Title: Evaluating the Impact of Treatment Effectiveness and Sideeffects on Prescription Decisions: The Role
1Evaluating the Impact of Treatment Effectiveness
and Side-effects on Prescription Decisions The
Role of Learning and Patient Heterogeneity
- Tat Chan
- Chakravarthi Narasimhan
- Ying Xie
- Washington University in Saint Louis
- May 11, 2007
- Yale Conference on Consumer Insights
2Medicines are never harmless!
- In September 2004, Merck Co. (MRK ) pulled its
blockbuster painkiller Vioxx from the market
because a study linked it to heart attacks and
strokes. - All drugs have risk does the benefit exceed the
risks? - How would physician and patient trade off the
benefits (effectiveness of a treatment) and the
risks (side effects) of a drug?
3Conceptual Framework
effectiveness
switch/stay
detailing
Physician Patient
switching reasons
Side effects
Prescription choice
Patient feedback
First prescription
cost
4Identifying Effectiveness and Side Effects
- Two data sources
- Market outcomes treatment choice for each
physician-patient pair - Self-reported reasons for switching treatments
ineffective or side effects - Manski (2004), BLP (2004)
- (1) infers the overall evaluation of treatments
- (2) identifies the treatment effectiveness and
side effects across treatments as well as the
heterogeneity in their impacts - Model Match of market outcomes and self-reported
switching reasons to identify the impacts of
effectiveness and side effects on prescription
choices
5Data
- Prescription and Promotion Data available for a
physician panel for the Erectile Dysfunction
Category from Sept. 2003 to Oct. 2004 - Two new drugs launched during this period
(Levitra in August 2003, and Cialis in November
2003) - 828 physicians, 13619 prescriptions
- 7324 prescriptions to revisiting patients, with
1652 switched treatment - Reported reasons for switching prescriptions, and
to which other drugs these patients switched - Observed promotional activities
- 26509 detailing visits 4648 detailing visit with
a meal - Observed patient characteristics severity,
insurance status, ethnicity, age
6Switching Reasons
- Among 1652 visits where the revisiting patients
switched the drug treatments - 929 ineffectiveness
- 161 side effects
- 365 patient request
- 198 other reasons such as marketing
- Assumptions
- Effectiveness and side effects of previous drug
known by patients and physicians - Model the probability of switching due to
ineffectiveness vs. side effects - Patient request or other reasons either due
to ineffectiveness or side effects
7The Model
- Physician i will choose the alternative drug j
that provides the highest expected utility for
patient h at occasion t conditional on the
physicians information ,
Past detailing, Past patient feedback
- Ongoing patient who used j before same drug
prescribed if - Otherwise switch treatment
8The Model
- Physician i will choose the alternative drug j
that provides the highest expected utility for
patient h at occasion t conditional on the
physicians information ,
effectiveness
- Ongoing patient who used j before same drug
prescribed if - Otherwise switch treatment
9The Model
- Physician i will choose the alternative drug j
that provides the highest expected utility for
patient h at occasion t conditional on the
physicians information ,
Side effects
- Ongoing patient who used j before same drug
prescribed if - Otherwise switch treatment
10The Model
- Physician i will choose the alternative drug j
that provides the highest expected utility for
patient h at occasion t conditional on the
physicians information ,
Patient characteristics detailing (persuasive)
- Ongoing patient who used j before same drug
prescribed if - Otherwise switch treatment
11The Model
- Physician i will choose the alternative drug j
that provides the highest expected utility for
patient h at occasion t conditional on the
physicians information ,
- Ongoing patient who used j before same drug
prescribed if - Otherwise switch treatment
Switching cost
12Model Specification
- Model Assumption
- Heterogeneity of drug effectiveness and side
effects across patients - Effectiveness and side effects may be correlated
across drugs -
13Model Specification switching reasons
- Suppose a revisiting patient h switches from drug
j to drug k on occasion t, and states side
effects as the switching reason, two additional
condition hold in this case
i.
ii.
We then estimate the joint probability of i, ii,
and iii
14Model Specification Learning
- As in Erdem and Keane (1996)
- Prior Belief
- Learning from two sources
- Detailing
- Feedback from Ongoing Patients
- Bayesian Learning
15Preliminary Results effectiveness and side
effects
Switching cost 2.6687
16Preliminary Results persuasive detailing
17Preliminary Results Patient characteristics
- Cialis is more likely to be prescribed to older
patients - Cialis is more likely to be prescribed to white
- Levitra is less likely to be prescribed to
African American, and more likely to white - Levitra is more likely to be prescribed to
patients with Medicare
18Summary
- Tease out the effect of treatment effectiveness
and side effects on prescription choice - The next step to identify the role of patient
feedback and detailing in driving physicians
learning of treatment effectiveness and side
effects
19Policy Experiments
- Competitive relationship across drugs
- Own- and cross-elasticities of prescription share
due to change of detailing effort - How much due to patient-physician evaluation of
effectiveness vs. side effects? - How much due to heterogeneity in effectiveness
and side effects as well as their correlations
across drugs? - Time-varying informative role of detailing
- Through both detailing and own prescription
experience, physicians learn more about
effectiveness and side effects of new drugs - Informative role of detailing may decline
overtime - Can that explain the declining detailing efforts
observed in data? - What is the optimal detailing strategy at the
physician level overtime?
20 21Prescription Choice
Detailing
Prescription Choice
Physician Influence
Learning Effectiveness Side effects
Effectiveness
Patient Influence
Side effects
Out of pocket cost
22Research Objective
- The Role of Treatment Effectiveness and Side
Effects - How does the heterogeneity of treatment
effectiveness and side effects across patients
affects treatment choices? - The Role of Detailing
- Persuasive function detailing, detailing with
meals - Informative function
- Uncertainty of effectiveness and side effects of
new drugs - detailing, detailing with meals
- Narayanan et al(2006)
- How does detailing perform the persuasive and
informative functions? - How does detailing affect competitive
relationships overtime?
23Prescription Trend
24Detailing Trend
25Model Specification
- Mean Effectiveness (by severity) and Side
Effects
- Model Assumption
- Heterogeneity of drug effectiveness and side
effects across patients - Effectiveness and side effects may be correlated
across drugs -
26Model Specification switching reasons
- Suppose a revisiting patient h switches from drug
j to drug k on occasion t, and states side
effects as the switching reason, two additional
condition hold in this case
i.
ii.
Therefore, we estimate the following joint
probability
which can be expressed as conditional probability
as follows
27Preliminary Results effectiveness and side
effects
28Preliminary Results without learning