Transcript The impact of information on patient preferences in

The impact of information on patient preferences in different delivery patterns:
a contingent valuation study of prescription versus OTC
Karine Lamiraud*, Konrade von Bremen*, Cam Donaldson**
*University of Lausanne, Institute of Health Economics and Management (IEMS), DEEP-HEC
**Institute of Health and Society, Newcastle University, UK
Toulouse, January 25 – 26 2008
INTRODUCTION
1. Background
1. CV
methods
 Increasing interest in the use of CV for measuring WTP
2. Major
concern
Influence of additional information
(Tversky & Kahneman, 1981)
 Empirical studies (Donaldson & Schakley 1997; Lee et al 1998; Protière et al 2004)
3. Impact of
information
 Information about health attributes of the programmes
 Inconsistent effects of additional information on WTP values
The impact of the level of information on WTP may depend upon the delivery
mechanism. This has not been assessed yet.
Toulouse, January 25 – 26 2008
INTRODUCTION
2. Objectives
 This study aims at assessing the impact of
information on preferences in different delivery
patterns (prescription versus OTC systems)
Toulouse, January 25 – 26 2008
INTRODUCTION
3. Policy relevance
 This investigation is driven by the fact that many countries have
expanded efforts to move prescribed drugs onto OTC status
 Previous works have focused on assessing the change in benefits
using a consumer surplus approach based on observed demand
curves (Ryan and Yule, 1990; Temin, 1983; Shih et al., 2002)
 However, the methods employed were not able to test the susceptibility
of benefits to the level of information
 A CV approach is used
Toulouse, January 25 – 26 2008
Outline
 Theoretical background

WTP study

Empirical analysis

Results

Conclusion
Toulouse, January 25 – 26 2008
THEORETICAL FRAMEWORK
1. Theoretical framework
Two simple uses of well-known economic theories help us derive two hypotheses
 Agency and Information
 “information hypothesis”
 Demand theory and the impact of moving to OTC
“switching hypothesis”
Toulouse, January 25 – 26 2008
THEORETICAL FRAMEWORK
2. Agency and Information
 When visiting a doctor, the patient may mostly
rely on the doctor’s knowledge
 Opting for an OTC medicine requires full
individual choice and self-administration
« Information assumption »
 The impact of information on WTP will be more significant in the OTC scenario
than in the prescription one
Toulouse, January 25 – 26 2008
THEORETICAL FRAMEWORK
3. Demand theory
 The figure depicts the demand curve for the case of a hypothetical drug
The cost of the drug is assumed to be fully borne by the consumer
 We assume that POTC < Pp (Ryan and Yule, 1990)
Cost
A
Pp
B
POTC
O
C
Qp
QOTC
Number of
consumers
Toulouse, January 25 – 26 2008
THEORETICAL FRAMEWORK
4. « Switching assumption »
(S1) the total WTP is expected to be higher in the OTC scenario than in the prescription one
(S2) more “0” answers are to be expected in the prescription scenario
(S3) if zero “answers” are excluded WTP is expected to be higher in the prescription scenario
Cost
A
Pp
B
POTC
O
C
Qp
QOTC
Number of
consumers
Toulouse, January 25 – 26 2008
THEORETICAL FRAMEWORK
5. A test of the validity of the CV approach
 Through basing our analyses on the
assumptions arising from these theories, a
useful (and new) test of the validity of
the CV approach is thus provided
Toulouse, January 25 – 26 2008
DATA
1. WTP Study
These effects will be assessed within a study measuring patient preferences
concerning a new class of drugs in influenza disease
Neuraminidase Inhibitors (NAIs)
 They offer symptom reduction by 1.5 days
 Since their release, NAIs have been prescription drugs
 Due to benign side effects, they might also qualify for OTC
 A CV study was performed to compare preferences between both delivery systems. It took
place in the French speaking part of Switzerland (winter 2001, summer 2001)
Toulouse, January 25 – 26 2008
DATA
2. General information to all study participants
All participants were first provided with some information on influenza as it was
considered that respondents should have a comparable understanding of the disease
 Influenza is in general a benign disease
 Self-limited to 5-7 days
 Symptoms = fever cough and runny nose with sudden appearance
 Risk of complications for children, the old and frail, and chronically ill
Vaccination is available
Vaccination does not give a complete protection against influenza.
Big economic impact due to influenza
Pandemic outbreak can be disastrous to the country
Toulouse, January 25 – 26 2008
DATA
3. Testing for informational effect
Respondents were randomly assigned into basic or extended information
Basic Information
Extended Information
 able to reduce symptom by 1.5 days
 able to reduce symptom by 1.5 days
 needs to be taken within 48 hours
after onset of symptoms
 needs to be taken within 48 hours after
onset of symptoms
 side effects are very mild
 side effects are very mild
 not able to prevent/cure the flue
 not active against cold
 does not replace vaccination
 is not active out of the period of
influenza epidemics of about 4 months
Toulouse, January 25 – 26 2008
DATA
4. WTP questionnaires
After being provided with some information all respondents were asked to imagine that
they were affected by influenza and to answer 2 WTP questions framed into 2 scenarios
 First scenario
 How much are you willing to pay for a
prescription drug which is able to reduce the
duration of the flue by 1.5 days?
This drug is not reimbursed, the consultation
would need to be paid for
 Second scenario
 How much are you willing to pay for an OTC
drug which is able to reduce the duration of the
flue by 1.5 days?
This drug can be bought without restriction in
the pharmacy
Toulouse, January 25 – 26 2008
DATA
5. Bidding ranges
A payment card system was used to facilitate answers
Frs 0
Frs. 1 - 20
Frs. 21 - 40
Frs. 41 - 60
Frs. 61 –80
Frs. 81 - 100
Frs. 101 –150
Frs. 151 - 250
Frs. 251 – 500. Frs. 501- 1'000
More than Frs.1'000
 To exclude starting point bias the order of ranges was randomized (increasing or decreasing)
Toulouse, January 25 – 26 2008
Outline
 Theoretical background

