subjects on same strategy and
randomize non-adherent subjects to remaining options
Summary
• Use ideas from behavioral economics to inform
interventions
• Use appropriate methodology from clinical trials to
properly assess hypotheses of interest
• Promising area for further development of adaptive
treatment and tailoring strategies
• LDI Center for Health Incentives (Kevin Volpp,
Director) is leading multiple trials in diverse areas of
behavioral health