WTP study

Empirical analysis

Results

Conclusion
Toulouse, January 25 – 26 2008
EMPIRICAL ANALYSIS
1. Overview of empirical strategy
The empirical analysis aims at testing the information and switching
hypotheses. We conducted 3 main analyses.
 We examined the factors associated with WTP :
- information level in the OTC and prescription scenario (I)
- scenario (S1)
 We examined consistency with S2 via the distribution of zero
answers between both the OTC and prescription scenarios
 We investigated the distribution of non zero answers between the
OTC and the prescription scenario in order to test S3
Toulouse, January 25 – 26 2008
EMPIRICAL ANALYSIS
2. Econometric model
logY  SOTC   (S p * Ii ) 1  (SOTC * Ii ) 2  X i   ui  vit
*
it
i
The individual
t
The scenario (s = prescription vs OTC)
Yis*
The respondent’s true valuation for scenario t
Ii
X it
Information level (I = 0,1 ie extended vs basic)
 1   2 would be consistent with the “information” assumption
on S1,  is expected to be positive and significant
 A finding that
 Based
Explanatory factors (socio-demographic and health status variables….)
Toulouse, January 25 – 26 2008
EMPIRICAL ANALYSIS
3. Selection of covariates
The selection of covariates included testing whether covariates had a different
impact on WTP for the prescription and OTC drugs
logYit  SOTC   1 (S p * I i )   2 (SOTC * I i )  1 X1S p  ...   n X n S p  1 X1SOTC  ...   n X n SOTC  ui  vit
*
The following equality tests were then performed
1   1
2  2
j j
Toulouse, January 25 – 26 2008
EMPIRICAL ANALYSIS
4. Econometric estimation
We fitted a random-effects interval censored regression model
logYis*  S  (S0 * Ii ) 1  (S1 * Ii ) 2  X is   ui  vis
 Answers lie in intervals and are not point estimates The respondent’s true valuation known
to lie within the interval defined by lower and upper thresholds
 Answers might be correlated for a given individual
Assumptions
 Zero answers can be considered as very small WTP (]0,1])
 A lognormal conditional distribution for valuations is proposed **
log
Yis*lie between log ta and log tb except for the first and last intervals
*Cameron & James, 1986 ; * *Cameron & Huppert, 1989
Toulouse, January 25 – 26 2008
RESULTS
1. The study population
1594 subjects were enrolled in the study
Information level
Basic
Extended p
Females (%)
55.8%
55.9%
55.8%
0.7
Age (mean, std)
35.4 ±12.1
35.5
35.4
0.7
Mandatory school (%)
6.4%
6.5%
6.3%
0.43
High school (%)
10.8%
11.5%
10.1%
Skilled training (%)
49.7%
50.0%
49.5%
University (%)
33.0%
32.0%
34.1%
Working (%)
90.3%
90.1%
90.4%
Physician Health care professional (%)
14.6%
13.1%
16.0%
Non physician Health care professional (%)
52.0%
53.5%
50.6%
Supplementary insurance (%)
49.4%
48.4%
51.4%
Perfect subjective health status (%)
74.0%
74.4%
73.8%
0.80
Vaccinated against flue (%)
Administered in winter time (%)
29.9%
29.9%
29.6%
0.90
56.2%
55.9%
56.4%
0.80
Basic information level (%)
49.5%
0.6
0.1
Toulouse, January 25 – 26 2008
RESULTS
2. WTP descriptive statistics
Willingness to Pay for influenza symptom reduction by 1.5 days
Prescription scenario
OTC scenario
100%
100%
80%
76%
64%
62%
38%
36%
24%
0%
20%
0%
WTP > 0
WT0 = 0
Basic
WTP > 0
WT0 = 0
Extended
p (basic vs extended) = 0.24 (chi2)
WTP > 0
WT0 = 0
Basic
WTP > 0
WT0 = 0
Extended
p (basic vs extended) = 0.09 (chi2)
 People receiving extended information were more willing to pay for the drug
However this tendancy is significant in the OTC scenario only (at a 10% level)
Toulouse, January 25 – 26 2008
RESULTS
3. Econometric results
Panel interval regression estimations over the full range of responses
OTC scenario
Limited information in the prescription scenario
Limited information in the OTC scenario
Ascending ranges in the prescription scenario
Ascending ranges in the OTC scenario
Questionnaire filled in during winter time
Male gender
Age (<24, (25,64), >65)
High school*
Skilled training*
University*
respondent having a job
Senior health care professional in the OTC scenario**
Non Senior health care professional in the OTC scenario**
Senior health care professional ** in the prescription scenario**
Non senior health care professional in the prescription scenario**
Basis + supplementary insurance
Subjective health status: perfect
Suffering from chronic disease
Affected by influenza during the past two years
Vaccinated against influenza
constant
sigma_u
rho
*
mandatory schooling = reference , ** lay people = reference
Coef
0.26
-0.12
-0.22
-0.16
0.03
0.37
0.05
-0.43
0.88
0.73
0.96
0.27
-0.32
-0.70
-0.67
-0.70
-0.02
-0.15
0.23
0.24
0.58
1.79
2.02
0.72
p
0.02
0.36
0.04
0.22
0.79
<0.01
0.73
<0.01
<0.01
<0.01
<0.01
0.243
0.16
<0.01
<0.01
<0.01
0.84
0.30
0.35
0.08
<0.01
<0.01
***
Toulouse, January 25 – 26 2008
RESULTS
4. Information impact
An increased level of
information:
-pushes the WTP to upper
levels in the OTC scenario
- has no impact on the WTP for
prescription medicine
 When asking for medical advice, the patient relies on the doctor’s knowledge
 When opting for OTC, the patient requires relevant information to make informed choices
This is in line with the “information assumption”
Toulouse, January 25 – 26 2008
RESULTS
5. Other results
Delivery patterns
 the WTP increases when the
drug is delivered OTC (S1)
the WTP
increases
 when the participant answers the
survey during the winter
 with incomes (assuming education
level is a proxy for income)
 if respondents are risk averse (as
expressed by being vaccinated)
 the insurance coverage does not
influence the WTP
Toulouse, January 25 – 26 2008
RESULTS
6. Discussion (1)
We must discuss the possibility that some people might have thought that the
OTC drug had to be sold at a higher price for some reason
 misunderstanding ?
some costs can be avoided ?
It is unlikely that the
seriousness of the disease
has been underestimated
However, physicians value the
OTC drug (not the prescription one)
as much as the general population
does
(phsyisicans, however, do not face
the same issues as time or money
to get a prescription)
Toulouse, January 25 – 26 2008
RESULTS
9. Discussion (2): a sequence effect?
The prescription form of the drug could have benefitted from a sequence effect (WTP
is expected to be larger for the first good of a series of good, Payne et al 2000)
Our methodology does not allow us to assess whether a sequence effect is present or not
 However we claim that our findings are not affected by a sequence effect if there is one
 If such an effect were present, our current results suggesting that the OTC version is
preferred would even be reinforced
Toulouse, January 25 – 26 2008
RESULTS
8. Discussion (3): protest zeros?
We also checked whether some « 0 » answers could be attributable to protest zeros*
A probit model was estimated on the subgroup of those who answered “0” to both scenarios
 A bivariate probit model was run on the propensities to pay for a prescription and OTC drug
Some people may not feel concerned or express aversion to contributing to a publicly funded health care system
Toulouse, January 25 – 26 2008
RESULTS
8. Discussion (4): protest zeros?
Probit model explaining zero answers to both scenarios
Coef
p
Limited information
.073
0.344
Ascending ranges
-0.05
0.511
Questionnaire filled in during winter time
-0.18
0.035
Male gender
-0.01
0.885
Age (<24, (25,64), >65)
0.30
0.006
High school
-0.45
0.018
Skilled training
-0.43
0.005
University
-0.38
0.023
Health care professional
0.28
0.004
Basis + supplementary insurance
0.02
0.801
Subjective health status: perfect
0.06
0.490
Suffering from chronic disease
-0.08
0.625
Vaccinated against influenza
-0.22
0.019
Personal history of influenza over the past 2 yrs
-0.21
0.025
constant
-1.08
0.000
 People declaring zero values to both scenarios cannot be regarded as protesters
Toulouse, January 25 – 26 2008
CONCLUSION
1. Policy issue
Our results:
Welfare gains are associated with switching to OTC
The benefits of switching to OTC status are likely to depend on the
information level
For the kind of drug considered here (safe and easy to monitor), the
requirement for patient decision making is the provision of comprehensive
information
Toulouse, January 25 – 26 2008
CONCLUSION
2. Validity of the CV approach

One important contribution of the study was to use the CV method to assess the benefits
from OTC while previous studies were based on market behaviors of actual
consumers

We provide a useful test of the validity of the CV approach given that all the results
come out in line with predictions from the agency and demand theories